Attribution in Marketplace Advertising: What You Can and Can't Measure

Marketplace ad reporting is last-click, within a window — and that has real limits

Most marketplace ad platforms report sales using a last-click (or last-touch) attribution model within a defined window after the click — commonly something in the range of 7 to 14 days, depending on the platform and format. If a shopper clicks your ad and buys within that window, the sale is credited to the ad. This is straightforward to understand, but it systematically misses or misattributes several real patterns in actual shopper behavior.

What last-click attribution over-counts

Sales that would have happened anyway. If a shopper already intended to buy your product, searched for it by name, and happened to click your ad along the way (even though they would have found and bought it organically), the ad gets full credit for a sale advertising didn't actually cause. This is the core reason ACOS/ROAS alone can overstate advertising's real contribution — see ROAS vs. ACOS vs. TACOS for why TACOS partially corrects for this by looking at total revenue rather than only ad-attributed sales.

The last ad clicked, even when it wasn't the deciding factor. If a shopper clicked an ad for Product A, browsed away, later clicked an ad for Product B, and then bought Product A, most reporting will still credit whichever ad was clicked last before the specific purchase — not necessarily the ad that actually drove the decision.

What last-click attribution under-counts

View-through effects. A shopper who saw your ad (an impression, not a click) and was influenced by it, but didn't click, generally gets little to no credit in standard click-based reporting — some platforms report limited view-through data for certain formats (particularly Display), but it's typically far less complete than click-based data.

The "halo effect" of advertising on organic ranking and other products. Increased sales velocity from advertising can lift a listing's organic search ranking, which then drives additional organic sales that ad reporting doesn't credit to the campaign at all — this is part of why TACOS trending down over time (see ROAS vs. ACOS vs. TACOS) is a meaningful signal that standard ACOS reporting alone won't show you.

Cross-device and cross-session behavior. A shopper who sees an ad on their phone but completes the purchase later on a different device or in a different session may not always be captured as the same attributed journey, depending on the platform's tracking and identity-matching capabilities.

Sales on other products in your catalog. If an ad for Product A leads a shopper to browse your storefront and buy Product B instead, some platforms' attribution rules will credit that sale to the ad (a "halo" sale within your own catalog), while others attribute more narrowly to just the advertised ASIN/SKU — the exact behavior varies by platform and is worth confirming rather than assuming.

What this means for how you use the reporting

Use platform-native ACOS/ROAS for fast, tactical decisions — which keywords to keep, which bids to adjust — where the reporting is directionally useful even if not perfectly precise about causation.

Use TACOS and broader revenue trends for strategic decisions — whether to keep investing in advertising at all for a given product — since these are less distorted by attribution-window quirks and cannibalization.

Be skeptical of attribution-window changes when comparing performance over time. If a platform changes its default attribution window (lengthening or shortening it), historical comparisons before and after the change can look like a real performance shift when it's actually a measurement artifact.

Worked example

A seller notices their Sponsored Products ACOS looks excellent (18%, well under their 30% target) for a product that also happens to rank in the top 3 organic results for its main keyword. Because the product already ranks highly, many shoppers who click the ad were likely to find and buy it organically anyway — the ACOS looks strong partly because it's capturing sales advertising didn't need to create. The seller checks TACOS instead and finds it's roughly flat over the past two months even as ad spend has been raised meaningfully, a sign that the extra spend is mostly buying already-likely sales rather than creating new demand — a genuinely different conclusion than the standalone ACOS number suggested.

Mistakes to avoid

• Treating ad-platform-reported ACOS/ROAS as a complete measure of advertising's causal impact, rather than a directionally useful but incomplete signal.
• Comparing performance across a change in the platform's attribution window without accounting for the change itself.
• Ignoring halo/cross-catalog effects entirely when deciding whether an underperforming individual product's ad campaign is "worth it."

FAQs

Can I get a more accurate measure of advertising's true impact than platform attribution provides? Partially — TACOS is a useful directional corrective, and a deliberate incrementality test (see Incrementality Testing for Marketplace Ads) can get closer to a true causal estimate for sellers with enough scale to run one.

Does a shorter attribution window mean lower reported ACOS? Generally a shorter window will report fewer attributed sales (since fewer purchases happen within a tighter window after the click), which would tend to make ACOS look worse, not better, all else equal — so an attribution window change and its direction matters when comparing periods.

Should I trust ACOS at all if it has these limits? Yes, for its intended purpose — ACOS/ROAS remain useful for relative, day-to-day campaign optimization (is Campaign A more efficient than Campaign B, right now) even though they're imperfect measures of advertising's absolute causal contribution.