Retargeting measurement
The Retargeting Campaign That Counted Cart Returners As New Demand
A realistic retargeting scenario showing how recent cart intent can turn normal return behavior into an overstated lift and ROAS claim.
Archetype: Retargeting campaign measuring attributed return visits and purchase matchbacks
Bias mechanism: The treatment group is enriched for people who recently browsed, compared, saved, or abandoned a cart, so the campaign is credited for conversions that many users were already likely to complete.
Business model pressure
A retargeting program earns budget by showing efficient return visits, attributed purchases, and recovered carts from people who recently showed product interest. The advertiser wants to know whether the media created incremental demand or mainly reminded high-intent visitors who were already moving toward purchase.
Advertiser proof claim
A dashboard reports strong attributed ROAS from exposed visitors who returned and bought within the lookback window, but the core question is how many of those visitors would have returned through direct, email, organic search, price comparison, or saved-cart behavior without the retargeting impression.
Statistical result
| Metric | Naive read | Stratified read | Modeled benchmark |
|---|---|---|---|
| Purchase-rate lift | 29.6 pts | 0.9 pts | 0.6 pts |
| Attributed incremental purchases | 22,460 | 720 | 455 |
| Attributed purchase value | $2,156,186 | $69,091 | $43,690 |
| Attributed value after media cost | 5.26x | 0.17x | 0.11x |
The retargeted-cart dashboard is 31.2x larger than the recent-intent holdout estimate in this worked example.
Readout audit questions
Use this pass before the case becomes a budget argument, buyer proof point, or channel lesson. The goal is to separate what the dashboard observed from what the campaign plausibly changed.
| Check | Question to ask | Evidence in this case | Safer claim boundary |
|---|---|---|---|
| Credited outcome | What is the report counting as campaign impact? | Separate attributed outcomes from incremental outcomes. In this case, the naive lift is 29.6 pts while the stratified read is 0.9 pts. | Treat path, click, form, survey, or matched-outcome reporting as descriptive until the counterfactual is visible. |
| Comparison group | Were treatment and control groups comparable before the campaign? | The treatment group is enriched for people who recently browsed, compared, saved, or abandoned a cart, so the campaign is credited for conversions that many users were already likely to complete. | Ask for a protected holdout, balanced market, suppression test, or matched comparison before using causal verbs. |
| Budget threshold | Would the decision survive the adjusted read? | The naive read is 31.2x larger than the stratified estimate, and adjusted ROAS is 0.17x after media cost. | Use the result for diagnosis, repair, or test planning unless the adjusted evidence clears the decision threshold. |
| Next evidence | What would change the conclusion? | Review whether lookback windows and lag patterns credited demand that was already likely to return. | Name the missing evidence request before turning one worked example into a general rule. |
For the next review, pair this case with Attribution window and conversion lag checklist and the claim confidence rubric.
Cart-returner decision logic
Before this case becomes proof that retargeting recovered demand, read the result through the recent-intent audience. The useful question is whether comparable cart visitors would have returned and purchased without another ad.
| Review point | Evidence to request | Decision consequence |
|---|---|---|
| Eligible audience | The full pool of recent product viewers, cart abandoners, saved-item visitors, price-comparison visitors, suppression exclusions, and who had a fair chance to enter retargeting or holdout. | If the report starts only with exposed returners, describe the result as attributed return behavior rather than incremental demand. |
| Intent balance | Pre-period product views, cart value, recency of abandon, category visits, email or direct-return exposure, loyalty status, and discount sensitivity for both retargeted and holdout visitors. | If retargeted visitors had stronger prior purchase intent, remove causal lift and ROAS language until the comparison lane is balanced. |
| Timing window | Time from cart event to impression, impression to return, return to purchase, and whether normal checkout delay or reminder channels were credited to the ad. | If the purchase window mainly captures users already inside a natural return cycle, call the result recovered-cart attribution rather than created demand. |
| Suppression and leakage | Protected holdout records, frequency caps, email and search overlap, cross-device matching rules, and whether suppressed visitors still saw retargeting from another route. | If holdout protection is weak, use the case for measurement repair before expanding budget. |
| Claim boundary | A prewritten ladder separating reach, return visits, attributed purchases, recent-intent lift, and randomized incrementality. | If renewal depends on causal proof, require a recent-intent holdout or equivalent counterfactual before using sales-lift or ROAS proof language. |
Worked downgrade
The headline version of this case says retargeted cart visitors produced a 29.6 point purchase-rate lift, 22,460 attributed incremental purchases, and $2,156,186 in attributed purchase value. That version sounds decisive because the visitors really did come back, buy, and match to the campaign window.
The weaker read appears when the eligibility file shows that retargeted visitors were more likely to have recent carts, saved items, product comparisons, direct-return paths, email reminders, and higher pre-period category intent before the ad could change behavior. After comparing visitors within similar return-propensity strata, the estimated movement falls to 0.9 points and $69,091 in attributed value against $410,000 of media cost. The modeled benchmark is 0.6 points.
The safer readout sentence is: retargeted recent-intent visitors returned and purchased at a higher observed rate, but the evidence does not prove incremental demand because the campaign selected people who were already likely to complete the purchase. The next action is to rerun with randomized suppression inside the recent-intent audience, locked conversion windows, channel-overlap reporting, and a protected holdout before using lift or ROAS language.
The advertiser-facing story
The report looks persuasive because exposed visitors are already close to purchase. They viewed product pages, compared options, saved items, abandoned carts, or came from high-intent sessions before the retargeting touch. When those visitors return and buy, the attributed result can look like recovered demand, even if the ad mostly arrived during a normal decision delay.
What broke
Retargeting eligibility is not neutral assignment. It is built from signals that predict conversion even without another ad. If the comparison group includes people who did not recently show the same cart, product, price, or saved-item intent, the readout confuses audience selection and natural return behavior with media-caused lift.
Better design
Define eligible visitors before launch, randomize a protected holdout inside the same recent-intent pool, keep suppression and frequency rules fixed, and report lift against mature conversion windows. The readout should separate return visits, recovered carts, attributed purchases, incremental purchases, and the channels that would have reminded the visitor anyway.
Propensity-strata audit
The adjusted estimate compares retargeted visitors and comparison visitors within similar return-propensity strata. That does not replace a protected randomized holdout, but it shows how much of the headline purchase gap was carried by people who already had cart, product-view, price-check, or saved-item intent.
| Propensity stratum | Recent-intent visitor records | Retargeted purchase rate | Recent-intent holdout purchase rate | Within-stratum difference |
|---|---|---|---|---|
| 2 | 12,388 | 9.1% | 7.9% | 1.2 pts |
| 3 | 23,108 | 12.3% | 12.3% | -0.0 pts |
| 4 | 10,772 | 16.9% | 17.6% | -0.6 pts |
| 5 | 2,226 | 26.9% | 23.3% | 3.6 pts |
| 6 | 213 | 28.4% | 35.6% | -7.1 pts |
| 7 | 3,417 | 42.0% | 43.9% | -1.9 pts |
| 8 | 22,919 | 54.0% | 53.0% | 1.0 pts |
| 9 | 46,150 | 66.5% | 64.7% | 1.9 pts |
| 10 | 4,807 | 77.7% | 77.3% | 0.4 pts |
Takeaway
A strong retargeting readout should not stop at attributed purchases from people who returned after an ad. It should show the full recent-intent eligible audience, suppression rules, frequency exposure, conversion-lag maturity, prior-cart balance, and how much of the result survives a protected recent-intent counterfactual.
Move from this case to the next review.
Use the case to tighten claim language, choose a better measurement method, and review the next campaign readout without treating a worked example as a rule.