What Is Programmatic Advertising? Types, Costs & How It Works (2026)

Programmatic advertising means software buys your ads for you—one impression at a time, in a split-second auction, choosing who sees the ad based on data rather than a person negotiating a placement weeks in advance.

Say you spend a few minutes on a running-shoe site and then open a news article. Somewhere in the two hundred milliseconds before that page finishes loading, the publisher describes the empty ad slot, dozens of buyers decide whether you are worth bidding on, one of them wins, and the shoe brand's ad appears. Nobody phoned anybody. Nobody signed anything. That is programmatic.

Below: the four buying types, what each costs, the platforms involved, and how to run it without wasting budget.

TL;DR: Programmatic in 60 seconds

What it is. Automated buying and selling of digital ad space. Software decides, per impression, whether to bid, how much to pay, and which creative to show.

The four types.

• Open auction (RTB)—anyone can bid, prices move in real time, widest reach.
• Private marketplace (PMP)—invite-only auction on inventory a publisher has set aside.
• Preferred deal—a fixed price and first refusal on each impression, with no volume commitment.
• Programmatic guaranteed—a reservation: agreed price, agreed volume, guaranteed delivery.

What it costs. You pay per thousand impressions (CPM). Standard display sits at the low end, connected TV at the high end, and roughly a fifth to a third of what enters a buying platform goes on fees before any of it reaches a publisher. Realistic entry budgets start in the low thousands per month.

Who it is for. Anyone buying digital media beyond search and social—from a regional retailer running streaming ads to a global brand coordinating display, video, audio, and street screens on one plan.

The acronym key

Programmatic has more initialisms than any other part of media.

What is programmatic advertising?

Programmatic advertising is the automated buying and selling of digital ad inventory, executed by software that decides in milliseconds whether to bid on a single impression, how much to pay, and which creative to serve. Most trades happen through real-time bidding auctions, though automation also powers private marketplace and guaranteed deals. In short: no manual insertion orders, and decisions are made at the level of the individual impression.

The money involved is not small and it is still speeding up. US programmatic ad revenue reached $162.4 billion in 2025, up 20.5% year over year—an extra $27.6 billion of spend in a single year, and a faster growth rate than the 18% recorded the year before.

How programmatic evolved

It began in the early 2010s with open-exchange bidding for display, then expanded into private marketplaces, preferred deals, and automated guaranteed arrangements that secured premium placements on negotiated terms while still running through programmatic pipes. Today it accounts for over 90% of US digital display spend, and it has long since moved past banners into video, connected TV, audio, and public screens.

The 4 types of programmatic advertising

Programmatic is not a single auction. It is a set of transaction types that trade scale, price control, and certainty against each other. All four run on the same infrastructure—deal IDs, floors, priorities—but they differ in who can access the inventory, how the price is set, and whether delivery is guaranteed.

Open auction (real-time bidding)

The open exchange is what most people picture when they hear "programmatic." An SSP broadcasts a bid request for one impression; eligible DSPs evaluate it against their targeting and pacing rules and return a bid within milliseconds. The exchange clears the auction—usually first-price now—against the publisher's floor and any deal priority, then returns the winning ad markup to the ad server. It offers the broadest access and the most fluid pricing, with visibility into domains, apps, and supply paths through ads.txt, sellers.json, and the SupplyChain object.

💡 Read next: Programmatic vs. real-time bidding (RTB): what advertisers need to know.

Private marketplaces

A PMP is an invitation-only deal in which one or more publishers expose specific inventory to a chosen set of buyers. Trading is still programmatic, usually a private auction with dynamic pricing, but the seller controls who bids, what is included, and the floors and suitability rules. PMPs are typically packaged by context, audience, or format—a connected TV bundle across premium apps, for instance—and traded through deal IDs, which gives buyers clearer placement quality and often higher win rates than the open exchange.

Preferred deals

A preferred deal is a one-to-one agreement at a pre-negotiated fixed CPM that gives the buyer first look at a publisher's inventory. The buyer takes or passes on each eligible impression; anything passed over flows down to a private or open auction. There is no delivery guarantee, but the price is stable and the priority sits above general auction demand.

Programmatic guaranteed

Programmatic guaranteed, sometimes called automated guaranteed, is a reservation. Buyer and seller agree upfront on price, volume, dates, and targeting, and the platform automates trafficking and delivery. There is no auction for that supply—impressions are held for the deal and paced to meet the guarantee, with cancellation and makegood terms written into the agreement. Because it runs on the same rails, it keeps the reporting and convenience of programmatic while offering the certainty of a direct booking.

How much does programmatic advertising cost in 2026?

Two questions sit underneath this one. What does a thousand impressions cost, and how much of what you spend actually buys impressions at all? The second matters more, and almost nobody publishes it.

The fee stack

When a dollar enters a DSP, it does not arrive at a publisher intact. The Association of National Advertisers established the shape of the problem in its 2023 supply-chain study: transaction costs—DSP and SSP fees—took 29 cents of every dollar, leaving 71 cents to reach publishers, of which a further 35 cents was lost to impressions that were fraudulent, unmeasurable, non-viewable, or served on made-for-advertising sites. What survived, roughly 36 cents on the dollar, the ANA called TrueAdSpend.

The picture has improved since, and it is now measured quarterly rather than annually. In the first quarter of 2026, the market-level TrueAdSpend index stood at 43.3%, recovering after volatility in late 2025—and the ANA attributed the recovery mainly to better delivery quality rather than to cheaper transactions.

The gap between buyers is now wider than the gap between years. In the fourth quarter of 2025, advertisers enforcing disciplined quality controls converted 56.7% of programmatic spend into qualifying impressions, against 37.5% for lower performers. The same benchmark found that optimizing toward quality-adjusted metrics rather than headline CPM delivered close to 40% reductions in cost per conversion, in cases where the nominal CPM actually went up.

That is the single most useful thing a buyer can know about programmatic pricing: the cheapest CPM is frequently the most expensive outcome.

What a cleaner supply path is worth

Kimberly-Clark applied log-level data to its supply-path decisions and cut CPMs by 20% while reducing the number of sites it bought from nearly 29,000 to about 1,700. Fewer paths, better inventory, lower price. This is what supply-path optimization does when it is done properly, and it is why we built Smart Supply—a DSP-agnostic layer that applies supply selection and optimization to enforce direct, brand-safe, IVT-filtered routes to inventory.

💡 Read more: Supply-side optimization with AI—a walkthrough for auditing supply paths, prioritizing higher-quality exchanges, and reducing fees.

Indicative CPM ranges by channel

Published CPM benchmarks come from platform reporting and vendor rate cards rather than audited industry data, and they vary widely by audience, geography, and inventory quality. Treat the following as directional bands for planning, not as quotes.

Realistic minimum budgets

Managed programmatic partners generally look for a few thousand dollars a month before a campaign can generate enough impressions to optimize against. Self-serve platforms go lower. Below roughly $3,000 a month, most campaigns cannot reach statistical significance on any single test, which means you are buying impressions rather than learning anything.

💡 Read more on the metrics that decide whether that spend was worth it: 15 digital marketing metrics and KPIs to measure performance.

How programmatic advertising works

Programmatic is a millisecond handshake between supply and demand. Each time a page, app, or streaming session loads, the publisher describes the impression, buyers evaluate it, an exchange clears the auction, and the winning creative renders—followed by verification and logging. That loop then feeds back into bidding and pacing so the next impression is decided on fresher signals.

1. A person loads content. The publisher's page, app, or stream calls its ad server and, where used, a header-bidding wrapper. The SSP assembles a description of the opportunity: slot size and format, page or app bundle, device and network details, geography, consent strings, and any publisher-provided identifiers. This is encoded as an OpenRTB request and sent to one or more exchanges under tight timeouts.
2. Exchanges fan out the request. Each exchange validates it, attaches deal information, supply-chain metadata, and brand-safety signals, then broadcasts it to eligible DSPs. Depending on the environment, the request may carry app or channel metadata for connected TV, advertising identifiers where permitted, and content signals for contextual matching.
3. The DSP decides. Inside the bidder, the DSP matches the request against active campaigns and their constraints—audience membership, contextual rules, geography, frequency caps, pacing, remaining budget. It can consult first-party segments, data management platform or customer data platform audiences, and brand-safety lists, then score the impression and compute a price. The response carries the bid, a creative that fits the spec, the advertiser domain, and tracking endpoints.
4. The auction clears. The exchange runs the auction, enforcing publisher floors and deal priorities. If the impression is tied to a private marketplace, preferred deal, or guaranteed arrangement, those rules apply before or instead of open bidding. The winner is cleared, win and loss notifications fire, and ad markup returns to the publisher's ad server. On connected TV with server-side insertion, the ad is stitched into the stream instead.
5. Render and measure. As the creative loads, impression beacons fire and third-party verification begins. Video players emit quartile events; fraud and suitability vendors may run post-bid checks. In server-side environments, measurement beacons are proxied so completion data still registers accurately.
6. The feedback loop. Log files from exchanges, SSPs, and DSPs flow back into reporting. Conversions arrive by pixel or server-side event, pacing updates, and models retrain on fresh outcomes. Supply authenticity is monitored through ads.txt, sellers.json, and the SupplyChain object. These signals shape the next auction decision.

Programmatic advertising examples

Three campaigns, three channels, three different jobs—each with a reported outcome and something worth copying.

The Economist: contextual targeting to grow subscriptions

The team profiled a segment of intellectually curious non-readers, mapped where those people spent time online, and identified the topics that signaled receptivity. Creative was written to mirror those contexts: short, provocative lines that clicked through to article excerpts and subscription offers. Bidding prioritized pages whose semantic cues matched the ad theme, so the same person could encounter very different messages depending on what they were reading. The campaign produced 64,000 paid subscriptions with a 25:1 payback on estimated lifetime value, and surfaced more than five million previously unseen, retargetable users.

What to copy: write the creative for the context, not the audience segment. The targeting and the copy were designed together, which is why it worked.

💡 Read more: The Economist: raising eyebrows and subscriptions.

Kellogg: changing the currency to viewable impressions

Kellogg moved buying into a programmatic stack and made viewable impressions the unit it optimized against, rather than impressions served. With unified control of frequency, pacing, and creative rotation, impression quality rose and wasted spend fell. Reported viewability moved into the 70–80% range, and audience matching improved two to threefold over previous methods.

What to copy: change the metric you optimize against before you change anything else. The buying didn't get smarter; the definition of success got stricter.

💡 Read more: Kellogg dishes up offline sales with programmatic buying.

Missing People: live-updating public screens

The UK charity connected a live case feed to digital out-of-home networks so screens could update by geography and time, showing the most relevant appeal wherever it mattered most. Proximity to transport hubs, commuting peaks, and local events determined which creative variant appeared on which screen. Proof-of-play logs and response tracking showed reported success rates on featured cases rising from around 50% to about 70%.

What to copy: treat the creative as a data feed rather than a fixed asset. Anything that changes hourly—stock, weather, price, availability—can drive a screen the same way.

💡 Read more: How charity Missing People is saving lives by moving to programmatic.

A B2B example: buying fewer places, better

Kimberly-Clark's supply-path work, described in the cost section above, is the clearest published example of the operator-side version of this discipline. It is not a targeting story or a creative story. It is a procurement story, and it produced a cheaper CPM and a shorter list of publishers at the same time. For considered-purchase and B2B advertisers, where audiences are small and waste is expensive, this is usually the highest-return project available.

Is Google Ads programmatic?

Partly, and the distinction matters when you are comparing platforms.

Many Google Ads placements are bought programmatically through Google's own systems—the Display Network, YouTube, and several Discovery formats all transact through automated, auction-based delivery. Display & Video 360 is a full demand-side platform, used to buy open-web inventory, connected TV, audio, and out-of-home well beyond Google's own properties.

Search ads are the exception people usually mean. They run on their own auction, priced per click against keyword intent rather than per impression against audience and context, and they are not part of the programmatic display market as the industry measures it.

The practical difference is access and control. Buying inside a single platform gives you that platform's inventory and that platform's reporting. Buying through an independent DSP gives you inventory across many sellers, log-level data, and the ability to compare supply paths—which is why advertisers running meaningful budgets tend to use both rather than choosing.

Programmatic advertising formats

Programmatic spans every major digital channel and, increasingly, physical screens. Total US display revenue—which includes banner, rich media, sponsorship, and native—reached $81.6 billion in 2025, growing 9.8%, the slowest growth of any format even as the money moved toward automated buying within it.

Display

Display covers graphical ad units and rich media rendered within page content or app views across the open web and mobile apps. Buying is overwhelmingly automated, with impressions auctioned in milliseconds and delivered through standard ad servers and SDKs. Creative ranges from static images to HTML5 and expandable units; viewability and suitability are verified through third-party measurement and supply-chain transparency tools.

Within programmatic mixes, non-video display has slowly ceded share to video and native, but it remains the broadest and most varied inventory pool available for addressable reach.

💡 Learn more: What is programmatic display advertising?

Video and connected TV

Programmatic video spans in-stream units on web and mobile, out-stream placements in articles and feeds, and connected TV inventory inside streaming apps on smart TVs and OTT devices.

Video is the fastest-growing format in digital advertising, reaching $78.0 billion in 2025 on 25.4% growth, and 26.5% of all US digital ad revenue. Connected TV is the engine of that growth: US CTV ad spend is forecast at roughly $37.95 billion in 2026, up about 15%.

Technically, video transacts through VAST and VMAP, with server-side ad insertion common on connected TV to protect stream quality. Identity is usually resolved at household or device level, and completion rates run higher than web video because viewing is lean-back. Inventory spans premium broadcasters, streamers, and long-tail apps, available through open auctions, private marketplaces, and guaranteed deals.

💡 Read next: Programmatic video advertising explained | OTT vs CTV: key differences and similarities | TV media buying: definition, process, costs, and strategies

Native

Native ads are rendered to match the host environment, using the OpenRTB Native specification—headline, image, brand, call to action—assembled into in-feed units, content recommendation tiles, and sponsored listings. The design aim is editorial harmony, which tends to produce longer attention and deeper on-page engagement than standard banners.

Programmatic native trades through the same pipes as display, with native-specific creative fields and rendering rules set per publisher template. Note that industry revenue reporting folds native into total display rather than breaking it out separately, so native-specific share figures should be treated with care.

💡 Dig deeper: 10 reasons to use native advertising

Audio

Programmatic audio covers ads inserted into music streaming, digital radio, and podcasts, delivered through client- or server-side dynamic insertion into live and on-demand content.

US digital audio revenue reached $8.4 billion in 2025, growing 10.2%, with podcast advertising a fast-growing component at $2.9 billion, up 17.6%. The channel remains structurally underbought: listeners spend 31% of their media time with audio while the channel takes around 9% of ad spend.

Targeting typically covers genre, show, mood, location, and device. Creative centers on 15- and 30-second spots, sometimes with companion banners. Measurement blends impression logs with brand-lift and site-visit methods suited to screen-off consumption.

Digital out-of-home

DOOH covers digitized roadside billboards, street furniture, transit screens, and place-based networks in retail, gyms, offices, and airports—all now addressable programmatically. Impression delivery is modeled from sensor and mobile panel data alongside proof-of-play logs, with buys transacted against impression goals or share of voice across screen loops.

US programmatic DOOH spend is projected to reach $1.23 billion in 2026, following growth of 23% in 2025 and 34% the year before (eMarketer). Growth is driven by dynamic triggers—time, weather, events—and flexible dayparting across national networks.

Creative has to respect out-of-home constraints: large type, high contrast, very short copy, with variants rotating by location or time.

The platforms behind programmatic

Programmatic runs on a small set of connected systems. On the buy side, DSPs decide what to bid and what to show. On the sell side, SSPs package and price inventory. In between, exchanges run the auctions, while data platforms supply the audience and measurement fuel.

Demand-side platforms

A DSP is the buyer's control room: software that ingests bid requests from exchanges, evaluates them against campaign rules, and decides whether to bid, at what price, and with which creative. At scale, bidding platforms routinely process more than one million bid requests per second.

A DSP centralizes audience targeting, budgets, pacing, frequency management, and creative delivery, and connects to first-party segments, contextual signals, and verification vendors. Modern platforms also handle identity resolution, log-level reporting, and optimization models that learn from impression, click, and conversion feedback.

Supply-side platforms

An SSP packages a publisher's inventory, applies business rules—floors, blocks, deal priority—and exposes that supply to multiple demand sources through a unified auction. SSPs enforce ads.txt and sellers.json authorization, pass consent signals, and provide yield tools such as dynamic floors and deal discovery. They also transmit quality, viewability, and supply-path data so buyers and exchanges can make better-informed decisions.

💡 Read more: Rethinking the value proposition of DSPs in today's programmatic

Exchanges and data platforms

Ad exchanges connect DSP demand with SSP supply, normalize bid requests through OpenRTB, and clear auctions under tight latency constraints. They enforce deal terms, propagate brand-safety and supply-chain objects, issue win and loss notifications, and return ad markup once a winner is determined.

Data platforms power targeting and measurement:

• DMPs aggregate mostly anonymous advertising identifiers into audience segments for activation, and are strongest at taxonomy and lookalike modeling.
• CDPs unify person-level first-party data—CRM, transactions, site and app events—with consent, creating durable profiles used for targeting, suppression, and measurement.
• Data clean rooms allow privacy-safe collaboration between brands, publishers, and platforms by matching datasets without exposing raw personal data. They are used to activate first-party audiences, deduplicate reach and frequency, and measure incrementality under stricter controls.

Benefits of programmatic advertising

The advantages all trace back to the same thing—deciding one impression at a time instead of one placement at a time.

Strong personalization programs have been shown to lift revenue by around 5–15% and marketing return by 10–30%, which is why impression-level targeting and dynamic creative optimization matter inside programmatic buys—and why retargeting remains one of the most reliable applications of the technique.

Transparency here is not a slogan but a set of specific standards. ads.txt lets publishers declare authorized sellers. sellers.json discloses who is selling or reselling inventory. The Open Measurement SDK standardizes third-party viewability and verification across environments. Together they let buyers verify supply chains rather than take them on trust.

The democratizing effect is real too. Premium inventory that once required large minimum commitments is now accessible to a regional advertiser bidding only on impressions that meet its criteria.

💡 Read more on why brand safety has to start on the supply side rather than with DSP filters: Safety in programmatic advertising.

Challenges and how to solve them

Scale and speed create their own problems: the same automation that finds an audience in milliseconds will find a bot just as fast.

Invalid traffic and fraud

The problem. Automated buying attracts bots, spoofed domains, and resold inventory that adds fees without value—and the underlying traffic picture has worsened. AI-driven bots now account for over half of global web traffic, with close to 40% classified as malicious, which complicates attribution, audience modeling, and optimization alike.

Industry standards have held the line where they are applied. Invalid traffic has stayed below 1% across TAG Certified Channels in the US for four consecutive years, and TAG's 2025 Impact and Compliance Report records €3.45 billion in European fraud savings, a 69% reduction against what would have been lost without those programs.

What works.

• Transact with certified partners. Prioritize exchanges, SSPs, and DSPs holding Certified Against Fraud status.
• Enforce supply authenticity. Require ads.txt, check sellers.json, and inspect the SupplyChain object to verify who is authorized to sell and every hop in the path.
• Filter pre-bid. Turn on IVT filters in your DSP and with verification vendors, maintain rolling exclusion lists, and review placement reports weekly.

Inventory quality and made-for-advertising sites

The problem. Waste from low-quality supply and opaque chains erodes return. MFA exposure had fallen to between 0.4% and 0.6% through 2025, but rose to 1.1% in the first quarter of 2026, with the ANA naming AI-generated filler as an emerging subtype that needs its own mitigation.

What works. Prefer direct paths to publishers and drop duplicative resellers. Buy sensitive campaigns through vetted marketplaces with viewability floors written into the deal terms. Demand log-level data and fee disclosure from platforms, and audit supply-path outcomes quarterly rather than annually.

Privacy and data regulation

The problem. Consent rules limit tracking and data sharing. Google's April 2025 decision not to deprecate third-party cookies in Chrome removed one disruption, and the company subsequently retired around ten Privacy Sandbox APIs in October 2025—but enforcement of GDPR and state-level US privacy law continues, and expectations keep rising.

What works. Capture first-party signals with a clear value exchange and unify them in a CDP. Expand contextual and cohort-level tactics where user-level identifiers are weak. Pair platform attribution with incrementality testing, shorten lookback windows where required, and document retention and access controls.

💡 Learn more: In a cookie-less world: new challenges and opportunities

Automation outrunning creative

The problem. Algorithms optimize delivery, but weak creative stalls performance regardless. Research indicates creative can account for around 47% of sales contribution.

What works. Build a creative test plan alongside your bidding tests, with stated hypotheses, rotating variants, and holdouts. Use dynamic creative where the signals are reliable and cap variants at what you can actually learn from. Keep the brand and activation split in view; IPA analysis continues to support roughly a 60/40 tilt, adapted to category.

💡 Explore how generative AI is changing creative media.

Measurement and KPIs

Measurement in programmatic is a telemetry system wrapped around every impression. Ad servers, SDKs, and verification tags record exposure, interaction, and outcome events; platforms aggregate those into metrics that map to how people move from seeing an ad to acting on it. Standards bodies define the common denominators so the numbers mean the same thing across sites, apps, and devices.

Mapping KPIs to the funnel

The four layers run in the order the behavior does—seen, engaged, acted, and what it cost to get there.

• Exposure. Reach and deduplicated reach describe how many people were exposed at least once; frequency describes how often. Viewability applies a quality filter based on a shared definition (MRC criteria), so exposure counts reflect ads that could reasonably have been seen. Attention augments this with time- and behavior-based signals.
• Engagement. Qualified visit rate, depth or time on page, video quartiles and completion rate, and audio listens captured through standardized player events.
• Action. Conversions tied to business value and their efficiency ratios—CPA, customer acquisition cost, return on ad spend. Post-click quality metrics help separate low-intent clicks from real sessions.
• Efficiency. Working-media share expresses how much of each dollar reaches human, quality impressions after fees and filtration. This is the layer the cost section above is really about.

The four measurement approaches

Each answers a different question, and each is wrong in a different way.

Verification and viewability are standardized through the Open Measurement SDK, which gives third-party vendors a common way to observe whether display, video, and in-app ads met viewability criteria and to report fraud filtration consistently.

The pressure on all four approaches is increasing as signal fragments. The IAB's 2026 State of Data research estimates that improvements in AI-driven measurement could recover more than $26 billion in media investment over the next one to two years—a figure that describes how much is currently being misallocated for want of reliable measurement.

Keeping the system honest

Measurement runs against baselines. 

• Establish a reference period, collect comparable data weekly, and report movement relative to that baseline. 
• Standardize definitions up front—viewability rules, attribution windows, filtration—so results from different partners line up. 
• Calibrate through scheduled lift studies and routine model refreshes, and feed what you learn back into audience definitions, creative libraries, and supply decisions.

What changed in 2026

If you read this page a year ago, five things are different.

1. Digital advertising crossed a threshold. US digital ad revenue reached $294.6 billion in 2025, up 13.9%—a record, and notable because 2025 had no Olympics, World Cup, or US presidential election to inflate it.
2. Automated buying pulled further ahead of manual. Non-programmatic placements, excluding search, fell 13.9% to $18.0 billion. The direction of travel is no longer contested; the only question is what proportion of automated spend is well spent.
3. Streaming beat primetime at the upfronts. In 2026, US connected TV upfront commitments ($17.73 billion) exceeded primetime linear TV upfront commitments ($16.98 billion) for the first time. Note that definitions vary—the IAB's narrower CTV measure produces a much smaller number than eMarketer's—so compare like with like before quoting a market size.
4. Commerce media kept compounding. Commerce media revenue rose to $63.4 billion, up 18%, though that is slower than the 23% of the previous year as brands press retailers harder on incrementality. The category has also broadened past retail media networks into off-site and in-store environments.
5. Quality replaced cost as the main efficiency lever. The shift described in the cost section—that the spread between well-run and poorly run programmatic now exceeds the spread between good and bad years—is the most consequential change for anyone actually buying media.

The future of programmatic

Nothing below is a prediction. All four are already happening, unevenly.

Generative AI in creative and optimization

AI has moved from experiment to infrastructure: rapid concepting for variants, automated copy and visual swaps by segment, and predictive signals feeding bid strategies. McKinsey's most recent State of AI survey found 88% of organizations regularly using AI in at least one business function, while nearly two-thirds have yet to scale it across the enterprise—adoption is near-universal, embedding is not.

Expect your playbook to pair dynamic creative and versioning with clear human review standards and measurement that separates creative lift from bidding effects.

💡 Read our take: The future is now: how AI Digital embraces AI technologies

Convergence with television and out-of-home

Television and public screens are increasingly bought the same way as web and mobile. eMarketer projects connected TV to overtake linear TV advertising around 2028, at roughly $47 billion against traditional TV's $45 billion. Whether the crossover lands in 2028 or a year either side matters less than the direction: the money is moving to surfaces that come with audience, context, and timing controls built in, and planning is following it.

Sustainability in media buying

Carbon accounting has reached briefs and quarterly reviews. GroupM's omnichannel carbon calculator, built with Scope3 data, gives planners emission factors by channel and targets halving emissions per impression by 2030. Scope3's research indicates that lower-carbon supply paths can maintain or improve effectiveness when buyers select inventory carefully and cut wasteful hops—which is the same discipline as supply-path optimization, arriving from a different direction.

Programmatic as the default

Programmatic is already the standard for display and is spreading across every other channel. With streaming and public screens adopting the same infrastructure, most campaigns in 2026 default to automated, data-driven transactions, with manual buys reserved for exceptions. The teams that win will combine well-selected supply, clean measurement, and disciplined creative testing—treating programmatic as the operating model rather than a tactic.

Conclusion

Programmatic matters in 2026 because it buys media the way marketers actually plan: one impression at a time, with data deciding the who, where, and how much. It delivers scale across display, video and connected TV, native, audio, and public screens; it gives live control over targeting, price, and safety; and it has matured toward better transparency, cleaner supply paths, and stronger measurement.

The lesson of the last two years is that the gap between advertisers is now larger than the gap between years. Two brands can spend identical budgets through the same platforms and convert wildly different proportions of that money into impressions a human being might see. Which one you are is a matter of governance, not luck.

The checklist

‍

Twelve things to actually do:

• Center on first-party data. Codify consent and value exchange; unify identifiers in a CDP and sync durable segments to the DSP with clear suppressions and burn windows.

• Enforce clean supply. Prefer direct paths and require ads.txt, sellers.json, and SupplyChain disclosure on every deal.

• Use vetted access when quality matters. Private marketplaces with viewability and, where used, attention floors.

• Audit supply paths quarterly. Cut hops and partners that add cost without performance.

• Run a standing learning agenda. Short, written hypotheses for audience, frequency, and creative.

• Test in tight sprints. Simple A/Bs; promote winners to always-on budgets.

• Prove causality regularly. At least one incrementality test per quarter for major tactics.

• Make creative a performance lever. Refresh variants per segment, sequence messages, cap cross-channel frequency.

• Use dynamic creative selectively. Only where signals are strong, with human review before launch.

• Standardize measurement. One viewability and filtration spec, shared attribution windows across partners.

• Triangulate results. Platform attribution for direction, lift tests for causality, mix modeling for long-run contribution.

• Report against baseline. Tie optimizations to return and cost targets, not vanity metrics.

Talk to AI Digital. If you want a DSP-agnostic partner that pairs brand-safe supply with clear, cross-platform buying, we can help through our Open Garden model and managed service—built to align optimization with your KPIs and give you full visibility across channels.

Get in touch to request a programmatic health check: supply-path review, KPI map, and a focused test plan.