Write a CTV measurement brief before buying media

Start with a sentence finance and media can both use: we will increase, maintain, or reduce CTV investment based on its contribution to a specified business outcome. For a retailer that may be incremental purchases; for a business selling through a sales team it may be qualified opportunities with a later revenue check.

Choose one primary outcome and define its source, timing, exclusions, and owner. A sale should mean the same thing before and after the campaign. Decide how cancellations, returns, repeated purchases, and conversion delays will be handled. Supporting metrics can explain a result, but they should not replace the primary outcome after the result disappoints.

Delivery describes what the buying system served: spend, impressions, reach estimates, frequency, and completed views. Attribution describes outcomes linked to an exposure under specified matching rules and time windows. An incrementality study estimates the difference caused by advertising against a comparison condition.

These answer different questions. A high completion rate can show that video was delivered without showing that someone bought because of it. An attributed purchase can be useful for diagnosis while still coming from someone who would have purchased anyway. Google describes Conversion Lift as a controlled comparison of geographic groups receiving and not receiving the tested advertising; it also requires an eligibility and feasibility check.

Ask the provider to document the identity method, household or device unit, lookback window, modeled components, deduplication rules, and unmatched share. Separate click-based and view-based figures. If two platforms credit the same sale, do not add both claims together as two sales.

For DOOH followed by CTV, ask how a location signal becomes an eligible audience and then a matched household. Require a clear statement of uncertainty at each step. Plan around the provider’s documented measurement capabilities; an audience match should not be presented as proof that a named person saw a particular screen.

Record which events your first-party sales system can verify and which figures only exist in a platform report. Keep those definitions alongside the numbers so a stakeholder can understand what the campaign report supports.

A platform-supported randomized holdout may suit an eligible campaign. A geographic test may suit a broader question where outcomes can be measured consistently by region. Availability, data volume, expected effect size, and campaign constraints determine feasibility; there is no universal minimum budget or duration that makes every CTV study reliable.

Specify exactly what changes between test and control. If you add both DOOH and CTV, the study tests the combined package. To estimate the added value of CTV after DOOH, keep the DOOH treatment comparable and vary CTV where the design permits. Have the measurement owner check assignment, baseline balance, sample size, and the analysis before launch.

Record promotions, inventory changes, other media, and exposure that can cross into control areas. Google’s geo experiment guidance explicitly discusses contamination between regions. If a credible causal test is infeasible, label the report observational and use it to plan a stronger next test.

Show attributed and experimental outcomes side by side. Include uncertainty, cost definitions, and any departures from the original plan. If you calculate incremental return on ad spend, state whether the numerator is revenue or profit and exactly which costs appear in the denominator.

Agree scale, revise, and stop rules in advance. An inconclusive study does not establish zero impact; it also does not justify claiming success. The practical next step might be a larger test, better conversion data, different creative, or a different allocation. The editable brief gives that decision an owner and a date.