Not every shopper is identified. Loyalty penetration might be 60%, 70%, maybe 80% at best. The rest are anonymous, they transacted, but they can't be linked to a campaign exposure.
So what do you do with the gap? Ignore it, and you understate the campaign's impact.
Extrapolate, and you risk overstating it. Either way, the number is wrong, the question is which direction you prefer to be wrong in, and how transparent you are about the choice.
What extrapolation means
Extrapolation takes the behavior observed in the matched (identified) population and projects it onto the unmatched (anonymous) population.
If identified shoppers exposed to the campaign showed a 6% uplift, extrapolation assumes the anonymous shoppers also showed approximately 6% uplift. The total estimated impact becomes the matched result scaled to the full audience.
The assumption: matched and unmatched shoppers behave similarly. This assumption is often wrong. Loyalty members tend to be heavier buyers, more engaged, more responsive to promotions. Projecting their behavior onto lighter, less engaged anonymous shoppers likely overstates the effect.
The rules of honest extrapolation
If you extrapolate, you must state the rules. At Footprints AI, the methodology is visible:
1. Report matched results as primary. "Among identified shoppers, the campaign delivered 6.2% incremental uplift and 4.8x iROAS."
2. Report extrapolated estimates separately. "Extrapolating to the full estimated audience, total incremental impact is estimated at €X. This assumes similar response rates among unidentified shoppers."
3. State the match rate. "68% of estimated exposed shoppers were identified through loyalty data."
4. State the assumption. "Extrapolation assumes equivalent response in the unmatched population. If unmatched shoppers are less responsive (as is typical), the true total impact is between the matched result and the extrapolated estimate."
5. Never present extrapolated numbers as measured numbers. The distinction must be clear in the report, in the dashboard, and in any conversation about results.
When extrapolation is appropriate
Extrapolation makes sense when: - The match rate is reasonably high (>60%) so the matched sample is representative - The extrapolation factor is modest (1.3-1.5x, not 3x) -
The methodology is documented and consistent across campaigns - The brand understands and accepts the assumption
It doesn't make sense when: - The match rate is low (<40%) and the extrapolation multiplier is large - The matched population is known to be systematically different from the unmatched - The result is presented without methodology context
The bottom line
If you extrapolate, you must state the rules. Otherwise the proof collapses under scrutiny.
Report matched results first. Extrapolate with documented methodology. Be transparent about the match rate and the assumptions. Never present estimates as measurements.
Honest extrapolation with clear methodology builds more trust than inflated numbers without context. And trust is the currency that gets campaigns renewed.
Related Reading
- Service Model: Why Packaging Must Match Self-Serve vs Managed
- New-to-Category: The Underused Growth Metric in Retail Media
- Geo-Fencing Tests: The Simplest Store- Level Experiment in Retail Media
- Cost-to-Serve: The Hidden Reason Unbundled Pricing Breaks Operations
- Omnichannel ROAS: One Outcome Across Onsite, Offsite, and In-Store
Ready to see how this works in practice?
Footprints AI helps brands and retailers measure what matters. See our customer success stories or get in touch to discuss your retail media strategy.
