Answer in brief
Last-touch attribution gives the whole credit for an install to the most recent qualifying ad interaction before it. MMPs commonly use it by default because it produces one winner per install, which keeps one install from being counted for two channels in your report. Its weakness is that it ignores every earlier ad the person saw.
How last-touch picks a winner
The rule is simple: the latest eligible interaction before the install takes all of the credit. Eligible has a precise meaning, and the MMP checks it in order.
- It collects every touchpoint it can tie to the device or person: clicks and, if you allow them, ad views.
- It drops any touchpoint that is older than the lookback window for its channel and type.
- It ranks what is left. A click normally beats a view, and a match on an identifier normally beats a statistical one.
- Within the top rank, the most recent touchpoint wins. If nothing is left, the install is counted as an organic install.
Say a person sees a video ad on Monday, clicks a search ad on Wednesday and installs on Thursday. Under last-touch, the search ad gets the install. The video ad gets nothing, even if it was the reason the person searched.
Why it is commonly the default
Three practical reasons keep last-touch in place.
- One install, one owner. Your own report needs each install to belong to one channel. Without that, two channels are both credited for the same install and cost per install cannot be compared.
- It needs the least data. The MMP only has to know the latest interaction. Splitting credit needs a full history of the person's ad exposure, and on mobile that history is often incomplete.
- Everyone can check it. A network can look at its own click log and see why it won or lost. A weighted model is harder to audit.
Where last-touch misleads you
Last-touch rewards whoever is closest to the install. Channels that catch people who were already going to install, such as a search ad on your own brand name, look strong. Channels that create the interest look weak. Three checks reduce the distortion.
| Check | What it tells you |
|---|---|
| Assisted installs per channel | How often a channel appeared earlier in the path without winning. |
| Results after pausing a channel | Whether total installs fall, or the credit just moves to another channel. |
| Window length per channel | Whether a long window is letting old clicks claim installs that would have happened anyway. |
The alternative model is multi-touch attribution, which shares credit across the path. A common arrangement is to keep last-touch for billing and postbacks and read multi-touch reports beside it.
Last-touch and the ad platform's own numbers
An ad platform reports conversions by its own rules, which can include installs another channel won in the MMP. The gap between the two is a normal discrepancy and does not mean either system is broken. It grows when several channels reach the same people.
What to do next
Keep last-touch as your working model while you run a small number of channels. Write down the window you use for each channel so a change in the report can be traced to a change in settings. When you add a channel that works mostly through views, read how view-through attribution works before you decide whether views may win.
Questions and answers
Is last-touch the same as last-click?
Nearly. Last-click counts clicks only. Last-touch can also let an ad view win when no click qualifies, if view-through attribution is switched on.
Does last-touch attribution undercount upper-funnel channels?
It can. A channel that introduces people to the app but rarely gets the final click will win fewer installs than its influence suggests. Compare its assisted installs before you cut it.
Continue reading
Guide
Deterministic versus probabilistic matchingDeterministic matching ties an install to a click with an exact identifier. Probabilistic matching infers the link from device signals. When each is used.
Guide
View-through attributionView-through attribution credits an install to an ad that was shown but not clicked. How the view window works, the risk of over-crediting, and how to set it.
Guide
The Google Play Install ReferrerThe Google Play Install Referrer API gives an app the referrer URL, click time and install-begin time of its own install. How MMPs use it for attribution.
Guide
Re-engagement attributionRe-engagement attribution credits an ad for bringing an existing user back to the app. How it differs from install attribution and which settings define it.
Guide
Self-attributing networksA self-attributing network decides for itself whether its ad led to an install and tells the MMP the answer. What that means for your reports and settings.
How it works
Best MMP attribution starts with knowing how the match is madeHow an MMP attributes an install on Android and iOS, what lookback windows do, and which settings to check before you trust a report.
Cite this page
Airbridge. “Last-touch attribution and why it is the common default.” bestmmpattribution.com. Reviewed 2026-11-10. https://bestmmpattribution.com/last-touch-attribution