Install attribution is the process of assigning credit for each new app install to the ad, campaign, and network that drove it. It's the central product of every MMP and the foundation of every UA decision — without reliable attribution, you can't know which campaigns are working, you can't optimize bids, you can't allocate budget across channels.
The problem is harder than it sounds. A typical install journey: the user sees a TikTok ad, doesn't click; sees a Meta retargeting ad on day 3, clicks but doesn't install; searches the brand on day 5, sees the App Store listing organically, installs. Which source gets credit? Different attribution rules give different answers — and most ad networks will claim the install if asked, so without a neutral adjudicator (the MMP), multiple networks get credited and you pay multiple times for the same user.
Common attribution rules
- Last-click attribution (default for most MMPs): the last ad the user clicked before install (within window) gets 100% credit. Simple, easy to implement, generally aligns with "the ad that closed the install".
- View-through attribution: credit given when a user saw an ad but didn't click, then installed within a shorter window (1-24 hours). Weaker signal than click-through. Often used as a tiebreaker when no click attribution applies.
- Multi-touch attribution: credit distributed across multiple touchpoints (linear, time-decay, position-based, data-driven). More accurate in theory, harder post-ATT.
- First-click attribution: opposite of last-click — first ad the user clicked gets credit. Less common; sometimes used for "demand generation" measurement.
- Engaged-view attribution: video ads watched to a defined completion threshold (often 6 seconds or 50% completion) credit the install. Common in YouTube and TikTok attribution.
Post-ATT iOS attribution is a patchwork
- SKAN postbacks for installs from SKAN-enabled ad networks (most major networks now).
- MMP device-level attribution for users who opted in via ATT (20-40% of users).
- Apple Search Ads Attribution API for Apple Search Ads campaigns specifically.
- First-party signal — CRM matching, web-to-app journey continuity via deferred deep linking, user-entered identifiers.
- Meta and TikTok internal attribution — running their own user-level attribution within their walled gardens.
The MMP's job is to ingest all of these, deduplicate them (so the same install isn't credited to multiple sources), and produce a unified view.
Android attribution is simpler because GAID-based deterministic attribution still works for most users (~80-90% of Android users don't opt out of advertising ID). MMPs primarily use install-referrer (Google Play's official mechanism for click-to-install attribution) plus device-level matching. Privacy Sandbox for Android (phased rollout) will eventually introduce SKAN-like aggregated attribution by ~2027.
Attribution rule comparison — when each applies
| Rule | How it works | Strengths | Weaknesses |
|---|---|---|---|
| Last-click | Last ad clicked within window gets 100% credit | Simple, unambiguous, MMP default | Over-credits closing-the-deal campaigns; ignores upstream awareness |
| First-click | First ad clicked within window gets 100% credit | Credits demand-generation work | Over-credits initial awareness; ignores conversion-driving touches |
| View-through | Ad impression (no click) credited if install happens within shorter window (1-24h) | Captures brand / video impact | Weaker signal — user saw but didn't engage |
| Multi-touch (linear) | Credit split evenly across all touchpoints | Recognizes funnel contribution | Hard cross-network post-ATT; linear weighting is naive |
| Multi-touch (data-driven) | ML model assigns weights based on observed conversion correlation | Most accurate when working | Requires extensive data + platform support |
| Engaged-view | Video ads watched to defined completion (e.g., 6s or 50%) credited | Bridges click + view-through | Common on YouTube + TikTok; less standardized cross-platform |
MMP default is last-click within window (typical 7-30 days). Mature programs use multi-touch + incrementality testing on top of last-click for strategic decisions, but operationalize on last-click for daily UA bidding.