Featured measurement case
Matched Market Test: validating offline media campaign impact
This case frames an offline media measurement problem for TV/OOH campaigns where user-level tracking is not practical. The measurement question is whether treated markets generated more foot traffic than similar untreated markets would have predicted without the campaign.
- Situation / TaskTV/OOH campaigns are hard to evaluate at the user level, so the business needed a market-level way to validate whether offline media created incremental foot traffic.
- ActionI matched test markets to control markets using pre-campaign foot-traffic patterns, market size, media delivery, and spillover risk, then used R + CausalImpact to estimate the no-campaign counterfactual baseline.
- ResultThe analysis produced an impact-validation readout: actual timestamped foot traffic versus the predicted no-campaign path, with lift, uncertainty, and the key data-source limitation clearly separated.
- Statistical theoryBayesian structural time-series models estimate the missing no-campaign path by combining trend, seasonality if needed, and control-market predictors, then return a posterior interval for lift.
- LimitationThe main constraint was data source quality: offline timestamped foot traffic was available, but online store-visit or richer conversion data would have improved validation.
- Plain-English translationCompare campaign markets with similar non-campaign markets so we can estimate what foot traffic would have looked like without TV/OOH exposure.