A customer sees your Meta ad on Instagram. Two days later, they click a Google Shopping ad. The next day, they type your store's name straight into Google and buy. Which campaign "caused" the sale?
That's the fundamental question of marketing attribution. And the answer you choose determines where you invest your budget — and therefore, directly, your profitability.
Why attribution is a problem
On average, an e-commerce shopper has 4 to 8 touchpoints with a brand before buying. They see ads, read reviews, visit the site several times, get an email, see an organic post, and so on.
The problem: each ad platform uses an attribution model that favors itself. Meta claims the sale if the user saw an ad in the last 24 hours (view-through) or clicked in the last 7 days. Google claims the same sale if the last click came from Google Ads.
Result: the sum of conversions attributed by each platform is far higher than the actual number of conversions. You have no idea of each channel's real contribution.
The main attribution models
Last-click
100% of the credit goes to the last touchpoint before the purchase. It's the default model in Google Analytics and most tools.
Pro: simple and objective. Con: it completely ignores prospecting work. Meta awareness campaigns that start the journey get no credit, which pushes you to underinvest at the top of the funnel.
First-touch
100% of the credit goes to the first touchpoint. If the customer discovered your brand through a TikTok ad, TikTok gets all the credit even if the purchase happened 3 weeks later through Google.
Pro: values acquisition. Con: ignores the role of retargeting and conversion campaigns.
Linear
Credit is split evenly among all touchpoints. If the journey has 4 interactions, each one gets 25%.
Pro: recognizes every step. Con: doesn't distinguish an intentional click from a passive impression.
Time decay
Touchpoints closest to the conversion get more credit. An interaction 2 days before the purchase is worth more than one 14 days before.
Pro: a good compromise between first-touch and last-click. Con: still undervalues prospecting campaigns.
Data-driven
Machine learning algorithms analyze every conversion path to determine each touchpoint's marginal contribution. It's the model Google Ads and Meta use internally.
Pro: theoretically the fairest. Con: it's a black box, each platform favors its own channels, and it needs a minimum volume of data.
The fundamental problem with digital attribution
Here's the truth nobody wants to hear: no attribution model is perfect. They're all approximations of an endlessly complex reality.
The reasons:
- Cross-device is incomplete: a journey that starts on mobile and ends on desktop is often seen as two different visitors
- Offline is invisible: word of mouth, friends' recommendations and conversations on Discord/Telegram don't show up in any model
- The data is incomplete: with iOS 14, ad blockers and private browsers, you only see a fraction of the real journey
- Causation vs correlation: seeing an ad doesn't mean the ad caused the purchase
The pragmatic approach: triangulation
Rather than looking for the perfect model, the best e-commerce merchants use a triangulation approach:
- Blended ROAS / MER: your total revenue divided by your total ad spend. No attribution, just the overall economic reality.
- Incrementality tests: pause a campaign for 7 days and measure the real impact on revenue. If revenue drops by €5,000 when you cut €2,000 of ads, the incremental contribution is real.
- Post-purchase survey: ask customers directly, "How did you hear about our brand?" Simple and often revealing.
- Geographic analysis: compare performance between regions where you advertise and regions where you don't.
What ProfitPilotPro calculates instead of a miracle model
No attribution model that "solves" the journey. The dashboard cross-references Shopify orders with Meta, TikTok and Google spend:
- blended ROAS (revenue / ad spend) and true ROAS (net profit / ad spend), during the trial and on Starter, Pro and Agency — not on the Free plan
- server-side tracking of purchases, during the trial and on Starter, Pro and Agency — not on the Free plan
- the customer breakdown by source, from the Pro plan. It isn't LTV by channel
The trial lasts 14 days, no credit card required. Create an account.