| # | Strategy | Type | Effort | What you need first |
|---|---|---|---|---|
| 1 | First-party data and privacy-first targeting | Targeting | High | CRM records and consent framework |
| 2 | AI and predictive analytics | Optimization | Medium | Campaign history to train on |
| 3 | Programmatic buying in CTV | Buying | Low | DSP access and supply-path visibility |
| 4 | Viewability and attention metrics | Measurement | Low | An attention measurement vendor |
| 5 | CDPs and data clean rooms | Data infrastructure | High | Unified customer records |
| 6 | Dynamic creative optimization | Creative | Medium | Modular creative assets |
| 7 | Content personalization | Creative | Medium | Defined audience segments |
| 8 | Interactive and shoppable formats | Format | Medium | Platform support and a commerce path |
| 9 | A/B testing and experimentation | Measurement | Low | A holdout designed before launch |
| 10 | Cross-device targeting | Targeting | Medium | An identity graph |
| 11 | Sequential storytelling | Creative | Medium | Three or more creatives and frequency control |
| 12 | Contextual targeting in streaming | Targeting | Low | Nothing—works without user data |
| 13 | Frequency capping and pod management | Delivery | Low | Caps that operate across platforms |
Fig. The 13 strategies compared by type, implementation effort, and prerequisite.
| Business Type | Why Retargeting Works | High-Value Retargeting Audiences |
|---|---|---|
| eCommerce | Shoppers frequently browse products or abandon carts before purchasing | Product viewers, cart abandoners, previous buyers |
| B2B | Long decision cycles create multiple opportunities to re-engage potential buyers | Pricing-page visitors, content leads, demo visitors |
| SaaS | Prospects often compare solutions before starting a trial or subscription | Trial users, pricing-page visitors, inactive leads |
| Retail | Retargeting connects product discovery with online or offline purchase opportunities | Product/category visitors, loyalty customers |
| Travel | Customers typically research destinations, hotels, and flights across several sessions | Destination researchers, booking abandoners |
| Finance | High-consideration products often require research before conversion | Product-page visitors, application starters |
| Healthcare | Retargeting can support longer consideration journeys where permitted by privacy and advertising rules | Eligible informational or service-page audiences |
| Subscription businesses | Repeat revenue makes retention and re-engagement especially valuable | Existing customers, lapsed subscribers, renewal audiences |
| Attribution | Incrementality | |
|---|---|---|
| Core question | Which marketing touchpoint should receive credit for the conversion? | Would the conversion have happened without the advertising? |
| What it measures | How conversion credit is assigned across ads, channels, and touchpoints | The additional conversions or revenue actually caused by advertising |
| Common methods | Last-click, first-click, linear, position-based, and data-driven attribution | Holdout groups, control groups, geo experiments, and conversion lift tests |
| Main limitation | Can credit advertising for conversions that may have happened anyway | Requires controlled testing and sufficient audience volume |
| Value for retargeting | Shows which retargeting touchpoints are associated with conversions | Shows whether retargeting generated genuine incremental lift |
| Best use | Understanding the customer journey and allocating conversion credit | Evaluating causal impact and deciding whether additional spend creates additional business results |
| Performance Metric | Retargeting vs. Standard/Prospecting Targeting | What the Evidence Shows |
|---|---|---|
| CTR | Higher | LinkedIn reported +30% CTR for website retargeting; Jabra reported 2–3x higher CTR |
| Conversion rate | Higher among previously engaged audiences | TOPdesk reported a 20% increase after implementing remarketing |
| Relative CPA | Lower | LinkedIn reported 14% lower post-click cost per conversion; TOPdesk reported a 24% decrease |
| Symptom | Likely cause | Where to start |
|---|---|---|
| Creative rejected at publisher QA | Media file specifications outside the accepted range | Creative or production team, against the publisher's spec sheet |
| Bids won but few impressions recorded | Wrapper chain timeout or media file incompatibility | Ad ops, with the DSP and third-party ad server |
| Ad serves but reporting shows no impressions | Broken, expired, or non-secure impression tracking URL | Third-party ad server or verification vendor |
| Quartile data missing or partial | Tracking events omitted from the tag, or SSAI relay failure | Ad server first, then the publisher's SSAI vendor |
| Delivery fails on one device family only | Codec or player compatibility limited to that platform | Publisher or CTV platform, with device-level test evidence |
| Impression counts inflated or duplicated | Unresolved cachebusting macro | DSP or SSAI vendor |
| Frequency capping ineffective across publishers | Missing or inconsistent UniversalAdId | Advertiser's ad server configuration |
Fig. Common VAST failures and where they usually originate.
| Standard | Primary purpose | Relationship to VAST | Status in CTV |
|---|---|---|---|
| VAST | Delivers ad metadata, media file references, and tracking URLs from ad server to player | The foundation; all others extend or sit alongside it | Required. VAST 4.x is the CTV expectation |
| VPAID | Enabled interactive video ads and in-ad measurement by executing code in the player | Delivered inside a VAST response as an executable media file | Deprecated. Largely unsupported by CTV players |
| VMAP | Describes the ad break schedule for a piece of content: where breaks fall and how many ads each holds | A separate XML document referencing VAST responses for each slot | Used by publishers and SSAI vendors to structure pods |
| SIMID | Provides interactive ad experiences through a separate creative layer rather than code in the player | Introduced in VAST 4.2 as the designated VPAID replacement for interactivity | Emerging; adoption uneven across CTV platforms |
| OMID | Standardizes viewability and verification measurement through the Open Measurement SDK | Referenced by VAST 4.x through dedicated verification nodes | The established route for third-party CTV verification |
Fig. How VAST relates to VPAID, VMAP, SIMID, and OMID.
| Element | What it holds | What breaks if it is wrong |
|---|---|---|
| AdSystem | Name and version of the ad server that generated the response | Responses attributed to the wrong system during troubleshooting |
| UniversalAdId | A persistent creative identifier, typically an Ad-ID registry value | Frequency capping and cross-platform reporting fragment, because one spot is counted as several ads |
| Impression | The URL fired when the ad begins rendering; the billable event | Discrepancies between platform-reported and third-party impression counts |
| MediaFiles | One or more encodings of the creative, each with type, bitrate, and dimensions | The player finds no compatible file and abandons the slot |
| TrackingEvents | URLs for start, quartile, completion, mute, pause, and other events | Completion rates and quartile curves are missing or partial |
| VideoClicks | Clickthrough destination and click tracking URLs | Interaction reporting fails; on CTV this is often QR or remote-triggered |
| Duration | The declared length of the creative | Pod slot mismatches, causing the ad to be dropped from the break |
Fig. Core elements of a VAST response and what each controls.
| Campaign goal or condition | Best approach | Why |
|---|---|---|
| Brand awareness | CTV or both | CTV creates large-screen impact; OLV extends reach |
| Product launch | Both | CTV introduces the story; OLV reinforces details and drives action |
| Website traffic | OLV | Clickable ads provide a direct route to the website |
| Retargeting | OLV | User- and device-level signals support precise follow-up |
| Lead generation | OLV | Interactive formats shorten the path to form completion |
| Limited budget | OLV | Lower entry costs and flexible testing support efficient scaling |
| Premium audience targeting | CTV | Publisher and first-party data support household targeting in premium content |
| Cord-cutting audiences | CTV | Reaches households through internet-connected television screens |
| Rapid creative testing | OLV | Multiple formats and faster performance signals simplify optimisation |
| National brand campaign | CTV or both | CTV delivers impact, while OLV expands cross-device coverage |
| Omnichannel campaign | Both | Connects household awareness with personal-device engagement and conversion |
| Full-funnel growth | Both | Assigns distinct awareness and performance roles to each channel |
| Marketing objective | Best-suited format | Strategic rationale |
|---|---|---|
| Brand awareness | CTV or both | CTV creates high-impact exposure, while OLV adds cross-device reach |
| Product launches | Both | CTV introduces the story, while OLV reinforces product information |
| Website traffic | OLV | Clickable formats provide a direct route to the website |
| Retargeting | OLV or both | OLV supports user-level retargeting, while CTV reinforces intent at household level |
| Lead generation | OLV | Interactive ads and landing pages shorten the conversion path |
| Full-funnel campaigns | Both | CTV builds awareness, while OLV supports consideration and action |
| Comparison point | OLV | CTV |
|---|---|---|
| Delivery | Mobile, desktop, tablet, websites, apps, streaming platforms, and CTV | Smart TVs, streaming devices, and connected gaming consoles |
| Ad formats | Pre-roll, mid-roll, post-roll, outstream, in-feed, vertical, and bumper ads | Primarily full-screen pre-roll and mid-roll, plus pause and interactive ads |
| Skippability | Skippable or non-skippable, depending on the placement | Usually non-skippable within streaming content |
| Targeting | User-, account-, browser-, contextual-, and device-level signals | Primarily household-, device-, content-, and publisher-level signals |
| Typical CPM | Approximately $5–$30 | Approximately $20–$65 |
| Buying models | Open auctions, platform auctions, PMPs, direct deals, and reservations | Open auctions, PMPs, preferred deals, programmatic guaranteed, and direct reservations |
| Primary KPIs | Viewability, view rate, completion rate, clicks, conversions, CPA, and ROAS | Unique reach, incremental reach, frequency, completed views, brand lift, and sales lift |
| Term | What it describes | Where it fits |
|---|---|---|
| OTT | Delivery method | Internet-delivered video content accessed through streaming services and apps |
| OLV | Advertising format | Digital video ads delivered within online content across multiple devices |
| CTV | Device category | Internet-connected television screens on which OTT content and OLV ads are viewed |
| Industry | Primary suitability concern | Risk tolerance | Regulatory overlay | Typically acceptable | Typically unsuitable |
|---|---|---|---|---|---|
| Financial services | Trust, competence, association with loss or fraud | Low | Advertising disclosure rules, suitability of claims | Business and personal finance coverage, market analysis | Debt distress content, gambling, crypto speculation, insolvency reporting |
| Healthcare and pharma | Regulatory compliance, patient sensitivity, misinformation adjacency | Very low | Category-specific rules on claims, targeting and disclosure | Health education, condition-specific editorial, wellbeing content | Unverified health claims, alternative medicine, self-harm and eating disorder content |
| Luxury and premium | Prestige transfer, ad environment quality, exclusivity | Low on quality, moderate on topic | Limited | Design, culture, travel, high-end lifestyle publishers | High-density inventory, discount and coupon environments, MFA properties |
| Children's brands | Age appropriateness, parental scrutiny, adjacency to harm | Extremely low | Children's privacy and advertising restrictions | Family, education, parenting and age-rated entertainment | Violence and crime reporting, mature UGC, unrated user-generated video |
| Consumer goods | Reach efficiency, mild reputational exposure | Moderate to high | Limited | Broad entertainment, news, sport, lifestyle, mainstream UGC | Extremist content, explicit material, low-quality and synthetic inventory |
Fig. Suitability profiles by industry.
| Scenario | Safety priority | Suitability priority | Primary control | Common mistake |
|---|---|---|---|---|
| Illegal or harmful content | Absolute | Not applicable | Universal floor exclusions | Assuming platform defaults cover it |
| Breaking news | Baseline only | High and time-limited | Event-level contextual analysis | Blocking the news category permanently |
| Political content | Baseline only | High, brand-dependent | Category controls plus publisher tiers | Blanket exclusion of civic reporting |
| Children's products | Absolute | Very high | Age-appropriate allowlists, strict tiers | Relying on keyword lists for age suitability |
| Luxury and premium | Baseline only | Very high | Publisher quality and ad density criteria | Optimizing for CPM on prestige campaigns |
| Healthcare and pharma | Absolute | Very high, regulated | Regulatory controls plus contextual tiers | Treating regulatory approval as suitability |
| Performance and DR campaigns | Baseline only | Moderate | Graduated tiers with inventory quality filters | Over-blocking and starving delivery |
Fig. Risk priority by advertising scenario
| Dimension | Brand safety | Brand suitability |
|---|---|---|
| Core question | Should any advertiser fund this content? | Should this advertiser appear beside this content? |
| Who defines the standard | The industry, broadly by consensus | The individual brand |
| Risk threshold | Binary—permitted or excluded | Graduated across risk tiers |
| Enforcement method | Universal exclusions, pre-bid filters, verification tags | Contextual classification, allowlists, tiered policies by campaign |
| Scope of application | Identical across all campaigns and markets | Varies by campaign objective, audience and market |
| Consequence of failure | Reputational damage, public escalation, boycott risk | Diluted brand association, weaker response, wasted spend |
| Typical failure direction | Under-protection | Over-restriction and under-specification at the same time |
Fig. Brand safety and brand suitability compared.
| Approach | Level of control | Available scale | Privacy resilience |
|---|---|---|---|
| Build your own data infrastructure | Very high—you own the data and the rules | Limited to owned audiences and their extensions | Very high—consent collected directly |
| Buy third-party audience segments | Low—methodology sits with the provider | Very high—instant access to broad categories | Low–medium—dependent on eroding identifiers |
| Unify data with an independent layer | High—one governance view across platforms | High—combines owned, licensed, and contextual reach | Medium–high—reduces reliance on any single identifier |
Fig. Choosing a CTV data strategy.
| Strategy | Primary data source | Targeting precision | Best use case |
|---|---|---|---|
| Demographic, geographic, psychographic | Panels, census, survey data | Low–medium | Broad awareness and category entry |
| Behavioral and contextual | Intent signals; content metadata | Medium–high | Prospecting and privacy-safe scale |
| First-party and CRM-based | Owned customer data | Very high | Retention, winback, lookalike seeds |
| Retail media and commerce data | Retailer purchase and loyalty data | Very high | Product launches, category conquesting |
| Extended reach (third-party, ACR, cross-device) | Licensed audience data | Medium | Scaling beyond owned audiences |
| AI-driven segmentation and personalization | Combined multi-signal models | High | Efficiency and creative relevance at scale |
Fig. Core CTV targeting strategies at a glance.
| Signal | Data source | Precision | Scalability | Privacy resilience | Ideal use cases |
|---|---|---|---|---|---|
| Demographic | Census data, panels, registration data | Low–medium | Very high | High | Broad awareness, category reach |
| Behavioral | Browsing, purchases, app usage, modeled intent | High | Medium–high | Low–medium | Prospecting, in-market audiences |
| Contextual | Content genre, metadata, daypart, environment | Medium | High | Very high | Signal-loss resilience, brand suitability |
| First-party | CRM, loyalty programs, owned site and app data | Very high | Low–medium | Very high | Retention, upsell, lookalike seeds |
| Identity-based | Device graphs, login data, clean-room matches | High | Medium | Medium | Cross-device reach, frequency control |
Fig. CTV audience signals compared.
| Dimension | Linear TV | CTV targeting |
|---|---|---|
| Targeting approach | Program and daypart proxies for audience | Audience-first: households and individuals defined by data |
| Optimization | Fixed buys; adjustments between flights | Continuous, in-flight optimization of bids, budgets, and creative |
| Data sources | Panel-based ratings and demographic estimates | Demographic, behavioral, contextual, first-party, and identity data |
| Audience addressability | One ad per program for all viewers | Different ads to different households within the same content |
Fig. CTV targeting vs linear TV at a glance.
| Inventory type | Typical CPM |
|---|---|
| Standard OTT/CTV inventory | $25–40 |
| Premium content and live sports | $50+ |
| FAST / broad AVOD placements | Under $20 |
| Self-serve entry points | From ~$7 (e.g. Paramount Ads Manager) |
Fig. Typical CPMs by inventory.
| Inventory tier | Typical CPM range | Where it sits |
|---|---|---|
| FAST and free ad-supported | ~$15–25 | Tubi, Pluto TV, The Roku Channel, Samsung TV Plus |
| Premium AVOD, standard programmatic | ~$25–45 | Hulu, Peacock, HBO Max, Paramount+, Prime Video |
| Premium direct-sold, pods and live sports | ~$45–65 | Netflix direct, Amazon, premium sports adjacency |
| Audience-targeted with first-party or retailer data | ~$45–85 | Addressable premium typically adds 25–60% to open-market pricing |
| Linear TV, for comparison | ~$10–15 | Broadcast and cable |
Fig. Typical CPM ranges. Ranges compiled from published buy-side benchmark estimates (Adwave; NA Media Experts). These are vendor-stated market estimates, not audited data.
| If your goal is… | Start with | Why |
|---|---|---|
| Maximum reach, any budget | YouTube | The largest single media distributor on US television, with self-serve access at any spend level |
| Premium reach at scale | Netflix, Amazon Prime Video | Ad tiers now measured in hundreds of millions of viewers, with programmatic access maturing |
| Efficient frequency and incremental reach | Tubi, Pluto TV, The Roku Channel | Lowest CPMs in streaming, audiences that skew younger than linear |
| Brand safety and family audiences | Disney+ | Tightly controlled content environment, passionate franchise audiences |
| Live sports adjacency | Amazon Prime Video, Peacock, Paramount+ | NBA, NFL, Olympics, World Cup and Champions League rights sit here |
| Next-day broadcast and general entertainment | Hulu | Current-season network episodes, a decade of ad-product maturity |
| Prestige drama and unscripted breadth | HBO Max | Award-winning originals alongside a deep lifestyle catalog |
| Affluent audiences, no ad inventory | Apple TV | Sponsorship and integration only—the last major streamer without an ad tier |
Fig. OTT platforms at a glance.
| Criterion | Question to ask the vendor | What a weak answer sounds like |
|---|---|---|
| Goal alignment | Which business metric will this move, and over what period? | A list of features and capabilities |
| Data readiness | What data quality does this assume, and what happens without it? | "The model handles messy data" |
| Integration | What does implementation involve, and who does the work? | "It integrates with everything" |
| Governance | How is consent handled, and where does data reside? | Deferral to a security questionnaire |
| Explainability | How would we explain a decision to a regulator or CFO? | "The model is proprietary" |
| Scalability | What changes at ten times our current volume? | Silence on pricing thresholds |
| Lock-in | How do we export data and models if we leave? | "Nobody leaves" |
| Total cost | What are the costs beyond the license? | License pricing only |
| ROI validation | How can we verify results independently of your platform? | "Our reporting shows the lift" |
Fig. Enterprise AI marketing tool evaluation checklist.
| Application | Share of buyers planning to use | Concentrated among |
|---|---|---|
| Media planning and buying recommendations | 51% | Larger advertisers |
| Inventory discovery and evaluation | 51% | Larger advertisers |
| Creative testing, selection, or optimization | 50% | Small and mid-size advertisers |
| Media pre-planning | 45% | Small and mid-size advertisers |
| Performance analysis and outcome insight | 43% | Mixed |
Fig. Where advertisers are applying agentic AI, by intended use (Source).
| Category | Primary use case | Representative platforms | What it cannot tell you |
|---|---|---|---|
| Creative and content production | Generating and adapting copy, images, video, and design assets | ChatGPT, Canva AI, Adobe Firefly, Runway | Whether the creative drove incremental business results |
| Audience targeting and personalization | Segmentation, predictive targeting, lookalike modeling, activation | Meta Advantage+, Google Ads AI, Adobe Experience Platform | How the same audience performs outside that platform |
| Programmatic media buying | Automated bidding, budget allocation, campaign optimization | Google Ads AI, Display & Video 360, The Trade Desk Kokai | Whether the reported conversions were incremental |
| Measurement and marketing intelligence | Cross-channel analytics, attribution, marketing mix modeling | Google Meridian, Adobe Marketing Campaign Analytics, Triple Whale | What to change in the campaign tomorrow morning |
| Workflow automation | Process automation, reporting, approvals, integrations | Zapier, Make, HubSpot AI | Whether the automated process was worth automating |
Fig. AI marketing tool categories at a glance.
| OTT (Over-the-Top) | CTV (Connected TV) | |
|---|---|---|
| What it is | A content delivery method | A device category |
| Where it plays | Phones, tablets, desktops, consoles, smart TVs | Television screens only |
| Content source | Any internet-delivered video service | The same services, viewed on a TV |
| Ad formats | Interactive units, clickable overlays, companion banners, playables | Non-skippable 15s and 30s spots in ad pods, QR codes, shoppable units |
| Targeting basis | Device-level and individual signals | Household-level signals, plus content alignment |
| Typical use | Performance, retargeting, mobile reach | Awareness, brand building, premium placement |
| Examples | Netflix on a laptop, Hulu on a phone, Tubi on a tablet | Netflix on a Roku, Hulu on a smart TV, Tubi on a Fire TV Stick |
| Category | Typical pricing model | Published entry point | Implementation load | Ongoing resource requirement |
|---|---|---|---|---|
| Web analytics | Free tier; enterprise tier by contract | $0 | Low to moderate (tagging, event design) | Analyst time; data engineering for warehouse export |
| CRM-native reporting | Per seat plus contact volume | HubSpot Marketing Hub Professional $890/mo | Moderate; onboarding fees often mandatory | Marketing ops owner |
| Attribution platforms | Tracked accounts or monthly tracked users | Dreamdata Activation Starter $750/mo | Moderate to high; CRM and UTM hygiene critical | Marketing ops plus analyst |
| Marketing intelligence | Consultative, scoped to media volume | Quote-based | Higher; data integration across sources | Shared with vendor under managed models |
Fig. Marketing ROI tracking pricing models by category.
| Tool | Best for | Key features | Integrations | Pricing | Pros and cons |
|---|---|---|---|---|---|
| Google Analytics 4 | Web and app performance measurement on owned properties | Event-based tracking, conversion paths, data-driven attribution, BigQuery export | Google Ads, Search Ads 360, BigQuery, Looker, third-party tags | Free; Analytics 360 quote-based | + No license cost, near-universal familiarity, deep event granularity − Session-based, no view of non-clickable media, sampling at scale |
| HubSpot | Mid-market B2B connecting campaigns to pipeline in one system | Campaign attribution reporting, revenue attribution, lifecycle reporting, ad tracking | Native CRM, Google/Meta/LinkedIn ads, Salesforce sync, 1,000+ app marketplace | Marketing Hub Professional $890/mo, Enterprise $3,600/mo plus one-time onboarding | + Fast setup, marketing and sales data in one place − Attribution limited to tracked digital touchpoints; contact-tier costs escalate |
| Adobe Analytics | Large enterprises with dedicated analytics teams | Multi-dimensional segmentation, Attribution IQ, cross-device analysis, predictive models | Adobe Experience Cloud, CDPs, warehouses, ad platforms via connectors | Quote-based | + Analytical depth few tools match − Requires specialist staff; strongest inside the Adobe ecosystem |
| Salesforce | B2B enterprises measuring ROI against closed-won revenue | Campaign influence models, Einstein attribution, pipeline and revenue reporting | Native CRM, Marketing Cloud, MuleSoft, Data Cloud, external ad connectors | Quote-based | + Revenue figures finance already trusts − Upper-funnel and non-clickable media largely invisible without added tooling |
| Dreamdata | B2B revenue attribution across long, multi-stakeholder journeys | Multi-touch models, account-level journey mapping, pipeline attribution, warehouse access | CRM, marketing automation, LinkedIn, Google Ads, BigQuery, Snowflake | Free tier; Activation Starter $750/mo; advanced tier quote-based | + Purpose-built for B2B account journeys − Digital-touchpoint dependent; setup requires clean CRM and UTM discipline |
| AI Digital Elevate | Enterprises measuring across programmatic, CTV, retail media, and walled gardens | Marketing mix modeling, Path to Conversion, cross-DSP reporting, cookieless targeting, AI-assisted planning | 12+ DSPs, CRM and ad platform feeds, offline conversion inputs, BI export | Quote-based | + Vendor-agnostic measurement independent of any selling platform; complements existing analytics and CRM − Enterprise scope; less relevant for single-channel advertisers |
Fig. Best marketing ROI tracking tools compared (2026).
| Category | What it actually measures | Primary data sources | Main limitation |
|---|---|---|---|
| Web analytics | On-site and in-app behavior, tagged campaign traffic, session-attributed conversions | Site and app tags, ad platform imports, first-party events | Session-based, single-property view; no visibility into media not linked to a click |
| CRM reporting | Campaign-influenced leads, opportunities, pipeline, closed-won revenue | CRM records, form fills, sales activity, marketing automation | Strong on recorded touchpoints, blind to upper-funnel and non-clickable media |
| Attribution platforms | Credit assignment across observable digital touchpoints using multi-touch models | Ad platform APIs, site tags, CRM, identity graphs | Depends on identity resolution; weak coverage of CTV, offline, and walled-garden delivery |
| Marketing intelligence platforms | Cross-channel contribution to business outcomes, validated against modeled baselines | DSPs, SSPs, CRM, ad platforms, offline conversion feeds, retail data | Requires more integration work and organizational buy-in than single-purpose reporting |
Fig. How each software category defines marketing ROI.
| KPI | What it measures | Business risk or cost |
|---|---|---|
| Brand-safe impression rate | Percentage of measured impressions that meet the advertiser’s safety policy | Shows reputational exposure and the share of media delivered in approved environments |
| Invalid traffic rate | Percentage of impressions generated by bots, fraudulent devices, or other invalid activity | Indicates how much budget may have been spent on non-human or manipulated traffic |
| Page-level suitability score | Degree to which the content surrounding an ad matches campaign-specific risk thresholds | Identifies whether a placement fits the brand, even when the publisher is approved |
| MFA exposure percentage | Share of impressions or spend delivered to made-for-advertising inventory | Reveals potential waste, low-value reach, and reduced working-media efficiency |
| Unknown or unclassified rate | Share of inventory the verification provider could not evaluate | Highlights measurement gaps that should not automatically be treated as safe |
| Qualified-impression rate | Share of impressions meeting combined fraud, measurability, viewability, suitability, and quality standards | Shows how much purchased media had a credible opportunity to create business value |
| Suitability tier | Appropriate use | Typical controls |
|---|---|---|
| Strict | Regulated industries, children’s products, crisis-sensitive campaigns, and major brand launches | Approved publishers, narrow contextual categories, extensive exclusions, strict risk thresholds, and mandatory verification |
| Moderate | Most corporate, awareness, and performance campaigns | Page-level contextual analysis, standard exclusions, quality thresholds, and selected publisher controls |
| Permissive | Entertainment, gaming, news-related, or culturally flexible campaigns | Broader editorial access, fewer topic exclusions, and greater tolerance for legitimate sensitive content |
| Channel | Primary risks | Verification challenges | Recommended protections |
|---|---|---|---|
| Open web and display | Unsafe adjacency, misinformation, MFA sites, content farms, arbitrage and spoofed domains | Fragmented supply paths, changing page content and unverified long-tail publishers | Page-level contextual analysis, pre-bid filtering, ads.txt validation, allowlists, exclusion lists and supply path optimization |
| CTV and streaming | App spoofing, device spoofing, SSAI errors, mislabeled inventory, invalid traffic and excessive frequency | Limited show-level data, fragmented apps and devices, server-side delivery and restricted client-side measurement | App-ads.txt, authorized-seller checks, app and program-level validation, IVT detection, frequency controls and post-bid verification |
| Walled gardens and social platforms | User-generated content, misinformation, AI-generated content, rapidly changing feeds and unsuitable creator adjacency | Platform-controlled reporting, restricted log-level data and incomplete third-party access | Platform suitability settings, placement exclusions, independent verification, contextual measurement and cross-platform reporting |
| Factor | Brand safety | Brand suitability |
|---|---|---|
| Primary purpose | Avoid universally harmful or unacceptable content | Determine whether content fits a specific brand |
| Decision basis | Shared industry safety standards | Brand values, audience, campaign and risk tolerance |
| Typical examples | Terrorism, hate speech, illegal activity and graphic exploitation | Politics, alcohol, mature entertainment, sensitive news or competitive subjects |
| Risk levels | Generally safe or below the safety floor | Often classified as low, medium or high risk |
| Policy owner | Industry standards supported by advertiser controls | Individual advertiser or agency |
| Campaign impact | Prevents severe reputational and compliance exposure | Balances protection, relevance and campaign reach |
| Provider | Main capabilities | Best suited for |
|---|---|---|
| DoubleVerify | Viewability, IVT, fraud detection, suitability, and cross-channel verification | Advertisers seeking broad coverage across display, video, social, and CTV |
| Integral Ad Science | Media quality, contextual classification, brand safety, and suitability | Enterprise teams managing detailed suitability policies across channels |
| GeoEdge | Malvertising detection, creative scanning, and harmful-ad prevention | Publishers and platforms focused on ad security and creative quality |
| HUMAN Security | Bot detection, sophisticated IVT analysis, and fraud prevention | Advertisers prioritising traffic authenticity and advanced fraud protection |
| Channel | Key verification risks | Strategic priority |
|---|---|---|
| Display | Hidden placements, domain spoofing, pixel stuffing, low viewability | Page-level measurement, IVT detection, and domain transparency |
| CTV | Device spoofing, fraudulent apps, server-side ad insertion manipulation | App transparency, device validation, and certified measurement support |
| Audio | Background playback, muted delivery, fraudulent listening activity | Valid playback, device signals, and publisher-level reporting |
| Social | Restricted third-party access and platform-specific definitions | Verify available metrics and understand measurement limitations |
| Verification side | Primary objective | Typical insights | Common actions |
|---|---|---|---|
| Buy-side | Protect advertiser spend and campaign quality | IVT, viewability, suitability, domain authenticity | Block inventory, adjust bids, review vendors |
| Sell-side | Protect inventory value and publisher reputation | Traffic quality, unauthorised resale, page-level risks | Remove bad traffic, investigate partners, improve inventory controls |
| Industry body | Role in ad verification | Standards and areas covered | What advertisers should check |
|---|---|---|---|
| Media Rating Council (MRC) | Establishes measurement standards and independently audits verification services | Viewability, invalid traffic, brand safety, measurement methodology, reporting controls, and data quality | Confirm that the specific service, channel, format, and market being purchased are included within the provider’s MRC accreditation |
| Interactive Advertising Bureau (IAB) | Develops industry guidance for consistent digital advertising practices | Brand safety, suitability, fraud disclosure, measurement terminology, privacy, and reporting | Check whether the provider follows current IAB guidance and uses recognised definitions across campaign reporting |
| IAB Tech Lab | Creates technical standards that improve transparency across the programmatic supply chain | Ads.txt, app-ads.txt, sellers.json, supply-chain objects, content classification, and authorised seller verification | Assess whether the verification solution can identify unauthorised sellers, domain spoofing, and mismatches across supply paths |
| Verification pillar | What it measures | Why it matters | Common risks | Typical advertiser action |
|---|---|---|---|---|
| Viewability | Whether the ad appeared within the visible screen area | A served ad has limited value if users could not see it | Below-the-fold ads, background tabs, hidden placements | Exclude weak placements and adjust viewability targets |
| Invalid traffic | Whether the impression came from legitimate human activity | IVT wastes budget and distorts campaign reporting | Bots, crawlers, click farms, automated refreshes | Block suspicious sources and review supply partners |
| Brand safety and suitability | Whether surrounding content met brand requirements | Unsuitable placements can damage trust and reputation | Harmful content, misinformation, sensitive topics | Apply inclusion lists, exclusions, and suitability controls |
| Ad fraud prevention | Whether inventory or delivery was deliberately manipulated | Fraud makes illegitimate inventory appear valid | Ad stacking, pixel stuffing, domain spoofing | Block fraudulent domains and investigate supply paths |
| Scenario | Recommended model | Primary KPI | Main benefit | Key consideration |
|---|---|---|---|---|
| Performance marketing | ✅ Programmatic | CPA or incremental conversions | Continuous optimization at scale | Requires reliable conversion data |
| Brand awareness | ✅ Hybrid | Unique reach and brand lift | Scale plus premium environments | Deduplicate cross-channel reach |
| Product launch | ✅ Hybrid | Target-audience reach | Guaranteed impact with flexible extension | Secure priority placements first |
| Premium sponsorship | ✅ Direct | Attention or brand lift | Exclusive context and integration | Confirm incremental value |
| CTV | ✅ Hybrid | Incremental reach and frequency | Premium access plus audience targeting | Control overlap across platforms |
| Retail media | ✅ Hybrid | Incremental sales | Purchase data with broader reach | Validate retailer-reported results |
| B2B campaigns | ✅ Hybrid | Qualified pipeline | Data targeting plus specialist publishers | Small audiences can saturate quickly |
| Omnichannel activation | ✅ Programmatic-led hybrid | Business outcomes | Coordinated cross-channel delivery | Requires shared data and KPIs |
| Dimension | Programmatic advertising | Non-programmatic advertising | Strategic implication |
|---|---|---|---|
| Buying process | Automated through DSPs, SSPs, exchanges, and direct deal infrastructure | Manually negotiated with publishers or media owners through insertion orders | ✅ Programmatic reduces execution time; direct buying requires more coordination but supports custom terms |
| Targeting | Uses audience, contextual, behavioral, geographic, device, and first-party data signals | Relies mainly on publisher audience, content context, placement, geography, and negotiated packages | ✅ Programmatic supports granular audience activation; direct buying offers stronger contextual alignment |
| Optimization | Budgets, bids, audiences, creatives, and supply can be adjusted during the campaign | Changes usually require publisher approval or revised campaign terms | ✅ Programmatic is better suited to performance optimization and rapid testing |
| Optimization speed | Decisions can be made in real time or near real time | Changes may take hours, days, or longer to implement | ✅ Faster optimization can reduce wasted spend and improve budget allocation |
| Inventory quality | Ranges from open-exchange supply to curated premium inventory | Usually selected directly from known publishers or media owners | ❌ Open programmatic supply requires stronger quality controls; direct buying offers greater placement certainty |
| Transparency | Depends on the DSP, SSP, supply path, deal type, and reporting access | Placement and publisher are usually known, but pricing and delivery data may remain publisher-controlled | Neither model guarantees full transparency without clear contracts, verification, and governance |
| Measurement | Supports impression-level reporting, attribution, experimentation, and cross-channel analysis | Often depends on aggregated publisher reports and separately implemented tracking | ✅ Programmatic produces more granular data; direct campaigns may require modeled or independent measurement |
| Pricing | Dynamic CPMs in auctions or fixed CPMs in automated direct deals | Fixed CPMs, flat fees, sponsorship costs, or negotiated packages | ✅ Programmatic offers pricing flexibility; direct buying provides cost predictability |
| Scalability | Can activate large audiences across publishers, devices, markets, and channels | Scale is limited by individual publisher relationships and available inventory | ✅ Programmatic supports enterprise reach with less operational overhead |
| Brand safety | Managed through inclusion lists, blocklists, verification tools, curated supply, and PMPs | Supported by direct relationships with known publishers and agreed placements | ✅ Direct buying offers stronger inherent contextual control; programmatic requires active safeguards |
| Data ownership | Advertisers can activate first-party data and access detailed campaign-level signals, depending on platform terms | Publishers usually control audience data and provide aggregated reporting | ❌ Platform and publisher restrictions can limit advertiser visibility in both models |
| Best strategic role | Scale, audience targeting, omnichannel activation, testing, and continuous optimization | Premium context, guaranteed delivery, sponsorships, exclusivity, and custom integrations | ✅ Enterprise advertisers typically gain the most value by assigning each model a distinct role |
| Channel | Common inventory | Best use | Main targeting signals | Key measurement |
|---|---|---|---|---|
| Display | Web and in-app ads | Reach, retargeting, acquisition | Audience, context, device, first-party data | Viewability, clicks, conversions |
| Online video | In-stream and out-stream video | Awareness and consideration | Audience, content, context, device | Completion rate, attention, conversions |
| Connected TV | Streaming and smart-TV inventory | Premium reach and brand campaigns | Household, geography, content, first-party data | Reach, frequency, completion, sales lift |
| Digital audio | Streaming music, podcasts, internet radio | Awareness and niche audience reach | Audience, genre, location, listening context | Reach, listen-through rate, conversions |
| Digital out-of-home | Billboards, transit, retail, venue screens | Local reach and contextual campaigns | Location, time, weather, venue | Estimated impressions, mobility data, lift |
| Retail media | Retailer search, display, off-site, and CTV ads | Commerce and sales growth | Purchase, product, loyalty, shopper data | ROAS, sales, new-to-brand, incrementality |
| Native advertising | Ads matching publisher content | Content discovery and consideration | Context, interest, content category | Engagement, clicks, conversions |
| Mobile | In-app, rewarded, video, mobile-web ads | App growth and mobile commerce | App, device, location, behavior | Installs, in-app actions, purchases |
| Buying type | How it works | Control and quality | Best for | Main trade-off |
|---|---|---|---|---|
| Open Auction (RTB) | Buyers compete for impressions through real-time auctions | ❌ Quality and transparency vary | Broad reach, prospecting, testing, performance campaigns | Requires stronger fraud, brand-safety, and supply controls |
| Private Marketplace (PMP) | Selected buyers bid on curated publisher inventory | ✅ Higher control and generally better-quality supply | CTV, premium video, brand-sensitive campaigns | Less scale than the open auction |
| Preferred Deal | One buyer receives priority access at a fixed CPM | ✅ Predictable pricing and premium access | Advertisers wanting flexibility without volume commitments | Inventory and delivery are not guaranteed |
| Programmatic Guaranteed | Buyer and publisher agree on fixed pricing and reserved volume | ✅ Highest certainty and control | Launches, premium placements, predictable reach | Less flexibility and usually higher cost |
| Criterion | Why it counts | Questions to ask vendors |
|---|---|---|
| Measurement methodology | Determines whether numbers reflect reality or platform bias | Which methodology underlies each reported metric? How are models validated? |
| Data integration & ownership | Governs long-term flexibility and exit costs | Which native connectors exist? Is raw data portable? Who owns the dataset post-contract? |
| Cross-channel coverage | Partial coverage produces distorted budget decisions | Are CTV, programmatic, social, search, retail media, and offline all measured on one basis? |
| AI & forecasting | Turns measurement into forward-looking planning | What do AI features predict, and how are forecasts validated against outcomes? |
| Governance & speed | Analyst workload is a hidden cost center | How fast do models refresh? What can non-technical users self-serve? |
| Implementation & TCO | License price is a fraction of true cost | What are typical timelines, services costs, and internal resource requirements? |
Fig. Evaluation criteria for marketing measurement platforms.
| Platform | Category | Best for | Enterprise readiness | Key strength | Potential limitation |
|---|---|---|---|---|---|
| Funnel | Reporting & BI | Marketing data aggregation | High | 500+ connectors, low setup burden | Reports platform data; no independent measurement |
| Tableau | Reporting & BI | Enterprise-wide visualization | High | Powerful, flexible analytics | Needs analyst resources; marketing is not its focus |
| Microsoft Power BI | Reporting & BI | Microsoft-stack organizations | High | Cost-effective, deep MS integration | Same reporting-only ceiling |
| Dreamdata | Multi-touch attribution | B2B revenue attribution | Medium–High | Purpose-built for long B2B journeys | Digital-first; limited offline visibility |
| Rockerbox (DoubleVerify) | Multi-touch attribution | DTC and mid-market attribution | Medium–High | MTA plus MMM and testing in one product | Now part of a larger measurement vendor's suite |
| Triple Whale / Northbeam | Multi-touch attribution | Ecommerce performance teams | Medium | Fast, commerce-native insights | Ecommerce scope; signal loss exposure |
| Adobe Mix Modeler | MMM | Adobe-ecosystem enterprises | High | MMM blended with journey data | Meaningful implementation lift |
| Nielsen / Ipsos MMA | MMM | Global, media-heavy brands | High | Decades of modeling pedigree | Slower cycles; consulting-style engagements |
| Keen Decision Systems | MMM | Mid-market to enterprise planning | Medium–High | Fast, forward-looking scenario planning | Less granular than experiment-led tools |
| Haus | Incrementality | Experiment-led growth teams | Medium–High | Rigorous self-serve geo testing | Requires testable spend levels and volume |
| LiftLab | Incrementality + MMM | Continuous spend optimization | Medium–High | Experiments calibrate an agile MMM | Best value needs ongoing testing cadence |
| Measured | Incrementality | DTC and omnichannel brands | Medium–High | Incrementality-based views of every channel | Focused on paid media questions |
| Google Meridian | Unified (open-source MMM) | Data-science-equipped teams | Medium | Free, transparent, customizable code | Needs in-house modeling talent; Google-built |
| AI Digital Elevate | Unified marketing measurement | Independent cross-channel measurement | High | DSP-agnostic layer combining MMM, path to conversion, and AI planning across 12+ DSPs | Oriented to programmatic-active enterprises |
Fig. Marketing measurement platform comparison (2026)
| Business goal | Best-fit category | Why |
|---|---|---|
| Monitor KPIs and share performance across teams | Reporting & BI | Centralizes marketing data into accessible dashboards |
| Optimize digital campaigns and journeys week to week | Multi-touch attribution | Granular, touchpoint-level signals for fast optimization |
| Allocate annual budgets across online and offline channels | Marketing mix modeling | Privacy-resilient, holistic view of channel contribution |
| Prove marketing caused outcomes; validate other models | Incrementality & experimentation | Controlled experiments isolate true causal lift |
| Consolidate conflicting numbers into one source of truth | Unified marketing measurement | Combines methodologies across all channels, independently |
Fig. Which measurement category fits your business goal.
| Business priority | Recommended deal type | Why it fits and key trade-off |
|---|---|---|
| Guarantee delivery for a fixed launch date | Programmatic Guaranteed | Inventory and impression volume are reserved before launch, reducing delivery risk. The trade-off is lower flexibility once budget and volume are committed. |
| Secure premium CTV or publisher inventory | Programmatic Guaranteed | Reserved buying protects access to scarce or high-demand placements. Advertisers may pay a premium and must finalize targeting and creative requirements early. |
| Create predictable media budgets | Programmatic Guaranteed | A negotiated CPM and committed volume make spending easier to forecast. However, reallocating budget after activation may require renegotiation. |
| Access premium inventory without committing spend | Preferred Deal | The advertiser receives priority access at a fixed CPM but can decline individual impressions. Delivery is not guaranteed because inventory remains unreserved. |
| Test a new publisher or inventory package | Preferred Deal | Buyers can evaluate inventory quality and performance before making a larger commitment. The publisher may not always have enough eligible supply available. |
| Maintain performance-optimization flexibility | Preferred Deal or PMP | Both allow buyers to adjust bidding and impression selection as results change. They provide less delivery certainty than a guaranteed transaction. |
| Reach curated audiences across selected publishers | PMP | Invitation-only auctions offer more control and inventory curation than the open market. Buyers must still compete, so pricing and delivery can fluctuate. |
| Scale prospecting and audience expansion | Open Auction | Broad supply and real-time bidding support extensive reach and rapid audience discovery. Inventory quality and supply paths require closer monitoring. |
| Prioritize cost-efficient reach | Open Auction | Competition across large volumes of inventory can create efficient CPMs. The lowest price does not necessarily provide the strongest placement quality or business outcome. |
| Balance certainty, quality, and scale | Hybrid approach | Combining guaranteed, preferred, and auction buying lets each model serve a distinct campaign objective. The trade-off is greater operational and measurement complexity. |
| Criterion | Programmatic Direct | Programmatic Guaranteed |
|---|---|---|
| Deal structure | Broad category of automated, one-to-one transactions between a specific buyer and publisher | Reserved deal negotiated between a specific buyer and publisher |
| Inventory reservation | Depends on the deal type; inventory may be reserved or unreserved | A fixed volume of inventory is reserved in advance |
| Purchase obligation | Varies; some direct deals allow the advertiser to decline available impressions | Advertiser commits to purchasing the agreed impression volume |
| Pricing | May use a negotiated fixed CPM or controlled auction pricing | Negotiated fixed CPM established before campaign activation |
| Targeting | Supports publisher, placement, contextual, audience, geographic, device, and first-party data targeting | Targeting conditions are agreed in advance and must remain compatible with forecasted delivery |
| Flexibility | Generally higher, particularly with Preferred Deals and non-guaranteed transactions | Lower after the agreement is finalized because volume, timing, and spending are committed |
| Campaign scale | Can range from individual publisher deals to broader curated buying across multiple supply sources | Limited to the inventory volume and publishers specified in the agreement |
| Measurement | Reporting and measurement conditions vary according to the platform and deal structure | Measurement requirements, reporting fields, and delivery expectations can be negotiated before launch |
| Delivery certainty | Varies by deal type and available inventory | High, subject to the agreed targeting, creative, and campaign conditions |
| Ideal use cases | Flexible premium buying, publisher testing, curated reach, and campaigns requiring direct inventory access | Product launches, CTV, seasonal campaigns, sponsorships, premium placements, and fixed delivery requirements |
| Criterion | Programmatic Guaranteed | Preferred Deal | Private Marketplace (PMP) |
|---|---|---|---|
| Inventory reservation | Reserved in advance | Not reserved | Not reserved |
| Purchase obligation | Advertiser commits to an agreed volume | No purchase obligation | No obligation; buyer must win the auction |
| Pricing model | Negotiated fixed CPM | Negotiated fixed CPM | Auction-based, usually with a floor price |
| Flexibility | Lower after the agreement is finalized | High | High |
| Inventory quality | Premium, specifically packaged inventory | Premium or preferred-access inventory | Curated, invitation-only inventory |
| Advertiser control | High control over placement, timing, and volume | High control over whether to purchase impressions | Moderate to high, but auction-dependent |
| Ideal use cases | CTV, launches, sponsorships, and seasonal campaigns | Publisher testing and flexible premium buying | Curated reach and scalable audience buying |
| Reporting layer | What to show | Example | Executive question it answers |
|---|---|---|---|
| Headline business impact | Pre- and post-deployment comparison on financial metrics | CAC $142 → $118 across two matched quarters | Did this make us money? |
| Validated contribution | Incremental lift isolated by controlled test | Holdout test shows 11% incremental conversion lift | Would this have happened anyway? |
| Supporting efficiency | Time and capacity gains, with redeployment stated | 240 production hours redirected to creative testing | What else did we get? |
| Planning quality | Forecast accuracy and allocation improvements | Forecast error down from 18% to 9% | Are our next decisions better? |
| Caveats | Confounding variables during the measurement period | New creative launched in month two | What could be wrong with this? |
Fig. What an AI marketing ROI report should contain.
| Metric type | Examples | What it measures | Why it earns attention | Direct evidence of ROI? |
|---|---|---|---|---|
| Business impact | CPA, CAC, ROAS, CLV, LTV:CAC, incremental revenue | Financial return generated by marketing activity | Speaks the language finance already uses | Yes |
| Operational efficiency | Campaign production time, creative output, reporting turnaround, hours saved | Speed and capacity of the marketing function | Demonstrates adoption; frees capacity for higher-value work | No—supporting evidence only |
| Strategic planning | Forecast accuracy, budget allocation efficiency, scenario coverage | Quality of investment decisions before results land | Reduces the cost of being wrong; compounds across quarters | Indirect |
Fig. Classifying AI marketing metrics
| Category | Primary objective | Typical business outcomes | Key KPIs affected | Example AI capabilities |
|---|---|---|---|---|
| Media buying and campaign optimization | Reduce wasted spend and improve conversion efficiency | Lower acquisition costs, higher share of qualified impressions, less budget lost to low-value inventory | CPA, CAC, ROAS, conversion rate | Predictive bidding, audience modeling, supply-side inventory selection, budget reallocation |
| Personalization and customer engagement | Increase relevance across the customer lifecycle | Higher repeat purchase rates, longer retention, larger average order values | CLV, retention rate, repeat purchase rate, email conversion rate | Behavioral segmentation, recommendation engines, churn scoring, send-time optimization |
| Creative production and testing | Increase creative volume and testing velocity | Faster launches, more variants tested, lower production cost per asset | Creative output, production time, CTR, conversion rate | Generative copy and image production, variant generation, creative scoring, dynamic creative optimization |
| Marketing measurement and decisioning | Unify cross-channel data and improve accuracy of performance reads | Better budget allocation, fewer misallocated dollars, defensible executive reporting | Forecast accuracy, allocation efficiency, incremental revenue | Data unification, marketing mix modeling, path-to-conversion analysis, scenario planning |
Fig. AI marketing solution categories and the outcomes they affect.
| Metric | What it measures | When to use it | Primary limitation |
|---|---|---|---|
| Adjusted ROAS | Revenue per ad dollar, corrected for platform over-claiming and organic baseline | Comparing channel efficiency on a fair basis | Still revenue-based; ignores margin and long-term customer value |
| Blended CAC | Total acquisition cost across all channels ÷ new customers | Board-level efficiency and overall health | Hides channel-level differences; can mask which channel drives cost |
| LTV:CAC ratio | Customer lifetime value relative to acquisition cost | Judging sustainable, profitable growth | LTV is a forecast; slow to read; only as good as its inputs |
| Incremental CPA | Cost per genuinely incremental conversion, from lift tests | Validating a channel's true contribution | Requires experiments; not feasible for every channel |
| Marketing contribution margin | Marketing-driven revenue minus marketing and product costs | Profit-focused decisions and CFO conversations | Needs clean finance integration; harder to compute |
Fig. Marketing ROI metrics compared.
| Category | Core job | Business problem it solves | Where it stops |
|---|---|---|---|
| Analytics and data collection | Capture interactions, engagement, and on-site behavior | Creates the raw measurement foundation | Sees one channel at a time; needs stitching for a full journey |
| Marketing attribution | Assign conversion credit across touchpoints | Guides channel-level optimization | Signal-dependent; largely within-channel |
| Marketing mix modeling | Model channel contribution from aggregate history | Informs strategic, privacy-safe budget allocation | Not tactical; slower to read |
| Business intelligence and reporting | Consolidate and visualize data for executives | Turns scattered metrics into board-level reporting | Visualizes data; doesn't unify it at the source |
| Data integration | Connect and reconcile data across systems | Builds a single source of truth | Infrastructure, not insight |
| Marketing intelligence platforms | Combine measurement, analytics, forecasting, and optimization | Supports cross-channel investment decisions | Only as good as the data feeding it |
Fig. Digital marketing ROI tool categories at a glance.
| Approach | What it measures | Question it answers | Strengths | Limitations | Best used for |
|---|---|---|---|---|---|
| Attribution | How conversion credit is distributed across digital touchpoints | Which channels and touchpoints influenced this conversion? | Granular, fast, tactical; guides daily optimization | Depends on cookies and trackable signals; biased toward the last or lower-funnel touch; weak on offline and brand | Optimizing trackable digital channels in-flight |
| Marketing mix modeling (MMM) | The aggregate contribution of each channel and external factors to outcomes | How should I split budget across all channels, including offline? | Privacy-safe; whole-portfolio view; captures offline, brand, and untrackable media | Needs historical data; slower cadence; not built for tactical calls | Strategic budget allocation across the full mix |
| Incrementality testing | Causal lift versus a control group | Did this spend cause net-new results, or capture demand that existed already? | Proves causation; validates the other two methods | Requires experiment design; runs test by test; can't cover everything continuously | Validating high-spend channels and settling attribution disputes |
Fig. Three measurement approaches compared.
| Campaign Type | Suggested Starting Frequency Cap | Best Used For | Optimization Notes |
|---|---|---|---|
| Awareness | 2–4 impressions per user per week | Brand awareness and reach | Keep caps moderate while monitoring incremental reach and recall signals. |
| Consideration | 3–5 impressions per user per week | Product education and comparison | Use varied creative to explain different benefits or proof points. |
| Retargeting | 3–7 impressions per user per week | Lead generation and conversions | Watch CPA, conversion rate, and negative feedback closely. |
| CTV | 2–5 impressions per household per week | Premium video and brand impact | Use household-level controls where available and monitor cross-device overlap. |
| Online Video | 2–4 impressions per user per week | Awareness, engagement, and storytelling | Track completion rate, VTR, and creative fatigue by frequency level. |
| Display | 4–8 impressions per user per week | Efficient reach and remarketing support | Higher caps may be acceptable because individual impressions are less immersive. |
| Audio | 2–5 impressions per listener per week | Brand recall and local campaigns | Avoid excessive repetition because users cannot visually skip the message. |
| Promotional Campaigns | 5–10 impressions per user during the promotion window | Time-sensitive offers and events | Use only when urgency is real and campaign duration is short. |
| Marketing channel | How ROI typically behaves | Key factors that influence it | Common measurement challenges |
|---|---|---|---|
| Programmatic display | Efficient reach, but return is easily overstated by platform reporting | Supply quality, viewability, working-media share | Attributed vs. incremental revenue; MFA and invalid traffic |
| Paid search | High-intent, strong apparent ROI—but credit-inflated | Keyword intent, competition, landing-page fit | Claims conversions other channels seeded |
| Content marketing | Compounding and long-payback; usually undercounted | Topic authority, SEO, distribution | Last-click undervalues early touches |
| Email marketing | Consistently efficient on low incremental cost | List quality, segmentation, lifecycle stage | Hard to separate from demand created elsewhere |
| Social media | Ranges widely by objective and platform | Creative quality, audience match, funnel stage | Self-reported platform attribution |
| Connected TV (CTV) | Growing reach; long-tail, brand-led returns | Creative, frequency, incrementality | Fragmented measurement; view-through overcounting |
Fig. Marketing ROI by channel—what drives it and where it's mismeasured.
| Metric | What it measures | How it's derived | Best used for |
|---|---|---|---|
| Marketing ROI | Profitability of marketing investment | (Marketing-attributed revenue − marketing cost) ÷ marketing cost, as a % | Budget decisions; proving marketing's contribution to profit |
| ROAS | Efficiency of ad spend | Revenue from ads ÷ ad spend, as a ratio (e.g., 4:1) | Comparing channel efficiency; in-flight optimization |
| Revenue attribution | Which touchpoints get credit for a conversion | A model—first-touch, last-touch, multi-touch, or data-driven—not a single formula | Assigning credit before ROI or ROAS can be trusted |
Fig. ROI, ROAS and revenue attribution compared.
| Metric | What it answers | Review cadence | Decision it supports | Failure mode when used alone |
|---|---|---|---|---|
| Business metrics | ||||
| Marketing efficiency ratio (MER) | Is total marketing producing return? | Monthly | Total budget level | Hides channel-level problems |
| LTV:CAC | Are the customers we buy worth the price? | Quarterly | Acquisition strategy | Slow to reflect recent changes |
| Customer lifetime value | What is a customer worth over time? | Quarterly | Retention investment | Modeled, not observed |
| Contribution margin | Does marketing add profit, not just revenue? | Monthly | Pricing and mix | Requires clean cost allocation |
| Retention rate | Are we keeping what we bought? | Monthly | Lifecycle investment | Lags acquisition decisions |
| Campaign metrics | ||||
| ROI and ROAS | Did this activity return more than it cost? | Weekly | Tactical adjustment | Inflated by attribution bias |
| CPA and CAC | What did a conversion cost? | Weekly | Bidding and budget pacing | Ignores customer quality |
| Conversion rate | Is the path from click to purchase working? | Weekly | CRO priorities | Improves as traffic narrows |
| Cost per incremental conversion | What did causally attributable volume cost? | Per test | Reallocation decisions | Requires experimental design |
| Engagement metrics | Is the creative working? | Continuous | Creative rotation | Weak link to revenue |
Fig. Business metrics vs campaign metrics.
| Channel | Biggest optimization opportunity | Primary success metric | Main ROI risk | Supporting technology |
|---|---|---|---|---|
| Paid search | Separating branded from non-branded incrementality | Cost per incremental conversion | Paying for demand that already existed | Experimentation platforms; independent measurement |
| Paid social | Creative velocity and audience value tiering | Blended return; cost per new customer | Platform-reported conversions overstating contribution | Creative testing tools; holdout designs |
| Programmatic and CTV | Supply path and inventory quality | Working media share; incremental reach | MFA exposure and invalid traffic | Supply optimization; verification |
| SEO | Conversion value of existing traffic | Revenue per session | Optimizing for volume over commercial intent | Analytics; content performance tools |
| Lifecycle segmentation and timing | Revenue per recipient | Sending frequency eroding list value | CRM and lifecycle automation |
Fig. Where the ROI opportunity sits, by channel.
| Business objective | Validation method | Optimization priority | Supporting measurement | Expected business outcome |
|---|---|---|---|---|
| Protect margin | Incrementality test on high-spend channels | Remove non-incremental spend | Blended return; contribution margin | Same revenue, lower cost |
| Lower acquisition cost | Channel-level holdout | Reallocate toward efficient acquisition | CAC by channel; cost per incremental conversion | More customers at equal spend |
| Improve retention | Cohort analysis against exposure | Rebalance toward existing customers | Lifetime value; repeat purchase rate | Higher lifetime value per customer |
| Fund growth | Media mix modeling across horizons | Identify under-saturated channels | Marginal return by channel | Profitable expansion of spend |
| Source of loss | Underlying cause | Typical symptom | Effect on profitability | How to identify it | Optimization approach |
|---|---|---|---|---|---|
| Walled garden bias | Platforms measure and report their own contribution | Channel returns sum to more revenue than the business booked | Overinvestment in channels with low incremental value | Compare blended return against the sum of platform-reported returns | Independent measurement; holdout and geo testing |
| Poor media quality | MFA inventory, long supply paths, low-attention placements | Strong impression and viewability metrics, weak business outcomes | Working media diluted before it reaches a real person | Domain-level placement audit; quality benchmarks by environment | Supply path optimization; inventory selection by outcome |
| Audience overlap | Frequency managed per platform, never across them | Rising cost per incremental customer as spend scales | Repeat exposure billed as new reach | Deduplicated cross-channel reach and frequency analysis | Cross-platform frequency governance and suppression |
| Cross-channel blind spots | Disconnected reporting by channel and vendor | Conflicting conversion counts; unexplained performance | Budget allocated to the last click rather than the cause | Unified measurement across the full journey | Cross-channel attribution supported by incrementality |
| Invalid traffic | Bots, spoofing, fraudulent inventory | Clean-looking engagement, no revenue lift | Direct loss plus corrupted optimization signals | Verification and fraud detection at pre-bid | Pre-bid filtering; supply hygiene |
| Creative fatigue | Assets treated as a fixed input | Declining conversion rate at stable delivery | Falling efficiency mistaken for audience exhaustion | Frequency-adjusted performance decay by asset | Continuous creative testing and rotation |
Fig. Where marketing ROI leaks: causes, symptoms and remedies.
| Business type | Best-fit tools |
|---|---|
| B2C ecommerce | Triple Whale, Northbeam, Rockerbox |
| B2B SaaS | Dreamdata, Ruler Analytics |
| Enterprise | Adobe Analytics, Measured, Rockerbox |
| Mobile-first | AppsFlyer |
| Agency | Rockerbox, Triple Whale (multi-store), GA4 as a baseline |
Fig. Match of business type to the tools.
| Tool | Best for | Attribution models | Channels | Starting price | Key strength | Limitations |
|---|---|---|---|---|---|---|
| SegmentStream | Teams wanting AI-driven, decision-focused measurement | Probabilistic conversion modeling, incrementality | Paid, web, CRM; 16+ platforms | Custom (self-serve + enterprise) | AI-agent measurement layer, cookieless | Learning curve; weaker offline |
| Rockerbox | Omnichannel and mid-market brands | MTA + MMM + incrementality | Paid, organic, email, offline, CTV | Custom / sales-led | Unified measurement in one platform | Guided onboarding; now part of DoubleVerify |
| Northbeam | High-spend DTC and ecommerce | ML deterministic + probabilistic, view-through, MMM | Paid digital; limited offline | ~$1,000–1,500/mo (no free tier) | Creative-level attribution | Limited B2B and offline; technical |
| Triple Whale | Shopify DTC brands | Probabilistic MTA; MMM via add-on | Shopify, paid, email, SMS | Free; paid from ~$129–219/mo (GMV-tiered) | Deep Shopify and creative analytics | Shopify-first; feature-heavy |
| Ruler Analytics | B2B and lead generation | Six models plus MMM layer | Web, calls, forms, chat → CRM | From £199/mo | Closes the web-to-CRM revenue loop | Needs a CRM; not for ecommerce |
| Dreamdata | B2B SaaS | First/last-touch, U- and W-shaped, custom | Paid, organic, CRM, product | Free; paid from ~$750/mo | Account-level pipeline attribution | Needs clean CRM; overkill for B2C |
| AppsFlyer | Mobile-first apps | SKAN 4 + AdAttributionKit, deterministic | Mobile networks (10,000+) | Free tier; ~$0.07/conversion | Mobile measurement and fraud protection | Mobile-focused; cost scales with volume |
| Measured | Enterprise brands | Incrementality + test-calibrated MMM | Digital, TV, offline, walled gardens | Custom (~$50K+/yr) | Causal, defensible measurement | Needs volume; longer time to insight |
| Google Analytics 4 | Google-centric teams; a baseline | Data-driven plus two last-click | Digital only; Google-leaning | Free | Free cross-channel baseline | Google bias; no offline or CTV |
| Adobe Analytics | Large enterprises | Algorithmic plus custom models | Web, app, offline via CJA | Custom (~$50K–200K+/yr) | Enterprise depth and CJA | Complex; costly; steep learning curve |
Fig. Cross-channel marketing attribution tools comparison.
| Platform | Primary Purpose | Key Strengths | Main Limitation | Best For |
|---|---|---|---|---|
| Elevate | Cross-channel marketing intelligence and measurement | Independent performance view, MMM, cross-channel planning, AI-driven insights | Not a replacement for CRM or general-purpose BI | Marketing leaders managing complex paid-media ecosystems |
| Adverity | Marketing data integration | 600+ connectors, automated ETL, data harmonization | Often requires BI or measurement tools downstream | Data teams managing fragmented marketing sources |
| AlphaSense | Market and competitive intelligence | Extensive premium research library, AI-powered search, external market intelligence | Does not measure campaign performance | Strategy, research, and market intelligence teams |
| Crayon | Competitive intelligence | Competitor monitoring, change detection, sales enablement | Focuses on competitive signals rather than media performance | Product marketing and sales enablement teams |
| Klue | Competitive intelligence and win-loss analysis | Competitive insights connected to active deals and sales workflows | Not designed for marketing attribution | Revenue and competitive intelligence teams |
| HubSpot | CRM and marketing automation | Customer data, lifecycle workflows, lead management | Limited as an independent paid-media measurement layer | Growing teams connecting marketing and sales |
| Salesforce | Enterprise CRM and marketing ecosystem | Customer intelligence, automation, reporting, broad ecosystem | CRM intelligence alone does not provide independent cross-channel measurement | Large enterprises managing complex customer journeys |
| Model | Primary ROI metrics | Sales cycle | Measurement watch-out |
|---|---|---|---|
| Ecommerce | ROAS, AOV, LTV, repeat-purchase rate, contribution margin | Short, transactional | Over-crediting last click; ignoring returns |
| B2B SaaS | Pipeline contribution, CAC, LTV:CAC, payback period, recurring revenue | Long, multi-touch | Judging long cycles on short-window conversions |
| Lead generation | Qualified leads, closed revenue, lead-to-sale rate | Variable, offline close | Optimizing lead volume over lead quality |
Fig. What ROI looks like by business model.
| Channel | Typical reported ROI | Measurement reliability | Main attribution challenge | Average sales cycle | Recommended primary KPI |
|---|---|---|---|---|---|
| Very high (platform/vendor-reported) | Low—self-reported, rarely deduplicated | Credits owned-audience sales it did not cause | Short | Incremental revenue per send | |
| Paid search | High | Medium—but heavy organic overlap | Harvests existing demand; overstates lift | Short to medium | Incremental conversions |
| Paid social | Moderate to high | Low to medium—platform self-attribution | View-through inflation, cross-platform double-counting | Short to medium | Verified CAC |
| Display / programmatic | Lower, wide variance | Medium—depends on supply quality | Wasted impressions, viewability | Medium | Cost per incremental outcome |
| Connected TV | Emerging, upper-funnel | Medium—improving with closed-loop tools | Long lag between exposure and conversion | Long | Incremental reach and lift |
Fig. Channel ROI at a glance—and why the number needs a second look
| Measure | What it counts | Costs included | Decision it informs | Primary audience |
|---|---|---|---|---|
| ROAS | Revenue per dollar of advertising | Media spend only | Bidding, creative and channel-level efficiency | Media buyers, performance teams |
| ROI | Profit per dollar of total investment | Media, creative, technology, fees, staff | Budget justification and profitability | CMO, finance, executive team |
Fig. ROI vs ROAS—what each answers and when to use it.
| Industry | Where direct advertising adds value | Where programmatic adds value | Primary buying consideration |
|---|---|---|---|
| Retail | Premium publisher partnerships, seasonal sponsorships, homepage takeovers, and high-impact product launches | Shopper targeting, product-level retargeting, audience expansion, and coordinated activation across retail media networks | Connecting media exposure with online and in-store sales |
| eCommerce | Reserved placements during major promotions, product launches, and shopping events | Dynamic product ads, cart-abandonment retargeting, automated bidding, and conversion optimization | Product-feed quality, attribution, and cost per acquisition |
| Financial services | Trusted publisher environments, premium business media, and controlled sponsorships | Contextual targeting, qualified audience acquisition, and sequential messaging across the customer journey | Compliance, data governance, and brand suitability |
| Healthcare | Specialist publishers, professional media, and carefully reviewed content environments | Contextual activation, geographic targeting, and scalable awareness using privacy-conscious signals | Regulatory requirements, sensitive data, and messaging accuracy |
| B2B | Trade-publication sponsorships, event partnerships, newsletters, and account-specific publisher packages | Account-based targeting, content promotion, lead generation, and retargeting across long sales cycles | Lead quality, buying-committee reach, and CRM integration |
| Travel | Destination partnerships, premium travel content, and seasonal sponsorships | Real-time targeting based on location, browsing intent, seasonality, and booking behaviour | Demand volatility, booking windows, and geographic relevance |
| Automotive | Product-launch sponsorships, premium video environments, and specialist automotive publishers | Local dealer campaigns, model-level targeting, retargeting, and cross-device journey management | Regional inventory, long consideration cycles, and offline conversion measurement |
| Luxury | Exclusive publisher relationships, controlled editorial environments, bespoke creative, and high-impact placements | Curated PMPs, affluent audience segments, controlled frequency, and premium CTV or video inventory | Brand equity, scarcity, context, and exposure quality |
| Factor | Direct advertising | Programmatic advertising |
|---|---|---|
| Inventory access | Reserved inventory purchased from specific publishers, including premium placements, sponsorships, homepage takeovers, and custom integrations | Inventory accessed across publishers, ad exchanges, private marketplaces, preferred deals, and programmatic guaranteed arrangements |
| Targeting | Relies mainly on publisher audiences, content categories, geography, placement, and negotiated first-party audience segments | Uses first-party data, contextual signals, location, device, behavioural attributes, and real-time audience signals |
| Buying process | Managed through publisher sales teams, proposals, negotiations, insertion orders, and manual approvals | Executed through DSPs, SSPs, ad exchanges, and automated deal workflows |
| Pricing | Typically uses fixed CPMs, flat sponsorship fees, or guaranteed spending commitments | May use auction-based pricing, negotiated CPMs, preferred rates, or guaranteed pricing |
| Delivery certainty | Inventory and impression volumes are normally reserved in advance | Delivery depends on available supply, targeting restrictions, bid competitiveness, and deal type |
| Optimisation | Campaign changes may require publisher coordination or amendments to the agreement | Bids, budgets, audiences, pacing, frequency, and creative can be adjusted continuously |
| Transparency | Provides a clear publisher relationship and known placement, although reporting depth may vary | Can offer impression-level reporting, but intermediaries, fees, and complex supply paths may reduce visibility |
| Scalability | Limited by the number of publisher relationships and agreements a media team can manage | Enables campaigns to scale across publishers, formats, devices, audiences, and markets through one buying platform |
| Campaign flexibility | Lower flexibility when budgets, dates, and placements are committed in advance | Greater flexibility to pause activity, reallocate spend, and respond to performance signals |
| Brand control | Strong control over the publisher, placement, campaign timing, and surrounding content | Depends on deal quality, inventory controls, allowlists, blocklists, contextual filters, and verification tools |
| Operational requirements | Requires negotiation, trafficking coordination, reconciliation, creative approvals, and publisher-specific reporting | Requires platform expertise, data management, supply-path oversight, brand-safety controls, and ongoing optimisation |
| Best suited for | Premium campaigns requiring guaranteed delivery, exclusivity, predictable placements, or close publisher alignment | Campaigns requiring scalable audience targeting, cross-channel reach, real-time optimisation, and flexible budget allocation |
| MMM insight | Planning decision | Activation decision | Measurement follow-up |
|---|---|---|---|
| Brand media drives search demand | Protect upper-funnel budget | Coordinate TV, CTV, YouTube, and search timing | Track branded search, direct traffic, and total sales |
| Retargeting is overcredited | Reduce retargeting allocation | Tighten frequency and audience exclusions | Run holdout or geo test |
| CTV is more incremental in select markets | Reallocate by region | Adjust audience, inventory, and frequency | Compare market-level response |
| Podcast ads show delayed contribution | Add audio to the plan | Test host-read, show type, and frequency | Model adstock and lagged response |
| Retail media performance varies by network | Move spend to stronger networks | Adjust product, retailer, and audience strategy | Validate with incrementality testing |
| Promotions distort media results | Separate promo and media budgets | Plan campaigns around margin, not only volume | Include price and promotion variables |
Fig. How MMM insights translate into spend optimization.
| Industry | Main measurement challenge | Useful MMM inputs | Common optimization lesson |
|---|---|---|---|
| FMCG and consumer goods | Separating media from price, promotion, and distribution | Retail sales, promotions, price, distribution, media, seasonality | Media value depends heavily on retail conditions |
| DTC and ecommerce | Fast channel saturation and weakened attribution | Spend, revenue, new customers, margin, promos, creative cycles | Retargeting and paid social need incrementality checks |
| SaaS and subscription | Long buying cycles and delayed revenue | Pipeline, ARR, trials, opportunities, sales cycle length, media | Lead volume is not the same as incremental revenue |
| Retail and omnichannel | Online media driving offline sales | Store sales, ecommerce, loyalty, media, region, RMN data | Total sales measurement changes channel value |
| Travel and hospitality | Seasonality and demand shocks | Bookings, pricing, capacity, search trends, media, events | Media must be separated from natural demand cycles |
Fig. MMM priorities by industry.
| Common MMM finding | What attribution may suggest | What MMM may reveal | Budget implication |
|---|---|---|---|
| Brand media is undervalued | TV, YouTube, CTV, OOH, and audio appear weak because they produce fewer clicks | These channels may drive search demand, direct traffic, store visits, or later conversions | Protect or increase upper-funnel investment if incremental contribution is strong |
| Retargeting is overcredited | Retargeting appears highly efficient because users are close to purchase | Many conversions may have happened without exposure | Reduce waste and move budget toward prospecting or demand generation |
| Paid search depends on other media | Search looks like the main performance driver | Search may capture demand created by TV, video, social, or offline media | Avoid cutting upper-funnel spend simply because search has the strongest platform ROAS |
| Channels saturate | A channel looks strong on average | Marginal ROI may decline at higher spend levels | Cap investment and test new channels or audiences |
| Offline sales are hidden | Digital campaigns look underpowered in online-only reporting | Media may drive store visits, retail sales, or wholesale revenue | Include offline sales and retail data in the model |
| Promotions distort results | Sales spikes are credited to advertising | Discounting, pricing, or distribution may have driven the lift | Separate promotional effects from true media contribution |
| Regional performance varies | National results look average | Some markets may have strong response while others are saturated or inefficient | Reallocate by market, not only by channel |
Fig. Common findings across MMM case studies.
| Method | Granularity | Data type | Cadence | Key blind spot |
|---|---|---|---|---|
| Multi-touch attribution (MTA) | User-level, per touchpoint | Tracked digital events | Continuous | Offline media and untracked touchpoints |
| Marketing mix modeling (MMM) | Aggregate, channel-level | Time-series spend and outcomes | Periodic | Granular, in-the-moment decisions |
| Incrementality testing | Test vs. control groups | Experimental | Per test | Always-on, channel-by-channel reporting |
| Unified measurement | Combined | Multiple sources | Continuous + periodic | Complexity and resource requirements |
Fig. What each measurement method actually sees.
| Source | Conversions claimed | Crediting logic |
|---|---|---|
| Paid search | 62 | Last click within a 30-day window |
| Paid social | 48 | View-through within a 7-day window |
| Video / CTV | 21 | View-through within a 14-day window |
| Email / CRM | 15 | Last interaction before purchase |
| Total claimed | 146 | — |
| Actual customers | 100 | Verified against sales records |
Fig. How double counting inflates reported conversions.
| Approach | What it measures | Primary data sources | Best used for |
|---|---|---|---|
| Cross-channel measurement | Marketing performance across digital channels through one unified view | Platform data, web and app analytics, CRM, first-party events | Day-to-day optimization and budget allocation across digital |
| Cross-media measurement | Combined effect of digital and traditional media on outcomes | Modeled and aggregated data, panels, sales data, media spend | Strategic planning across TV, radio, print, and digital |
| Omnichannel measurement | The quality and consistency of the customer experience across touchpoints | Customer interaction data, service, retail, and journey data | Improving experience, retention, and engagement |
Fig. Cross-channel, cross-media, and omnichannel measurement compared.
| Network | Reach & scale | First-party data | Ad formats | Onsite / offsite / in-store | DSP & buying |
|---|---|---|---|---|---|
| Amazon Ads | The largest RMN by far: $68.6B ad revenue in 2025, up 22%; eMarketer expects its retail media revenues to top $75B by 2028, more than $65B ahead of the next-largest network | Purchase, browsing and streaming signals across the Amazon ecosystem, incl. Prime | Sponsored Products/Brands, display, video, streaming TV on Prime Video, audio | Onsite + extensive offsite; limited in-store (Amazon Fresh) | Amazon DSP; Amazon Marketing Cloud clean room |
| Walmart Connect | The clear no. 2 and fastest-scaling challenger: $6.4B global ad revenue in 2025, up 46%; the US's largest omnichannel retailer | Online + in-store transaction data linked across thousands of US stores | Search, display, CTV (via the Vizio acquisition), in-store screens, self-checkout and TV-wall placements | All three, incl. the most developed in-store offer | Walmart DSP (built with The Trade Desk) |
| Target Roundel | National multi-category reach; new leadership installed in 2026 with a mandate to double the media business within five years | Target Circle loyalty and guest purchase data | Onsite search and display, offsite programmatic and social, brand partnerships | Onsite + offsite; in-store developing | Roundel Media Studio self-serve; offsite via partners |
| Instacart Ads | Reaches 95% of North American households across 100,000+ store locations and 1,800+ retail banners; close to $1B in annual US ad revenue | Basket-level grocery purchase data spanning hundreds of retail banners | Sponsored product, display, shoppable video and recipes | Onsite + offsite; also powers other retailers' networks via Carrot Ads | Self-serve Ads Manager, expanded in 2026 |
| Kroger Precision Marketing | America's leading pure-grocery chain, with high purchase frequency | Loyalty-card data refined by in-house science unit 84.51° | Onsite search and display, offsite programmatic, CTV partnerships | Onsite + offsite; in-store pilots | Self-serve and managed; offsite via preferred DSPs |
| Costco | Warehouse-club membership base with unusually high average order values; a newer entrant scaling through 2025–26 | Deterministic member purchase data—every basket ties to a membership | Onsite sponsored products, offsite extensions to members | Onsite + offsite; in-store minimal | Managed service; offsite via partner platforms |
| Best Buy Ads | The leading US consumer-electronics specialist—a high-consideration category audience | Tech purchase history and My Best Buy loyalty data | Onsite, offsite, in-store screens; strong fit for non-endemic finance and telco brands | All three | Managed and self-serve; offsite via partner DSPs |
| Home Depot (Orange Apron Media) | The world's largest home-improvement retailer; DIY and pro-contractor audiences | Project-level purchase signals across DIY and trade customers | Onsite, offsite (Pinterest and Reddit integrations announced at its 2026 InFronts), in-store | All three | Managed and self-serve via Orange Apron Media |
Fig. Top RMNs (Sources: eMarketer H1 2026 retail media forecast; company earnings via Marketplace Pulse and AdExchanger; eMarketer December 2025 forecast).
| Channel | Primary objective | Typical placements | Targeting data | Best used for | Core KPIs |
|---|---|---|---|---|---|
| Onsite | Win the digital shelf | Sponsored listings, PDP/display and search | Real-time browsing, search and purchase data | Conversion, share of results and trade support | PDP views, add-to-cart rate, orders and ROAS |
| Offsite | Scale across the open web and CTV using retail data | Display, video, social and CTV via DSPs | Retailer audiences activated through external partners | Upper- and mid-funnel campaigns with sales attribution | Reach, VTR, assisted orders and iROAS |
| In-store | Influence shoppers at the shelf | Endcap and aisle screens, audio and checkout displays | Store, location, daypart and loyalty data | Product launches, promotions and seasonal features | Store lift, units per store and geographic iROAS |
Fig. Onsite vs offsite vs in-store: a quick chooser.
| Step | What happens | Data or tech involved | Outputs to watch |
|---|---|---|---|
| 1. Collect first-party data | Site/app events, POS, loyalty and CRM data are stitched into shopper profiles | Identity resolution, consent management, product catalogue | Addressable audiences; compliant data foundation |
| 2. Build the ad platform | Onsite, offsite and in-store inventory is packaged, with a buying interface for advertisers | Ad server/auction, DSP/SSP connections, screen CMS, clean room | Surfaces and specifications brands can activate |
| 3. Target audiences | Brands select segments, budgets, bids and campaign goals | Audience builder, lookalike modelling, eligibility rules | Reach, eligible impressions, forecast ROAS |
| 4. Deliver ads | Sponsored listings, display, video, CTV and in-store ads are served | Real-time decisioning, website/app integration, DOOH players | Viewable impressions, CTR, PDP views, store coverage |
| 5. Measure outcomes | Ad exposure is connected to online and in-store orders | Sales matching, incrementality tests, clean-room joins | Revenue, ROAS and iROAS, new-to-brand customers, basket lift |
Fig. Five-step RMN flow at a glance.
| Category | CDP | DMP |
|---|---|---|
| Primary data source | First-party customer data from owned channels, such as CRM, website activity, app behaviour, purchases, loyalty data, support interactions, and consent preferences. | Anonymous audience data from third-party providers, publisher networks, cookies, device IDs, behavioural signals, and data marketplaces. |
| Identity model | Builds persistent customer profiles linked to known or identifiable users. | Builds anonymous or pseudonymous audience segments based on inferred behaviour or attributes. |
| Data retention | Designed for long-term customer intelligence. Profiles can grow over time as new interactions, purchases, and preferences are added. | Usually built for short-term campaign activation. Audience data often loses value quickly as cookies, device signals, and intent signals expire. |
| Privacy readiness | Stronger fit for consent-based data governance when implemented correctly, because consent and preferences can be managed at the customer-profile level. | More exposed to privacy and provenance challenges because data often comes from third-party or indirect sources. |
| Audience activation | Activates known customer segments across email, SMS, CRM, web personalization, apps, analytics, paid media, and customer engagement platforms. | Activates anonymous audience segments mainly across DSPs, ad exchanges, ad networks, and programmatic buying environments. |
| Personalization capabilities | High. Supports customer-level personalization, lifecycle messaging, suppression rules, next-best-action campaigns, and retention strategies. | Moderate. Supports audience-level targeting but usually cannot deliver deep one-to-one personalization. |
| Integrations | Connects with CRM systems, marketing automation platforms, ecommerce tools, analytics platforms, customer support systems, data warehouses, and ad platforms. | Connects mainly with DSPs, ad exchanges, ad networks, publisher platforms, data providers, and media-buying tools. |
| Implementation complexity | Higher. Requires clean first-party data, identity resolution, event tracking, consent management, governance, and cross-team alignment. | Lower to moderate. Easier to activate for media campaigns, but dependent on data availability, integrations, and third-party signal quality. |
| Ideal use cases | Retention, personalization, customer journey orchestration, loyalty, churn prevention, customer lifetime value modeling, and first-party data activation. | Prospecting, awareness campaigns, audience extension, lookalike modeling, programmatic display, and top-of-funnel media activation. |
| Strategic role | Builds durable customer intelligence and supports long-term customer relationships. | Expands campaign reach and supports scalable media targeting. |
| Factor | Ad network | Ad exchange |
|---|---|---|
| Inventory access | Aggregates inventory from multiple publishers and packages it into audience, category, or placement-based bundles. | Provides access to impression-level inventory through automated marketplaces, often via DSPs and SSPs. |
| Transparency | Usually offers limited visibility into exact publishers, placement-level pricing, and network margins. | Offers stronger visibility into impressions, publishers, bids, supply paths, and performance data. |
| Pricing | Often uses fixed CPM, CPC, CPA, or managed package pricing. Easier to budget, but harder to audit. | Uses auction-based pricing through RTB, PMPs, or programmatic guaranteed deals. More flexible, but less predictable. |
| Targeting | Relies on pre-built audience, content, or vertical segments provided by the network. | Allows advertisers to activate first-party data, contextual signals, audience rules, and DSP-level targeting. |
| Control | Lower control because the network manages inventory selection, pricing structure, and delivery. | Higher control because advertisers can set bids, exclusions, frequency rules, data strategies, and supply preferences. |
| Technical complexity | Easier to use because the network abstracts much of the programmatic setup. | More complex because buyers need DSP access, reporting tools, optimization skills, and supply-chain oversight. |
| Best fit | Brands that need speed, simplicity, managed execution, or limited technical involvement. | Advertisers that need transparency, scale, first-party data activation, and greater control over media spend. |
| Platform | Best fit | Solution type | Model ownership | Refresh cadence | Validation approach | Implementation effort | Main limitation |
|---|---|---|---|---|---|---|---|
| Measured | Performance and retail teams wanting MMM plus causal validation | Managed effectiveness platform | Vendor-led, with client inputs | Varies by setup | Incrementality and causal calibration | Medium to high | Cost and setup may exceed smaller-team needs |
| Recast | Teams needing frequent planning and probabilistic MMM | Managed Bayesian MMM platform | Vendor-led | Frequent/weekly positioning | Causal MMM and GeoLift | Medium | Requires comfort with uncertainty and Bayesian outputs |
| Nielsen/Circana | Enterprise, CPG, retail, store-level analytics | Enterprise/consultative MMM | Vendor-led | Project- or program-based | Granular commercial and media analytics | High | Heavier process, less self-serve |
| Analytic Partners | Global enterprises needing planning infrastructure | Enterprise commercial analytics | Vendor-led with strategic partnership | Program-based | Benchmarks, forecasting, scenario planning | High | Significant investment and stakeholder commitment |
| Google Meridian | Technical teams wanting open-source model ownership | Open-source MMM | Client-owned | Client-controlled | User-defined validation and priors | High | Requires Python and analytics expertise |
| Meta Robyn | Digital-heavy teams with data science resources | Open-source MMM | Client-owned | Client-controlled | Calibration possible with lift/geo tests | High | Not turnkey; adoption burden sits internally |
| Mutinex | Marketing teams seeking accessible planning and investment guidance | Commercial MMM/planning platform | Vendor-led | Platform-based | Validation-led positioning | Medium | Buyers should verify exact testing features |
| Adobe Mix Modeler | Adobe-native enterprises | Enterprise stack-native MMM/MTA | Adobe environment | Platform-based | MMM plus event-level measurement where available | Medium to high | Strongest fit inside Adobe ecosystem |
Fig. Media mix modeling platform comparison.
| Approach | Best fit | Internal resources required | Speed to value | Strengths | Trade-offs | Example tools |
|---|---|---|---|---|---|---|
| Open-source MMM | Teams with data science capability and appetite for model ownership | High | Medium to slow | Flexible, transparent, lower software cost | Requires technical build, maintenance, interpretation, and governance | Google Meridian, Meta Robyn |
| Managed MMM | Teams that want expert support without building from scratch | Medium | Medium | External expertise, structured process, lower internal lift | Vendor dependency, variable transparency, recurring fees | Recast, Measured, Circana/Nielsen-style services |
| Enterprise MMM platform | Large organizations with complex portfolios and executive reporting needs | Medium to high | Medium | Governance, scenario planning, stakeholder workflows, planning infrastructure | Higher cost, heavier implementation | Analytic Partners, Adobe Mix Modeler |
| Hybrid MMM plus experiments | Teams that want MMM calibrated by causal testing | Medium | Medium | Stronger validation, better confidence in recommendations | Test design and budget required | Measured, Recast GeoLift, Mutinex-style validation workflows |
| Stack-native MMM | Organizations already invested in a large martech ecosystem | Medium to high | Medium | Easier integration with existing data and planning systems | Less attractive outside that ecosystem | Adobe Mix Modeler |
Fig. Which MMM approach fits your organization?
| Requirement | Minimum expectation | Why it affects model quality | Red flag |
|---|---|---|---|
| Historical media spend | Ideally 18–24 months of consistent spend data | Helps the model detect patterns across time, seasonality, and channel changes | Spend history exists only by platform screenshot or invoice |
| Outcome data | Revenue, sales, leads, subscriptions, store visits, or other KPIs aligned by time period | MMM needs business outcomes, not only media outputs | Sales and marketing use different definitions of conversion |
| Channel taxonomy | Consistent naming across platforms, agencies, and finance systems | Prevents duplicated or misclassified spend | “Paid social” means different things in different reports |
| External variables | Seasonality, pricing, promotions, holidays, macro factors, stockouts, competitor activity where available | Helps avoid giving media credit for non-media effects | Major sales events are missing from the dataset |
| Offline inputs | TV, radio, out-of-home, print, sponsorship, retail, trade, or store-level inputs where relevant | Prevents digital channels from absorbing unexplained offline impact | Offline spend lives in a separate budget file |
| Experiment history | Lift tests, geo tests, holdouts, or incrementality studies where available | Helps calibrate or validate model assumptions | No test history exists, and no one plans to create one |
| Internal owner | A marketing analytics, growth, or effectiveness lead responsible for adoption | Keeps MMM connected to decision-making | The vendor owns the entire process |
| Decision cadence | Monthly, quarterly, or annual planning rhythm | Determines how often outputs must refresh | MMM arrives after budget decisions are already made |
Fig. MMM readiness checklist.
| Evaluation area | Adobe CJA | Warehouse-centric analytics |
|---|---|---|
| Time to value | Faster if the business already uses Adobe systems | Slower if models, pipelines, and governance need to be built |
| Data ownership | Strong within the Adobe ecosystem | Stronger if the warehouse remains the source of truth |
| Flexibility | Structured, enterprise-ready journey analysis | More flexible for custom schemas, metrics, and modeling |
| Activation | Stronger when connected to Adobe Experience Cloud | Depends on the connected CDP, reverse ETL, CRM, and marketing tools |
| Total cost of ownership | Can be higher because of enterprise suite costs and add-ons | Can be lower at scale, but requires data engineering investment |
| Approach | Best for | Strengths | Limitations |
|---|---|---|---|
| CDP-first | Marketing teams that need faster audience activation | Built-in customer profiles, audience segmentation, consent tools, and channel integrations | Can become expensive at scale and may limit data-model flexibility |
| Warehouse-native | Mature data teams with strong analytics infrastructure | Greater data ownership, flexible schemas, deeper analysis, and stronger long-term control | Requires engineering support and more setup before marketers can activate insights |
| Connector-based | Teams that need to connect existing tools quickly | Faster implementation, useful for reporting, and lower initial technical lift | Can create fragile pipelines if identity, schema, and governance are not standardized |
| Maturity stage | Typical tools | Best for | Main limitation |
|---|---|---|---|
| Starting out | Platform-native lift studies (Meta Conversion Lift, Google geo experiments) | Small teams running first tests on a single platform | Self-reported and confined to one environment |
| Scaling | Independent platforms (Haus, Measured, Rockerbox) | Mid-market teams testing across several channels | Requires data integration and analytical capacity |
| Mature and cross-channel | Marketing intelligence suites, Google Ads Data Hub, BI platforms | Large teams running always-on, cross-channel measurement | Higher complexity and resourcing demands |
Fig. Matching incrementality tools to testing maturity.
| Design parameter | What it controls | A weak setup looks like | A strong setup looks like |
|---|---|---|---|
| Conversion volume | Statistical power | Too few conversions to detect the target lift | Volume sized to the minimum detectable effect |
| Test duration | Signal quality | Ends before delayed conversions land | Long enough to capture the full conversion window, no longer |
| Control group quality | Comparability | Control differs systematically from exposed | Control closely matches exposed on key traits |
| Audience separation | Freedom from contamination | Overlap and cross-channel leakage | Clean, enforced split between groups |
| External activity | Confounders | Promotions or launches run across both groups | Test window kept free of competing changes |
Fig. How test design choices affect reliability.
| Method | Measurement accuracy | Implementation complexity | Required scale | Privacy readiness | Best-fit channels | Recommended use |
|---|---|---|---|---|---|---|
| Geo holdout | High when markets are well matched | Moderate to high | Large | Strong — no user-level data needed | CTV, audio, OOH, upper-funnel video | Validating brand and upper-funnel spend at scale |
| Audience split | High within the platform's own walls | Low | Moderate to large | Moderate | Paid social, search, high-volume display | Fast reads on lower-funnel digital channels |
| Platform lift study | Moderate — self-reported by the platform | Low | Moderate | Moderate | Single-platform social and search | A quick first read, to be externally validated |
| Synthetic control | Moderate to high, depends on model quality | High | Moderate | Strong | Cookieless and sparsely tracked inventory | Privacy-first environments without clean holdouts |
| Ghost bidding | High within programmatic supply | High | Moderate to large | Strong | Programmatic display, video, native | Clean control groups inside live DSP buys |
Fig. Incrementality testing methods compared.
| Dimension | Walled environments | Open environments |
|---|---|---|
| Data access | Held inside the platform | Available to the advertiser |
| Measurement | Defined by the platform | Defined on common, comparable terms |
| Cross-platform comparison | Difficult by design | Supported |
| Pricing transparency | Limited | Visible end to end |
| Control over spend | Constrained | Retained by the advertiser |
Fig. Visibility in walled and open ecosystems.
| Pillar | What it provides | What its absence looks like |
|---|---|---|
| Unified data collection | One consistent source of marketing data | Conflicting numbers and manual reconciliation |
| Cross-channel measurement | Channels measured in relation to each other | Each channel claiming the same conversion |
| Consistent reporting frameworks | Shared KPIs and definitions | Meetings that argue over whose figure is right |
| Forecasting and predictive intelligence | A forward view of performance | Plans built by extrapolating last quarter |
| Actionable decision support | Insight turned into a next move | Dashboards admired, then ignored |
Fig. The five pillars of marketing visibility
| Capability | What it answers | What it leaves open |
|---|---|---|
| Marketing reporting | What happened, by channel and period | Why it happened, and what the numbers mean together |
| Marketing measurement | How much each activity produced | How the activities connect and influence one another |
| Marketing visibility | How channels, spend and outcomes relate, and what drives them | Little, structurally — though it relies on the quality of the first two |
Fig. Reporting, measurement and visibility compared
| Dimension | Walled garden | Open measurement |
|---|---|---|
| Data ownership | Platform | Advertiser |
| Attribution logic | Platform-defined; differs across vendors | Advertiser-defined; consistent across channels |
| Cross-platform visibility | Limited by design | Built into the architecture |
| Identity resolution | Within-platform only | Spans channels and devices |
| Reporting transparency | What the platform chooses to expose | Full inspection of inputs and methodology |
| Vendor dependency | High | Low |
Fig. Walled gardens vs open measurement at a glance
| Method | Question it answers | Data type | Best for | Key limitation |
|---|---|---|---|---|
| Attribution | Which touchpoints contributed to a tracked conversion? | User-level, digital | Daily and weekly tactical optimization | Misses offline; correlation, not causation |
| Marketing mix modeling | How does each channel contribute to total business outcomes? | Aggregated, historical | Strategic budget allocation across all channels | Slower cadence; needs years of consistent data |
| Incrementality testing | What additional outcome did this specific campaign cause? | Test vs control groups | Validating channel and campaign impact | Narrow scope; periodic rather than continuous |
Fig. How attribution, MMM, and incrementality compare
| Layer | Purpose | Primary capabilities |
|---|---|---|
| Reliable data collection | Capture clean, consented, cross-channel signal | First-party data infrastructure, server-side tagging, identity resolution |
| Smarter measurement models | Convert raw signal into business meaning | Attribution, MMM, incrementality testing |
| Centralized reporting infrastructure | Deliver one shared view of performance | Unified dashboards, governed data pipelines, standardized metrics |
| AI-driven forecasting | Predict and prioritize decisions before they're made | Scenario modeling, automated budget allocation, predictive analytics |
| Automated insights and workflows | Reduce manual analysis through pattern detection | Anomaly detection, AI-generated commentary, alerting |
| Governance and continuous testing | Keep the system honest as data and behavior change | Tracking audits, naming conventions, structured experimentation |
Fig. The six layers of a scalable marketing measurement framework.
| Where attribution breaks | What it looks like in reports | Practical response |
|---|---|---|
| Fragmented journeys | Conversions logged as "direct" or "unattributed" | Map real journeys; widen lookback windows to match the sales cycle |
| Privacy signal loss | Tracked conversions undercount actual sales | Invest in first-party data and consented, server-side measurement |
| Walled-garden double-counting | Summed platform conversions exceed recorded sales | Audit cross-platform overlap; measure on neutral, independent terms |
| Last-click bias | Branded search and retargeting look disproportionately strong | Adopt multi-touch and data-driven models; validate with incrementality |
| Method over-reliance | One dashboard treated as the single source of truth | Combine attribution with MMM, incrementality, and journey analysis |
Fig. Common attribution failures and how to operationalize the fix.
| Approach | Primary purpose | Key strengths | Main limitations | Best-fit use case | Business question it answers |
|---|---|---|---|---|---|
| Attribution | Assign credit to touchpoints in tracked digital journeys | Fast, granular, good for in-flight optimization | Misses offline and incremental effect; correlation not causation; degraded by signal loss | Day-to-day digital channel and campaign optimization | Which tracked touchpoints preceded this conversion? |
| Incrementality testing | Measure the causal lift a campaign produced | Most rigorous read on cause; isolates true contribution | Narrow scope per test; needs controlled conditions; periodic, not continuous | Validating whether a channel or campaign genuinely adds value | What outcome did this activity actually cause? |
| Marketing mix modeling (MMM) | Estimate each channel's contribution to total outcomes | Captures offline, brand, and cross-channel effects; unaffected by signal loss | Slower cadence; needs years of consistent data; aggregate not user-level | Strategic budget allocation across the full mix | How should total budget be split across all channels? |
| Customer journey analysis | Map the real sequence and interaction of touchpoints | Reveals how channels work together; exposes gaps and friction | Descriptive rather than causal; data-intensive to maintain | Understanding journey structure and experience | How do buyers actually move toward a purchase? |
| Unified marketing measurement | Combine the above into one decision framework | Triangulates strengths; reduces single-method blind spots | Requires governance, integration, and analytical maturity | Enterprise measurement and forecasting at scale | What is the most reliable overall read on performance? |
Fig. How modern measurement approaches compare.
| Model | How credit is assigned | What it tends to over-reward | Where it falls short |
|---|---|---|---|
| Last-click | All credit to the final touchpoint | Branded search, retargeting, direct traffic | Ignores every earlier interaction entirely |
| First-click | All credit to the first touchpoint | The initial awareness channel | Ignores everything that closed the deal |
| Linear | Equal credit across all touchpoints | Nothing in particular; flattens everything | Treats a glance and a demo as identical |
| Time-decay | More credit to touches nearer conversion | Late-funnel channels | Undervalues early demand creation |
| Data-driven | Algorithmic weights from historical patterns | Whatever the model's data favours | Correlation, not causation; needs volume; opaque |
Fig. How common attribution models assign credit
| Team profile | Deal volume | Cycle length | Recommended foundation | Key platform requirement |
|---|---|---|---|---|
| Low-volume, high-value | Tens per year | Long | Pipeline influence + self-reported + win/loss | Account-level stitching |
| Mid-market, growing | Hundreds per year | Medium to long | Position-based or time-decay, account-level | Configurable windows |
| High-volume, mature data | Thousands per year | Mixed | Data-driven, validated against incrementality | Neutral, integrated intelligence platform |
| Enterprise, multi-segment | High, varied | Nine to eighteen months | Stage-specific models + MMM + incrementality | Per-segment windows, platform independence |
Fig. Attribution platform selection by team profile
| Method | Question it answers | What it sees | B2B strength | B2B limitation |
|---|---|---|---|---|
| Attribution | Which touchpoints were present before conversion? | Trackable individual touches | Tactical channel optimization | Correlation only; misses the dark pipeline |
| Incrementality | What would have happened without this channel? | Lift vs. a held-out control | Exposes demand-harvesting channels | Needs full-cycle test duration |
| Marketing mix modeling | Where does the next budget dollar work hardest? | Aggregate spend vs. outcomes, including offline | Privacy-durable, strategic budget view | Needs volume; thin signal in low-deal-count B2B |
Fig. Attribution vs. incrementality vs. MMM.
| Model | Business question it answers | Minimum data maturity | Ideal sales-cycle length | Primary blind spot | B2B fit |
|---|---|---|---|---|---|
| First-touch | Which channels source net-new awareness? | Low | Any | Ignores all mid- and late-cycle influence | Limited |
| Last-touch | What is present at the moment of conversion? | Low | Short | Ignores everything before the final touch | Poor |
| U-shaped | What drove first contact and lead conversion? | Medium | Short to medium | Underweights sustained nurture | Good for MQL-focused teams |
| W-shaped | What drove first contact, lead, and opportunity? | Medium | Medium to long | Underweights touches between anchors | Strong for opportunity-focused teams |
| Time-decay | Which recent interactions reflect buying intent? | Medium | Medium to long | Sensitive to decay-rate misconfiguration | Strong if tuned to cycle length |
| Data-driven | What does our own data say each touch is worth? | High | Any, given volume | Unstable below high conversion volume | Strong at scale, risky below it |
Fig. B2B attribution model decision matrix.
| Organization | Measurement conditions | Most suitable platform type | Priorities |
|---|---|---|---|
| Emerging ecommerce brand | Concentrated digital spend, limited analysts, Shopify or marketplace revenue | Accessible geo-testing or ecommerce-focused platform | Simple setup, power guidance, direct sales integrations, practical recommendations |
| Scaling multichannel brand | Social, search, CTV, retail, and offline outcomes | Hybrid incrementality and MMM platform | Cross-channel data, geo tests, halo measurement, scenario planning |
| Large enterprise | Many brands, agencies, countries, and data owners | Managed enterprise measurement ecosystem | Governance, experiment calendar, model calibration, documentation, executive reporting |
| Advanced analytics team | Strong internal data science and custom business requirements | Configurable platform or internal framework | Data access, transparent methods, APIs, model export, reproducibility |
| Attribution-led performance team | Mature daily reporting but weak causal validation | Testing layer integrated with MTA and MMM | Comparison of attributed and incremental ROAS |
| CTV-heavy advertiser | Household exposure, cross-device conversion, stores or call centers | Geo and omnichannel measurement platform | Aggregate outcomes, offline integration, regional power, cross-publisher comparison |
| Privacy-constrained business | Limited identifiers and incomplete event paths | Geo testing, MMM, or aggregate causal inference | PII minimization, clear aggregation thresholds, stable business data |
Fig. Incrementality platform fit by organizational maturity