Lesson 3 of 5 · 8 min read · intermediate
Attribution models: who gets the credit?
Attribution decides which ad gets credit for a sale. Learn last-click, multi-touch and view-through models, their blind spots, and how fraud exploits them.
A customer sees a video ad on Monday, a social ad on Wednesday, searches the brand on Friday and buys. Which ad earned the sale? Attribution is the set of rules that answers that question. Because budgets follow credit, attribution quietly decides where billions of advertising dollars go next.
Think of a football goal. Who deserves credit: the striker who scored, the midfielder who passed, or the defender who won the ball? Last-click attribution gives the whole goal to the striker. Multi-touch attribution tries to share it. Incrementality asks a harder question: would the team have scored anyway?
From ad to sale: attribution
A shopper sees a video ad, then a social ad, then clicks a search ad and buys.
- Three ads, one person: A shopper sees a video ad, then a social ad, then clicks a search ad and buys.
- The sale: They place an $80 order. Now every ad platform wants the credit.
- Last-click says: search did it: Last-click attribution gives 100% of the credit to the final click. Simple, but it ignores everything that built the desire to buy.
- Multi-touch spreads the credit: Multi-touch attribution shares credit across the video, social and search ads using rules or models. Fairer, but it needs to track people across sites, which privacy changes have made harder.
- Incrementality asks the real question: Would they have bought anyway? A holdout group that sees no ads shows the baseline. The difference is the incremental effect, the sales the ads actually caused.
- Why fraud loves attribution: If credit goes to whoever touched the sale last, fraudsters can fake that touch. Click injection and cookie stuffing steal credit for sales that would have happened anyway.
The common models
| Model | How credit is shared | Tends to favour |
|---|---|---|
| Last click | 100% to the final click before the conversion | Search, retargeting, affiliates, coupons |
| First click | 100% to the first touch | Awareness and discovery channels |
| Linear | Equal shares for every touch | Long journeys with many touches |
| Time decay | More credit to touches closer to the sale | Late-funnel channels |
| Position-based | Most credit to first and last, the rest shared | Both ends of the funnel |
| Data-driven | Algorithm estimates each touch's contribution from paths that did and did not convert | Depends on the data it sees |
Multi-touch attribution (MTA) covers every model that shares credit across several touchpoints. Google Analytics 4 retired its first-click, linear, time-decay and position-based options in 2023, leaving data-driven and last-click, a sign of how the industry has shifted toward algorithmic models.
Clicks, views and the plumbing
Attribution relies on connecting an ad exposure to an action. That is done with a tracking pixel or tag on the advertiser's site, a Conversions API sending events server-to-server, or in apps a mobile measurement partner. A click-through conversion follows a click. A view-through conversion credits an ad that was seen, but not clicked, within a lookback window such as one day or seven days.
Why attribution is getting harder
- Signal loss: the decline of third-party cookies and Apple's App Tracking Transparency (from iOS 14.5 in 2021) mean fewer journeys can be stitched together across sites and apps.
- Privacy-preserving frameworks: Apple's SKAdNetwork and AdAttributionKit report conversions in aggregate, with delays and limited detail.
- Walled gardens: each large platform measures its own contribution, and their totals often add up to more than 100% of actual sales.
- Offline and CTV: TV ads and in-store purchases are hard to link to individual exposures without panels or clean rooms.
How fraud feeds on attribution
Any rule that assigns money can be gamed. In mobile, click spamming and click injection exist to win last-click credit for installs they did not cause. In affiliate marketing, cookie stuffing plants tracking cookies to claim commissions on organic sales. On the web, bots that visit a retailer and then load ads can make retargeting look effective. This is attribution fraud: the sale is often real, but the credit is stolen.
A quick example of how much the model matters. Suppose a shopper in Singapore sees a YouTube ad, clicks a social ad two days later, and finally clicks a branded search ad before buying a $100 item. Last click gives search $100 and the others nothing. Linear gives each about $33. Position-based might give $40 to the video, $40 to search and $20 to social. Same sale, same data, three very different budget recommendations.
Using attribution well
- Pick a model on purposeKnow which channels your model favours, and do not let it be the only input into budgets.
- Clean the inputsRemove invalid clicks and impressions before attribution, not after. Fraudulent touches corrupt every model.
- Deduplicate across platformsCompare platform-reported conversions with your own sales records to find over-claiming.
- Calibrate with experimentsUse lift tests to check whether channels that look strong in attribution actually drive extra sales.
Key takeaways
- Attribution is a rule for sharing credit for a conversion among the ads a customer encountered.
- Last click favours bottom-funnel channels; multi-touch and data-driven models spread credit but still show correlation, not cause.
- Signal loss, privacy frameworks and walled gardens make journeys harder to stitch together.
- Attribution fraud steals credit for real sales, so filter invalid traffic before attributing and validate with experiments.
Questions people ask
What is last-click attribution?
Last-click attribution gives 100% of the credit for a conversion to the final ad a customer clicked before buying or signing up. It is simple and was the default for years, but it overvalues channels like branded search, retargeting and coupon sites that catch people already about to buy, and it ignores earlier ads that created interest. It is also the model most exploited by click fraud.
What is the difference between attribution and incrementality?
Attribution assigns credit for conversions to the ads that preceded them using a rule, such as last click or a data-driven model. Incrementality measures how many conversions an ad actually caused by comparing people who saw it with a similar group who did not. Attribution describes the path; incrementality tests cause and effect, and is used to check whether attribution is telling the truth.
What is a view-through conversion?
A view-through conversion is a sale or action credited to an ad that the person saw but did not click, within a set lookback window, often one to seven days. It helps value display, video and CTV ads that rarely get clicked, but long windows can over-credit ads, including unseen or fraudulent impressions. Require viewability and invalid traffic filtering for view-through credit.