← Every exhibit

Table Table 3 From the research bench

Which operational miscalculations undermine marketing mix modeling?

MiscalculationRoot causeEconometric failureCorrective protocol
Accepting high $R^2$ as proof of causalityOverfitting dozens of regressors to limited time periodsMistaking collinear trend-fitting for true causal elasticityEvaluate out-of-sample prediction and holdout tests
Omitting organic demand driversFailing to model pricing moves, product releases, and PR eventsMarketing coefficients artificially absorb baseline salesInstrument explicit controls for price changes and macro index
Ignoring spend endogeneityAutomated ad tools spend more when sales are already surgingRegression interprets correlation as advertising persuasionImplement instrumental variables or experimental priors
Using static adstock parametersForcing identical decay rates across diverse digital channelsOverestimates search longevity; underestimates brand half-lifeCalibrate channel-specific decay parameters via decay audits
Operating MMM without experimentsRelying entirely on observational historical dataModel drifts into statistically confident delusionMandate quarterly randomized holdouts to anchor priors

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.

Related exhibits