Programmatic Advertising Agency: What You Are Buying and What the Fee Structure Hides

Programmatic is the only line in your media budget where you buy an audience you cannot see, on inventory you never name, through a chain of intermediaries you will never meet. The invoice says "media". The number that matters is how much of that media became a real person looking at your ad on a page you would be comfortable being seen on. In Q1 2026, the better half of the ANA Programmatic Transparency Benchmark converted 54.0% of programmatic spend into qualified impressions. The weaker half converted 32.1%. Both cohorts were buying from the same exchanges, at roughly the same time, at roughly the same CPMs. Whatever separated them is not in your proposal, and it is not in the pitch deck either.

Sources and dating. The benchmark figures below come from the ANA Programmatic Transparency Benchmark series (Q1 2025 through Q1 2026) and the fee bands are third-party market data dated July 2026. Neither is anyone's rate card, including ours. Where a figure describes a published price, the source is named beside it so you can check it yourself.

Search this term and you will mostly find two kinds of page. Agencies describing themselves, and listing surfaces that rank firms by headcount. A directory subdomain ranks well here with no editorial judgment behind it, and at least one listicle ranks itself first among the agencies it reviews.

What it costs, in one paragraph. Published US bands put a flat programmatic retainer at $15,000 to $60,000 per month, or a commission at 10 to 20% of media spend, with a hybrid at roughly $10,000 plus 8 to 12%. Every one of those carries a media minimum, from $50,000 to $100,000 per month, and none of them includes the DSP seat fee or third-party data, which can add 15 to 30% to your media on their own.

What a Programmatic Advertising Agency Actually Manages

Strip the language away and programmatic advertising means buying display, video and audio inventory through automated auctions instead of through a media rep and a signed insertion order. The agency's job is to operate the buying seat, decide what to bid on, and keep the supply path from quietly degrading. In practice, that covers five things you are paying for.

• The DSP seat and the auction logic. Which demand-side platform you buy through, whose account the seat sits in, and the bidding strategy inside it.
• Deal construction. Private marketplace deals (PMPs), preferred deals and programmatic guaranteed lines, each with its own deal ID, floor price and publisher set.
• Audience and data. First-party segments, third-party data overlays, Customer Match, lookalikes, plus the consent basis for every one of them.
• Brand safety and verification. Exclusion lists, fraud and invalid traffic filtering, viewability thresholds and suppression rules for made-for-advertising (MFA) inventory.
• Measurement and reporting. Attribution, incrementality testing, and the placement-level reporting a buyer is entitled to read.

Notice what is absent from that list: the fee. Nothing about how the agency is paid tells you whether any of the five is done well. The quality of a programmatic engagement lives in the settings inside the seat, and the settings are the part you never see.

One more thing worth stating before the money. Programmatic display behaves nothing like paid search. If most of your intent sits in search queries, the buying decision, the fee structure and the reporting you should demand are different, and the trade-offs in PPC management services are the ones to read first.

The Fee Models and What Each One Rewards

The market runs on a handful of fee shapes, and the published bands for each (marketerhire.com's vetted-agency list, July 2026) are in the table below. Read each one as a threshold rather than a price. The fee percentage matters less to your budget than the media minimum that unlocks it.

Percentage of media spend

The most common shape in the market, and the easiest one to misread. The industry standard sits around 10 to 20%, but the source that publishes those bands says the shape matters more than the number. A commission on media spend rewards the agency for scale. Rewarding the agency for your outcomes is a different incentive, and it is not the one the contract creates. Push more budget through the same structure and the fee grows whether performance does or not.

The single question that matters here: does the percentage apply to gross media cost, or to the net that reaches publishers? Those two figures are not close. The ANA put publisher-share of spend at 36.5% in Q2 2025, then at 47.1% in November 2025 for buyers who enforced quality rules. A 15% commission on gross spend and a 15% commission on what actually arrived can differ by more than the commission itself.

Flat retainer

A retainer buys a fixed scope, and that is its strength. You know the monthly figure and you can hold the agency to a defined list of deliverables. It also creates the incentive you would expect: the cheaper the agency can deliver the scope, the better the margin.

That becomes a problem when the scope names activities instead of outcomes. If the retainer covers "campaign management" without specifying reporting cadence, log-level data access or the frequency of supply-path reviews, the flat fee protects the agency's margin instead of your budget. A fractional-expert variant sits at the edge of this model: $7,000 to $12,000 per month with no media minimum, buying senior judgment rather than a team, and usually leaving the DSP seats with you. That last detail is the one to read twice, because it is the only model in the set where the access rights come bundled.

Performance or outcome-based

Outcome fees are the most attractive in the pitch and the hardest to enforce. Everything depends on what the outcome is. Cost per qualified impression, cost per incremental conversion, and cost per acquisition are three different contracts, and only one of them is usually measurable on the data you control.

Before you accept an outcome-based fee, ask for the outcome's definition in writing, the data source that will measure it, and who arbitrates a dispute. If the answer routes through the agency's own dashboard, what you are buying is a reporting model with a variable fee attached.

Where the Inventory Markup Hides

The fee is the part of the cost you can read. The markup is the part you cannot, and it shows up in four places.

Take rate. The spread between what you pay and what the publisher receives. In Q2 2025 the ANA put $26.8bn of global programmatic budget in the wasted column for a single quarter, with only 36.5% of spend reaching publishers and waste itself up 34% to that figure. By November 2025 publisher share had risen to 47.1% among buyers who enforced quality rules, which is the part worth taking to your next meeting: the share of spend reaching publishers is a variable, and somebody has to move it.

Tech and seat fees. A DSP seat is a real cost, and it is commonly billed separately from the management fee. MarketerHire flags DSP seat fees as one of two SOW lines buyers routinely miss, and it is the one nobody questions at signing.

Third-party data overlays. The same source warns that third-party data costs "can add 15 to 30% to media on top of everything else". A media plan can grow by nearly a third without the agency's percentage moving at all, because the data bill is not media.

Quality drift. MFA exposure rose to 1.1% in Q1 2026 after sitting between 0.4% and 0.6% through 2025, and AI slop has been named as an emerging subtype at 1.3 to 2.4% of open web programmatic impressions. The counterpoint matters just as much: among buyers who enforced quality rules, MFA exposure fell to 0.39% in November 2025. Quality rules are the variable that moves your number, not budget size.

Two smaller signals worth carrying into your next meeting. A falling CPM is not automatically good news: market CPMs fell from $5.55 in Q4 2025 to $4.42 in Q1 2026, and your budget at a lower CPM simply buys more impressions, not better ones. And when a source presents CPM and CPC as an agency's "upfront price model", it has mislabeled media pricing as agency pricing.

What You Should Be Able to Audit, and Usually Cannot

The strongest access-rights language found in this market is short enough to memorize: working in your seat keeps the deal IDs, the audiences and the log-level history with you. Platform access and deal IDs stay in accounts you own. Webtonic publishes exactly that language and adds a budget floor of "mid five figures a month". What it does not publish is a single number for its own fee model. That combination is common, and it tells you where the industry is comfortable being clear.

Ask for these five rights before you sign, in the contract rather than in a slide:

• Seat ownership. The DSP seat sits in an account your organization controls, and it survives an agency change.
• Deal IDs on record. Every PMP and preferred deal ID is documented, so a new partner can pick the supply path up without rebuilding it.
• Log-level data (LLD). The granular impression-level file, not a summary export. Availability genuinely differs by platform: the TAG TrustNet LLD Register lists full log-level data with all required fields for Basis, no LLD support for Amazon Advertising, and Meta, TikTok and X as unknown. Log-level access is a checkable fact, not a claim.
• A sample placement report. One page, real data, your domains and your app list. Ask for it before signing, not in month three.
• The audit right, in writing. Once a year at minimum, with access to the seat and the invoices behind the plan.

One practical gate, and the one I would use first: ask every shop on your shortlist to audit 30 days of your current programmatic program before you sign. That request comes from the same source as the fee bands, and it is the cheapest test available. You learn what they find, how they explain the supply path, and whether they are willing to touch somebody else's account at all.

Programmatic Agency, DSP In-House or Managed Platform?

The in-house question has already been answered by the market, just not uniformly. Advertiser Perceptions' 2026 in-housing study found 64% of enterprise advertisers now manage at least one DSP in-house, up from 41% in 2023. At the same time, an ANA survey of 2026 in-house practitioners found that only 9% now name cost saving as the main reason for moving in-house. The decision has stopped being a price argument.

That reframes the three routes honestly. I run an agency, so I will say it plainly: these are not three versions of the same service.

• Agency. You buy expertise, seat access and a team without a hiring cycle. You accept that the settings live in someone else's hands unless the contract says otherwise.
• In-house. You buy control and continuity, and you pay for it in headcount. A senior programmatic lead costs $150,000 to $220,000 fully loaded, per a 2026 hiring analysis citing US labor data, with hiring cycles of 3 to 6 months. Tools, training and verification sit on top of that salary.
• Managed platform or self-serve buying. You buy a workflow that covers the parts of the job you can standardize. It works when your reporting and your content operations already live somewhere you control.

Most mid-market teams end up hybrid, and the honest split is by supply path. The channels where quality rules and log-level data exist get enforced, and the ones where they do not get monitored.

This is the layer where we sit, so judge the fit yourself. Allable does not trade programmatic inventory. What it does is keep your measurement on your side of the fence: connected paid accounts, search terms and cost data, plus the reporting you build your own dashboards from, so the numbers you audit are not the numbers the agency built for you. That feeds naturally into marketing attribution software thinking when you want to compare channels rather than platforms. Plans run Pro at EUR 99 per month, Business at EUR 199 per month and Scale at EUR 399 per month, with a 7-day trial that requires a card. If a vendor promises programmatic inventory at that price, the inventory is the product being sold.

What to Measure Monthly

One page, ten lines, no dashboard needed, and you can hold anyone to it. A report that cannot produce these is decoration.

1. Publisher share of spend. The ANA series is the benchmark to sit against, and it is public.
2. Qualified impression rate. Share of spend that became a measurable, viewable, fraud-free impression.
3. MFA share. Ask for the suppression rules in place and the exposure figure after them.
4. Effective CPM against the platform CPM. The gap between them is where fees hide.
5. Seat and tech fees as their own line. Never inside "media".
6. Third-party data cost. Watch for a plan that grew because the data did.
7. Deal ID performance. Which PMPs earn their floor price and which ones only consume budget.
8. Audience overlap. Two segments billed separately can be largely the same people.
9. Incrementality, at least quarterly. A holdout beats an attribution model you cannot inspect.
10. Access rights confirmation. A dated line confirming the seat, the deal IDs and the log-level access are intact.

If a monthly report cannot show lines one through three, the conversation to have is about reporting rights, not optimization.

Conclusion

The fee is the easiest line on that table to negotiate and the least important. What separates a 54.0% qualified impression rate from 32.1% is who owns the seat, how tight the quality rules are, and whether anyone outside the agency can open the log-level file. For a full picture of how the reporting layer is built on top of that data, the Google Ads reporting guide covers the same mechanics on the search side.

Before you sign anything programmatic in the next quarter, ask for one page of placement-level data from your own account, and for the deal IDs behind it. The ones who send it within a day and the ones who explain why they cannot are two different businesses, and the fee model looks identical from the outside. So which of the two is on your shortlist right now?