How to measure CTV without confusing attribution with…

In summary

The answer in brief

Measure CTV in two layers. Use attribution to describe observed journeys and support day-to-day optimisation. Use a controlled lift design—preferably user or geographic holdouts—to estimate conversions or revenue that would not have happened without the campaign. For wider budget allocation, use a marketing-mix model calibrated with experiments. Agree the outcome, control, conversion lag, minimum detectable effect and decision rule before spend begins.

Attributed conversions are not automatically incremental conversions.
Match the method to the budget decision it must support.
Protect a control group before the campaign changes the market.
Report uncertainty and inconclusive results, not only a point estimate.
Use experiments to calibrate models and attribution assumptions.

Which CTV measurement method answers which question?

Which CTV measurement method answers which question?
Method	Best question	Main limitation
Platform attribution	Which observed exposures preceded recorded actions?	Assigns credit; does not create a counterfactual
Branded-search or site lift	Did demand signals move after exposure?	Can be affected by other campaigns, news and seasonality
User holdout	Did exposed users convert more than a comparable control?	Requires eligibility, identity matching and enough volume
Geo experiment	Did test markets outperform comparable control markets?	Market spillover and poor matching can weaken the result
Marketing-mix model	How did channels contribute across time and the portfolio?	Causal estimates depend on model assumptions and controls
Incrementality-calibrated mix	How should evidence inform the next budget?	Needs disciplined experiments and consistent outcome data

Separate attributed response from incremental effect

Attribution and incrementality are useful because they answer different questions.

Attribution connects a conversion to one or more recorded touchpoints. It helps teams inspect journeys, creatives, audiences and delivery. Incrementality asks the causal question: how many outcomes occurred because the advertising ran, above what would otherwise have happened?

Simplicity Media’s partner page describes CTV measurement through geographic holdouts, branded-search lift, marketing-mix modelling and attribution. That is a useful method list, but any agency result still needs a disclosed design, baseline, comparison and uncertainty before it can support a budget claim.^[1]

Review the paid media guide

Write the measurement contract before buying media

The campaign and the test must share one outcome definition and one decision date.

Name the primary outcome, data source, owner, attribution window, expected conversion delay, exclusions and value rule. Decide whether the question concerns reach, qualified visits, leads, sales, profit or customer value. Do not switch to an easier metric after seeing weak results.

Document the counterfactual: a user, household or market that could have received the campaign but did not. Define the minimum effect worth acting on and estimate whether the available budget and conversion volume can detect it.

Primary and secondary outcomes
Treatment and control eligibility
Conversion lag and study length
Cross-channel contamination checks
Minimum useful effect
Scale, stop or retest rule

Use holdouts to estimate causal lift

A valid control group provides the clearest answer to what happened because of the campaign.

Google describes Conversion Lift as a controlled experiment comparing treatment and control groups, with user-based and geographic designs. Report absolute lift, relative lift, incremental value, incremental cost per action and incremental return where the data supports them.^[2]

Power matters. Google’s guidance says certainty depends on historical conversion volume, expected lift, budget, duration and holdout size. A low-certainty result can be directional or inconclusive; it should not be converted into a confident revenue claim.^[3]

Use MMM for the portfolio, not as a magic truth machine

Marketing-mix modelling is useful for longer-horizon allocation across channels and external conditions.

An MMM can estimate channel contribution while accounting for seasonality, pricing and other controls. Google’s open-source Meridian documentation also makes the limitation explicit: causal inference is difficult to validate directly without a well-designed experiment.^[4]

Calibrate the model with lift studies when the estimand and time window align. Compare model contribution, experimental lift and attributed journeys. Disagreement is diagnostic: check contamination, missing controls, identity loss, conversion lag and whether each method measured the same outcome.

Report the evidence so another person can challenge it

A defensible CTV result includes the design and its limits, not only the headline lift.

Keep a record of markets or users, campaign dates, spend, creative, reach and frequency, outcome definitions, exclusions, other media, model version and confidence interval. IAB Europe’s draft CTV Measurement Framework similarly separates delivery, attention, brand outcomes, business outcomes and incrementality.^[5]

End with a decision: scale, hold, redesign or rerun. State what the study can prove, what remains correlated and what may not generalise to another season, audience or spend level.

Compare media agencies

FAQ

Frequently asked questions

What is the difference between attribution and incrementality?

Attribution assigns credit among observed marketing touchpoints. Incrementality estimates the outcomes caused by advertising compared with a credible no-ad counterfactual.

Can CTV be measured with last-click attribution?

Last click can show the final recorded interaction but usually misses or misallocates upstream television influence. Use it for journey context, not as the sole causal claim.

What is a geo holdout test?

A geo test runs media in selected markets and compares outcomes with similar control markets where it is withheld. Market matching, spillover, other media and sample size determine how credible the estimate is.

When should a company use marketing-mix modelling?

Use MMM when the decision spans several channels and a longer period with enough consistent spend, outcome and control data. Smaller campaigns may need simpler tests.

What if the lift result is inconclusive?

Report it as inconclusive. Check statistical power, contamination, conversion lag and outcome quality, then decide whether a larger or better-controlled repeat is worth the cost.

Should branded-search lift be treated as sales lift?

No. It can be a useful demand signal, but other activity can move branded search and the signal is not revenue. Link it to controlled evidence and downstream business outcomes.

Sources and further reading

Editorial method: English and Portuguese SERP patterns were reviewed on 26 July 2026. Partner pages supplied practitioner questions; factual claims were checked against the primary and official sources below. Commercial and partner material was not treated as independent proof.

How to measure CTV advertising incrementality — Simplicity Media
Partner description of CTV capabilities and proposed measurement methods. Promotional performance claims were excluded from the evidence base.
Google Ads — About Conversion Lift
Official definition of user and geographic controlled lift studies and incremental metrics.
Google Ads — About certainty of lift
Official guidance on statistical certainty and inconclusive results.
Google — Meridian model-fit guidance
Primary documentation on causal goals, model assumptions, controls and validation limits in MMM.
IAB Europe — CTV Measurement Framework and Transparency Principles
Official framework landing page used for its measurement taxonomy, with the public-consultation status and dates retained.