← Every exhibit

Table Table 3 From the research bench

Which operational miscalculations undermine incrementality testing?

MiscalculationRoot causeExperimental failureCorrective protocol
Running tests without sufficient sample sizeUnderestimating sales variance relative to small ad liftYields wide confidence intervals that span zero; inconclusive testsPerform pre-test power calculations following Lewis and Rao (2015)
Treating pre-post changes as causalTurning off ads nationally and measuring before vs afterConflates seasonal demand swings and macro trends with ad liftAlways use concurrent control groups or synthetic geo-controls
Ignoring spillover and interferenceTreatment exposure leaks into control regions via word-of-mouthViolates SUTVA (Stable Unit Treatment Value Assumption)Establish geographical buffer zones around test territories
Evaluating incrementality on short windowsMeasuring conversions only during the active test periodMisses delayed conversions and ad-stock decay effectsExtend measurement windows post-intervention to track decay
Applying uniform incrementality across channelsAssuming paid social and brand search have identical liftOver-funds brand search while under-funding true prospectingMeasure incrementality independently across discrete channel tiers

Swipe or scroll horizontally if the table is wider than your screen.

Cite Embed

Reference & Evidence

Source: Table from this essay. Sources and interpretation are given in the article.