Reading CTV incrementality claims: five things to check

Incrementality is the share of results the ads caused. It is the number that should decide budgets, which is why platforms feature it prominently. But the word covers very different methods, from randomised household holdouts to geo tests to statistical models, and a headline figure can come from a single favourable case. Reading the method behind a claim matters more than the figure itself.

Real incrementality needs a group that could not see the ads: a holdout of households, a set of regions without ads, or a matched control. If a report compares exposed people with everyone else, without a designed comparison group, it is closer to attribution than to incrementality. Ask how the comparison group was chosen and whether it was fixed before launch.

A result measured by the platform that sold the impressions is useful but is not outside evidence. A result read by a third-party measurement partner using its own method is easier to trust and to compare with other channels. Ask whether the partner had access to raw exposure data and whether the platform shaped the design.

Lift on site visits within a day is not the same as lift on first purchases within a month. Check the metric, the time window and whether the metric was chosen before the campaign. A claim that switches metrics between the plan and the report deserves caution.

Small tests produce noisy results. Ask for the size of the exposed and comparison groups and for the uncertainty range around the lift, not only the central number. A lift with a range that includes zero is not yet a result.

Case studies show what happened for one advertiser with one product, budget and creative. They can show that a method works, not that your campaign will produce the same number. The only incrementality figure that should decide your budget is the one from your own test.

On the pages we read, tvScientific describes incrementality-driven attribution and incremental lift analysis, Tatari says buyers can measure the true incremental impact, Vibe's MCP page describes incrementality testing built in with geo holdout and cross-channel lift examples, and StackAdapt lists brand lift and sales lift studies. Several platforms do not describe incrementality testing on the pages we read; that is a reason to ask, not a conclusion about what they offer.

Source: tvScientific measurement and attribution page

Source: Tatari TV measurement page

Source: Vibe MCP page

Source: StackAdapt connected TV advertising page

Whichever platform you choose, design the test yourself, or with a partner you trust, and agree it before the first impression.

A good report states the design in one paragraph: how the comparison group was chosen, its size, the dates, the metric and the window. It shows the outcome in both groups, the difference, and a range of uncertainty around it. It says who analysed the data and whether anything changed during the test. It explains how the result was turned into cost per incremental result. If a report leaves any of these out, ask for them before you act on the number.

It also separates the test from the business case. The test tells you what the campaign caused; the decision to scale also depends on margins, capacity and what else you could do with the budget. Keep those two conversations apart, so the measurement is judged on its method and the budget on its economics.

Finally, be consistent. Apply the same five checks to every platform's claim, including the one you already use. The checks are not designed to catch out a particular vendor; they are designed to make different claims comparable, so that the budget follows the evidence rather than the most confident presentation. If a platform cannot answer them, run your own test before you rely on its numbers.

The same discipline applies inside your own team: agree the checks before the results arrive.