CTV Frequency Capping in 2026: A 6-Step Framework for Buying Reach Instead of Repetition

The average CTV campaign reaches just 19.64% of households at a frequency of 7.09, per Innovid's analysis of hundreds of billions of impressions served in 2024. CTV frequency capping is how you rebalance that — but the fix is not typing a smaller number into a DSP. It is buying reach instead of repetition.

The real problem is not frequency. It is the reach you traded away for it.

Every conversation about CTV frequency capping starts in the wrong place: what number should I cap at? Three? Five? Seven? It is the wrong question, and the data shows why.

In its 2025 CTV Advertising Insights Report (April 2025), Innovid analysed hundreds of billions of video ad impressions served across its platform in 2024. The headline finding was not that frequency was high in isolation. It was the pairing: an average delivered frequency of 7.09 against an average household reach of 19.64% of Innovid's 95 million-plus U.S. households.

Read those two numbers together and the waste becomes obvious. Four out of five households in the addressable pool never saw the campaign at all. The fifth that did saw it roughly seven times. That is not a frequency problem you fix by typing a smaller number into a DSP. It is a distribution problem.

It gets worse at scale. Innovid found that campaigns serving 200 million or more impressions saw delivered frequency climb to 10 or more. The bigger the budget, the more of it lands on households that already converted, already ignored you, or already got irritated.

Why there is no universal "correct" cap

You will find plenty of confident advice online naming a specific number as the CTV frequency sweet spot. Treat it with suspicion. The Advertising Research Foundation's own work on finding the frequency sweet spot in today's TV ecosystem lands on a less satisfying but more honest conclusion: effective frequency is contextual. It moves with the ad environment, the timing, the creative, the product category, and the audience.

A 15-second spot for a quick-service restaurant promoting a two-week offer does not behave like a 30-second spot for a considered financial product with a six-month decision cycle. Anyone quoting you one number for both is selling a rule of thumb as a finding.

So the useful framework is not "what is the right cap." It is "how do I stop buying repetition I did not intend to buy." That is six steps.

Step 1: Measure delivered frequency, not the cap you set

The cap in your DSP is an intention. Delivered frequency is what actually happened. On most buys they are different numbers, and the gap is where the money goes.

Pull your actual distribution, not the average. An average frequency of 7 can mean every household saw the ad seven times, or it can mean half saw it twice and half saw it twelve times. Those are completely different problems with completely different fixes, and the average hides both. Ask your DSP or ad server for the frequency distribution curve — the share of reached households at 1, 2, 3, 5, 10 and 15-plus exposures.

If you cannot get that report, that is itself the finding. You are flying blind on the single largest controllable source of waste in a CTV buy.

Step 2: Cap the household, not the device

A living room with a smart TV, a streaming stick and a games console is three devices. Cap per device and you have quietly authorised three times the exposure you planned.

Household-level capping requires identity resolution — IP-based or graph-based — and not every platform does it by default. Check which level your cap is enforced at before you trust the number. This is the single most common reason a buyer who "capped at 3" sees a delivered frequency of 9.

Step 3: Cap across the entire buy, not inside each platform

This is the structural flaw in most CTV plans. If you are running across a DSP, two direct publisher deals and a FAST platform, each one enforces its own cap in its own silo. Four platforms capped at 4 each is a household cap of 16, and none of the four is doing anything wrong.

The fixes, in order of preference: buy through a single DSP wherever the inventory allows it; use an ad server or frequency management layer that sees across all four; or, if neither is available, divide your target household cap by the number of independent sellers and accept that the enforcement is approximate. Buyers running a genuinely fragmented buy should read our breakdown of streaming ad CPM benchmarks by platform before assuming a cheap CPM on a fourth platform is actually incremental.

Step 4: Set the cap against the objective, then validate it

Because there is no universal number, the honest approach is to set a defensible starting point and then let your own data correct it.

Reasonable starting logic: campaigns built for reach and awareness should bias toward more households at lower exposure, because the marginal value of the eighth impression on an already-reached household is close to zero. Campaigns built for response in a short window can tolerate more repetition, because you are trying to catch a household in a narrow buying moment rather than build memory over months.

Whatever you choose, write down the number you set, the number that delivered, and the outcome. Three campaigns of that and you have something better than any published benchmark: a cap validated on your own audience, your own creative, and your own product.

Step 5: Watch the reach curve, not the frequency number

Frequency is an output, not an input. The input is how much of your budget is chasing new households versus re-serving old ones.

Track incremental reach per thousand dollars spent, week over week. Healthy campaigns show reach climbing steadily through the flight. When the reach curve flattens while spend continues, you have saturated your addressable pool — and every dollar after that point is buying frequency you did not ask for. That is the moment to expand the audience, change inventory, or stop.

This is also why CPM shopping can backfire. A cheap CPM on a small inventory pool saturates faster. Our CTV CPM benchmarks for 2026 are worth reading alongside your reach curve rather than on their own.

Step 6: Move budget off saturated households

The reason frequency management became a serious discipline in 2026 rather than a checkbox is that the tooling finally caught up. Innovid explicitly frames this in its report as shifting budget away from overexposed households and toward untapped ones — turning waste into performance.

Practically, that means suppression: build an audience of households that have already hit your target exposure and exclude them, then redirect that budget into prospecting. This is the same mechanic as a retargeting exclusion list, applied to reach rather than conversion.

It only works if your measurement is sound, which is the dependency most buyers skip. If you are not confident your exposure and outcome data line up, start with measuring CTV attribution properly before you start suppressing anything.

If you are not sure what your campaigns are actually delivering against what you set, a free audit of your current CTV delivery will show you the frequency distribution and the reach curve side by side — which is usually the first time a buyer sees where the waste is.

What this changes about how you plan

The shift is small but it reframes the whole exercise. Stop briefing "cap at five." Start briefing "reach this many households at a frequency we have validated, and tell me when the reach curve flattens."

That brief forces three things that a cap number does not: it makes reach a target rather than a by-product, it makes delivered frequency a reported metric rather than an assumption, and it gives you a defined stopping point instead of spending to the end of the flight out of habit.

The budget consequences are not marginal. If the industry average is a fifth of households at seven exposures, and you can move to a third of households at four, you have bought substantially more of the market for the same money — without negotiating a single CPM. For context on how that compares to the other lever most buyers pull first, see where CTV and linear TV budgets are actually going in 2026.

Methodology and sources

Delivery figures in this article are drawn from Innovid's 2025 CTV Advertising Insights Report, published 23 April 2025 and based on hundreds of billions of video ad impressions served by Innovid during calendar year 2024 across a base of more than 95 million U.S. households. The position that effective frequency is contextual rather than fixed reflects the Advertising Research Foundation's published work on frequency in the current TV ecosystem. Where we describe cap-setting logic by objective, we have labelled it as a starting point for validation rather than a benchmark, because we could not identify a credible cross-industry study establishing a single optimal CTV frequency, and we are not going to invent one.

Frequently asked questions

What is a good frequency cap for CTV advertising?

There is no single correct number, and any source quoting one without qualifying it is overreaching. What the data does support is that the industry's delivered average — 7.09 exposures against 19.64% household reach, per Innovid's 2024 analysis — is unbalanced for most objectives. Set a cap you can justify against your objective, measure what actually delivers, and adjust across three campaigns.

Why is my delivered frequency higher than the cap I set?

Almost always one of two reasons. Either the cap is enforced per device rather than per household, so a three-device living room gets three times the exposure; or you are buying across multiple platforms that each enforce their own cap independently, so four platforms capped at four deliver up to sixteen.

Does frequency capping reduce my total reach?

No — it does the opposite, provided your budget stays the same. Capping redirects impressions that would have gone to already-reached households toward new ones. Reach goes up. What falls is repetition, which is the part you were not intentionally buying.

How do I cap frequency across different DSPs and publishers?

Consolidate the buy into one DSP where the inventory allows, or use an ad server or frequency management layer that measures across every seller in the plan. Where neither is possible, divide your target household cap by the number of independent sellers and treat the result as approximate rather than enforced.

Is high frequency always bad?

No. Short-window response campaigns trying to catch a household inside a narrow buying moment can justify more repetition than a brand campaign building memory over months. The problem is not high frequency — it is unintended frequency, arriving because of device-level caps and siloed platforms rather than a decision anyone made.

Where to take this next

If you want a second opinion on what your CTV campaigns are actually delivering, there are two ways to start. Book a 30-minute intro call with Ryan and we will talk through your current setup, no pitch. Or request a free audit and we will pull your delivered frequency distribution and reach curve against the benchmarks above. You can also read more about how we approach programmatic buying and audience targeting, or get in touch directly.