CTV Advertising: How to Buy Real Reach Without Hidden Margins

The industry has spent the last decade using CTV and OTT interchangeably, and the confusion is not academic. It carries pricing consequences.

The IAB Tech Lab's guidance is clear: CTV describes a device, not a content category. It refers to internet-connected televisions and the devices that stream video onto them (smart TVs, Roku, Fire TV, Apple TV, Chromecast, game consoles). OTT, by contrast, refers to premium video content delivered over the internet regardless of the screen it lands on. The Media Rating Council's 2021 SSAI and OTT measurement guidance formalized the distinction: OTT is a content descriptor; CTV is a device descriptor; the two intersect but are not the same. We address the OTT question separately in the OTT advertising pillar.

The consequence for buyers is not academic. Any CTV insertion order that includes mobile web pre-roll, tablet interstitial video, and desktop long-form is not, definitionally, a CTV buy. It is an OTT buy with a CTV price tag. The buyer is paying big-screen CPMs for phone-screen impressions, and the media plan almost never separates the two before the flight goes live.

The reason the two categories are conflated so often is that the reseller layer profits from the conflation. A platform that sells CTV as a device-agnostic content bucket can allocate impressions to whichever device carries the highest available margin at that moment. The buyer sees a blended report. The platform captures the arbitrage.

Linear television CPMs in 2026 broadly cluster in the $10 to $15 range for national broadcast, with daypart spikes for live sports and primetime news taking the upper end higher, per S&P Global Kagan's TV advertising forecasts. CTV CPMs, by contrast, cluster in the $15 to $85 range depending on inventory tier: FAST channels and broad AVOD run $15 to $25, premium AVOD like Hulu, Peacock, Max, Paramount+, and Prime Video run $25 to $45 in programmatic and $45 to $65 in direct-sold pods, and first-party audience-targeted buys run $45 to $85, per IAB Tech Lab's CTV programmatic guide and multiple SSP-reported benchmarks.

The premium is defensible on paper. CTV delivers household-level targeting, digital-style attribution, cross-device frequency capping (in theory), and a cleaner viewability story than legacy linear ever offered. But the premium is only defensible if the buyer is using those advantages, and most are not.

A buyer paying $45 CPMs for Hulu inventory without a first-party audience overlay, without pod-position guarantees, and without cross-app frequency measurement is paying premium prices for benefits they never activated. The equivalent linear buy on the same content at a $12 CPM would have delivered the same message to a comparable household set with an honest, if simpler, reach curve. The CTV premium is not intrinsic to CTV. It is a premium the buyer earns back by using CTV's advantages, and it disappears when they don't.

Here is where the independent-agency conversation gets uncomfortable. The default CTV buying path for mid-market and independent-agency buyers routes through a small set of self-serve reseller platforms, of which MNTN is the largest and most publicly documented. The platforms are useful. They compress a genuinely complicated buying workflow into a manageable interface. That is the service they sell, and they sell it well.

The economics of that service are worth understanding in detail. MNTN's FY2025 disclosures report a full-year gross margin of 77 percent on $290 million of revenue, per its FY2025 earnings release filed with the SEC. Q2 2025 gross margin was also 77 percent, up 700 basis points year over year. What that number describes, in plain terms: of every dollar an advertiser pays MNTN, roughly 23 cents flows through to the underlying inventory owner (the publisher, the SSP, the ad server, the delivery infrastructure), and 77 cents stays with MNTN as gross profit before operating expenses.

That is the take rate. It is not an accusation. It is what the SEC filing says.

Compare that to the honest disclosures at the transactional layer of the programmatic supply chain. The ANA study, using log-file matching on $123 million of real advertiser spend across 21 brands, found that transaction costs (DSP plus SSP plus ad-server fees combined) run 29 cents on the dollar in traditional open-programmatic display, and the remaining loss is inventory quality, not intermediary margin. The CTV performance-platform layer sits on top of that stack, extracting a further margin above and beyond the DSP and SSP fees the ANA measured. A buyer paying a self-serve CTV platform for their Hulu access is paying the platform's markup plus the DSP fee, plus the SSP fee, plus the ad-server delivery cost, before the impression reaches a household.

The platforms are not hiding this. It is disclosed in their own filings. The problem is that the disclosure is not translated into a language brand-side buyers can use when they compare a self-serve CTV plan to a direct-negotiated one. The result is that a large share of independent-agency CTV budget flows through a stack whose intermediary layer keeps a majority of the dollar, and the buyer's internal reporting shows the top-line CPM without the take-rate context that would put it in perspective.

This is not an argument that the reseller stack is fraudulent, unethical, or bad for every buyer. For an advertiser spending $150,000 a year on CTV, the platform fee is a rational price for the operational simplicity. For an advertiser spending $8 million, it is not. Somewhere between those numbers is the threshold where a direct-negotiated approach (private marketplaces with the streaming platforms, programmatic guaranteed on premium inventory, or agency-of-record negotiated upfront positions) pays for itself many times over. Most mid-market brands are on the wrong side of that threshold and do not know it because no one on their current buying stack is incentivized to tell them.

The other thing missing from most CTV pitches is the reach and frequency conversation that linear TV buyers have been having for sixty years and that CTV is quietly recreating from scratch.

Household reach in linear TV is a measured, defensible number. A national primetime buy on a broadcast network delivers a documented reach percentage against a documented population, with Nielsen panel data and set-top-box return-path data cross-checking each other. The math has flaws. It is more accurate than what CTV currently offers.

The CTV equivalent is fragmented across a dozen platforms, each with its own household graph, its own frequency-cap logic, and its own definition of "unique." Innovid's benchmarking data, cited by IAB, put the average publisher-level duplication rate in cross-app CTV campaigns at roughly 32 percent, meaning that a third of the households a buyer thought they were reaching uniquely across, say, Hulu and Peacock were in fact the same households being counted twice. More consequentially, Innovid's cross-campaign analysis found that 8 to 15 percent of households in a typical CTV flight exceed the intended frequency cap, and those over-frequency households account for 42 to 60 percent of total delivered impressions.

Read that carefully. In a CTV campaign designed to reach 5 million households at a target frequency of 3, more than half the impressions are landing on a small minority of over-exposed households, and a third of the households were duplicated across publishers before the frequency cap was even measured. A linear buyer running that campaign in 1995 would have caught the problem inside two weeks with a Nielsen reach curve. A 2026 CTV buyer with the same problem may never see it, because their reporting comes from the platform that has no incentive to surface it.

The Nielsen Gauge report for December 2025 documented that streaming captured 47.5 percent of all U.S. television viewing that month, with cable at 20.2 percent and broadcast at 21.4 percent, per Nielsen's Gauge monthly viewing report. Streaming has won the audience. It has not yet won the measurement infrastructure that would let advertisers plan against that audience with the discipline they once brought to linear.

The other conversation missing from most CTV pitches is invalid traffic. Pixalate's Q4 2025 North American benchmarks put the CTV invalid-traffic rate in the United States at 19 percent, with Roku traffic showing 20 percent of measured bundle IDs classified as malformed, unidentified, or fraudulent. DoubleVerify's 2025 global insights report found that bot fraud accounted for 65 percent of all fraud in CTV environments in 2024, a share 14 percent higher than the same measurement in other digital channels, with 4 million compromised CTV devices generating invalid traffic daily.

Layer those numbers on top of the take-rate math. A buyer paying $45 in CPM to a self-serve CTV platform, on inventory where 19 percent of impressions are invalid, whose platform is retaining 77 cents on the dollar in gross margin, is paying an effective net CPM against verified-human reach that is meaningfully higher than the sticker price. The gap between the reported CPM and the effective per-verified-human CPM is the operational alpha available to a buyer willing to audit the stack.

None of this argues that CTV is a bad channel. It is a large, growing, and structurally important channel that is not going to reverse. The argument is one level up: the way most mid-market and independent-agency buyers are buying CTV in 2026 is the same posture linear buyers had in the mid-1990s before Nielsen currency, reach-curve modeling, and upfront negotiation discipline made linear a professionally-bought medium. CTV will get there. Some buyers will not wait for it.

Five questions belong on every CTV insertion order in 2026:

What percentage of the buy is genuinely CTV versus OTT delivered on non-television screens? The platform-level report should break device out separately. If it does not, the buyer is paying blended pricing on a mixed-device delivery.

What is the take rate? Not the CPM. The take rate. If the platform will not disclose the gross margin on the specific buy, the buyer should assume it is at or above the 77 percent MNTN benchmark and price the negotiation accordingly.

How is frequency capped across apps? If the answer is "we cap at the app level," the buyer should assume publisher duplication above 30 percent and over-frequency concentration above 40 percent, and reprice the reach-and-frequency assumptions in the plan.

What percentage of delivered impressions passed independent MRC-accredited verification? Not the platform's own verification. Independent verification. The brand lift methodology and the audit path we walk clients through on programmatic direct deals apply the same discipline to CTV.

Where does a direct-negotiated alternative pay for itself? For most portfolios above $2 to $3 million in annual CTV spend, a portion of the budget belongs in direct publisher deals or programmatic guaranteed, negotiated by an entity whose incentives are aligned with the advertiser rather than with the take rate on the buy.

Connected TV is now the largest single line item in most brand-side video plans and, in 2026, the largest single line item in the U.S. upfront. It is also, at the same time, the least professionally-bought category of premium video inventory that has ever existed at this scale. The reseller layer that dominates independent-agency access to the channel retains gross margins that would be extraordinary in any other category of media brokerage and that pass largely unremarked in the CTV conversation. The reach and frequency infrastructure that linear buyers built over sixty years does not yet exist in comparable form across the streaming ecosystem, and its absence is not a rounding error. It is the reason more than half of a typical CTV campaign's impressions land on a small minority of households the buyer never intended to over-serve.

These problems are not permanent. The IAB, the MRC, and the leading measurement vendors are all moving in the right direction, and cross-publisher frequency governance is closer to reality than it was two years ago. But no brand-side team should confuse where the industry is heading with where it is today. Today, a CTV plan built without take-rate discipline, without frequency-duplication scrutiny, and without an independent verification layer is a plan that overpays.

Criterion Global is privately held and founder-owned. It is not tied to a holding-company network, a media reseller, or a preferred vendor stack. That structure matters most in categories like connected TV where the intermediary layer captures more of the advertiser's dollar than it discloses in any single reporting view. The Criterion Global Budget Blueprint℠ diligence we run for clients evaluating CTV allocations is structured to surface exactly the take-rate, frequency, and verification math that platform-side reporting is not designed to show.

For further reading on adjacent categories, see the sibling piece on OTT advertising, the primer on programmatic direct deals, and the deeper discussion of the role of digital in media economics.

Criterion Global builds international and portfolio media strategy for consumer, luxury, and financial brands. To discuss a Criterion Global Budget Blueprint℠ assessment of a connected TV or advanced-video plan, contact the practice through the international media planning and buying engagement page.