How to Measure Connected TV Without Confusing Attribution With Lift

Measure Connected TV at three levels: whether the agreed advertising was delivered, what activity the reporting system associated with it and whether the campaign created additional business. Keep those questions separate in the report.

A television ad can reach a relevant household without producing a tracked visit. A tracked household conversion can occur without the ad causing it. Both facts matter when someone asks whether the campaign worked.

Read this fictional report before choosing a metric

A fictional retailer spends $8,000 on media. Its provider associates 80 orders with exposed households. The simple calculation is $100 in media cost per associated order: $8,000 divided by 80.

That is not yet an incremental acquisition cost. Some customers may have bought anyway, another channel may claim the same orders, and the orders may include existing customers. Creative, management, measurement, refunds and product costs are also outside that $100 calculation.

A useful meeting decision is: “Reconcile the 80 orders with store records, check the matching method and reporting window, and decide whether we have a feasible controlled comparison.” Calling all 80 new customers would skip the most important questions.

Start with delivery

Review impressions, reported reach, frequency, completed views and the inventory information your agreement provides. Ask how reach is deduplicated and which unit it represents. Do not add separate platform reach figures as though they describe unique households across the whole plan.

Video completion can help assess delivery and exposure conditions. It is not proof that someone watched attentively, remembered the message or purchased. Investigate unexpected delivery patterns before using them to explain sales.

Document the attribution method

Ask the provider what connects an exposure with a later visit or order. Record the matching method, eligible devices, lookback window, exclusions and known coverage limits. Household and cross-device matching can involve estimates and incomplete information.

Use a consistent window when comparing campaigns. Choose a review period that fits the purchase cycle and observed conversion delay. A standard number of days copied from another campaign does not establish that your results are mature.

Reconcile reported orders or leads with your own records where the permitted identifiers support that work. Separate modeled results from directly observed records. A missing match is an unresolved coverage question until you investigate it.

Test additional business when the decision needs it

A well-designed experiment can provide stronger evidence of causal lift than an attribution report. The design still needs credible comparison groups, reliable outcome measurement and enough information to detect a commercially meaningful difference.

A geographic comparison, for example, can be distorted by different promotions, service coverage, seasonality or other media. Decide how those issues will be handled before launching. Comparing two unlike regions after the campaign is not automatically an incrementality test.

If a suitable test is impractical, state the limitation. You can still improve creative, delivery and commercial reconciliation without claiming causation you have not measured.

End the report with a decision

Keep a short record: business objective, full costs, delivery findings, attributed outcomes, reconciled customer results, remaining uncertainty and next action. Include an owner and a review date.

In the retailer example, unresolved duplicate orders call for reconciliation. Confirmed delivery to unsuitable areas calls for a buying correction. Neither finding calls for a larger budget by default.

For help connecting the buying plan with a credible review, discuss Connected TV advertising with Experience Advertising. Bring the provider's reporting definitions as well as the performance screenshot.