Quick answer
A mobile measurement partner (MMP) is an independent company that matches ad clicks and impressions to app installs, credits one ad network per install and filters fraud. In-app behaviour analysis is a separate job, and the two only line up when they use the same event names.
What is a mobile measurement partner?
A mobile measurement partner (MMP) is an independent company whose SDK and servers decide which ad, if any, led to each app install, and then report that result to the advertiser and to the ad networks involved. "Independent" is the point: the ad networks are graded by a referee they don't own, using the same rules for every network.
An MMP's job has clear edges. An MMP matches ad clicks and impressions to installs and filters fraud; in-app behaviour analysis is a separate job, and the two only line up when they share the same event names. Most of the confusion around MMPs, and most of the "why don't these numbers match" tickets, come from treating those two jobs as one.
This page explains what an MMP does and doesn't do, how Apple's privacy frameworks changed the job, and how to make MMP data and product analytics agree. For what product analytics measures inside the app, see mobile app analytics.
How an MMP works
- The app ships with the MMP's SDK. On first launch, the SDK reports an install with device signals and any referrer data the platform provides.
- Ad networks report engagements. When someone sees or taps an ad, the network sends the MMP the click or impression, usually through a tracking link or server-to-server call.
- The MMP matches. It looks for a click or impression that fits the install within the attribution window. On Android, the Google Play Install Referrer API lets the app securely retrieve the referrer URL the Play Store recorded for that install, with click and install timestamps. On iOS, matching to a specific user needs App Tracking Transparency permission; without it, the MMP relies on Apple's aggregated attribution frameworks.
- The MMP picks one winner (typically last click within the window, sometimes a view-through impression) and labels everything else organic.
- Results go out as postbacks. The MMP tells each network which installs and in-app events it gets credit for, so the network can optimise bidding, and shows the advertiser a cohort report per channel and campaign.
- Fraud checks run throughout. Clicks injected seconds before an install, installs from emulators and impossible device patterns are rejected before they are credited.
What an MMP does and doesn't do
| An MMP does | An MMP doesn't do |
|---|---|
| Match ad clicks and impressions to installs and re-engagements | Explain why users drop off after onboarding |
| Pick one winning network per install and count the rest as organic | Analyse funnels, paths or feature adoption inside the app |
| Send postbacks to ad networks so they can optimise | Replace your product analytics' retention and engagement reports |
| Detect click injection, click spamming and fake installs | Run A/B tests or change the app |
| Handle deep links and deferred deep links into the app | Guarantee that its install count equals your first-open count |
| Receive and forward Apple's aggregated attribution postbacks | See users who never interacted with a tracked ad as anything other than "organic" |
The right column isn't a criticism. Those jobs belong to product analytics, and MMP reports aren't built for them: they're grouped by ad network and campaign, not by user behaviour.
SKAdNetwork, AdAttributionKit and ATT
Apple's developer documentation defines the limits an MMP now works within on iOS:
- App Tracking Transparency (ATT). An app that collects data and shares it with other companies to track people across apps and websites must ask for permission through the ATT framework. Without permission, user-level matching between an ad network's click and your install isn't allowed.
- SKAdNetwork. Apple's install-validation API "helps advertisers measure the success of ad campaigns while maintaining user privacy." The device, not the MMP, decides the winning ad. Postbacks go to the ad network (and to the developer if they opt in) with a level of detail set by a crowd-anonymity data tier. Since iOS 16.1, apps can update conversion values during three conversion windows, which gives up to three postbacks for an ad signed with version 4.
- AdAttributionKit. Available from iOS 17.4, it covers ads in the App Store and alternative app marketplaces, adds re-engagement attribution, and Apple's SKAdNetwork page now tells developers to use AdAttributionKit for app ad campaigns.
The practical effect is that on iOS an MMP increasingly forwards aggregated, delayed results that the device produced, rather than matching individual users itself. Campaign-level ROI still works; per-user ad history mostly doesn't.
MMPs and product analytics: who does what
| Question | Owner | Why |
|---|---|---|
| Which network and campaign brought this install? | MMP | It sees the ad engagements and runs the match |
| Should we raise the bid on campaign X? | MMP data plus revenue | Needs cost, installs and downstream value by campaign |
| Where do new users drop out of onboarding? | Product analytics | Needs every screen and event, for all users, paid or organic |
| Do users from campaign X retain better? | Both, joined by campaign | MMP supplies the source; analytics supplies retention |
| Which features drive repeat use? | Product analytics | Covered in app engagement metrics |
The last-but-one row is where the shared event names matter. If the MMP tracks af_purchase and the app's analytics tracks order_completed, nobody can confirm that the 312 purchases the MMP credits to a campaign are the same 312 purchases product analytics sees. Agree on one tracking plan: the same event name, the same trigger point (payment confirmed by the server, not the "Buy" tap) and the same currency, sent to both tools.
Worked example: 1,000 installs vs 940 first opens
Sample data: for one day, the MMP dashboard shows 1,000 installs, and product analytics shows 940 first opens for new devices. Neither is broken. Reconciling the 60-install gap:
| Reason | Installs | Why it happens |
|---|---|---|
| Day boundary | 25 | The MMP reports in UTC; the analytics project uses its own time zone, so late-evening installs land on the next day |
| No analytics consent or opt-out | 20 | The user declined analytics collection before the analytics SDK sent anything; the MMP counted the install under its own setup |
| Reinstalls | 15 | The MMP counted a reinstall as a new attributed install within its re-attribution rules; analytics recognised a device it had already seen and didn't count a new user |
| Total gap | 60 | 1,000 − 25 − 20 − 15 = 940 |
Two checks make this repeatable: compare the same time zone, and compare like with like (attributed installs vs new devices, not vs all launches). If the remaining gap is above a few per cent after those two, look at SDK initialisation order and at events that only one of the tools receives.
Common misconceptions
- "The MMP is our analytics." It answers "which ad?" It doesn't answer "what did users do next?", and it sees organic users only as "organic", with no behaviour behind the label.
- "MMP installs should equal store downloads." Store consoles count downloads; MMPs count first launches they can see. People who download and never open the app never reach the MMP.
- "SKAdNetwork gives user-level data." It gives campaign-level postbacks with delayed, tiered detail; that's the point of it.
- "Higher match rate means better measurement." A very high iOS match rate without ATT consent can mean signals that Apple doesn't allow; Apple's guidance also forbids fingerprinting devices.
Credit for later conversions on your own site or app, across several sessions, is a different model again; see attribution.
MMPs and TapCub
The usual question is "we already pay for an MMP, do we still need product analytics, and how do they connect?" Using the example above:
- TapCub isn't an MMP. It doesn't match clicks or impressions to installs and isn't an attribution provider. The iOS SDK doesn't read the IDFA and doesn't show an ATT prompt.
- Receive the MMP's result. Point your MMP's install postback at the endpoint on the MMP tab of the app's business settings. Each postback becomes an
$app_installevent carrying fields such asmedia_source,campaign,install_source_statusandmatch_type. App analytics › Quality & attribution › Attribution then splits installs into organic, non-organic and unattributed, by media source and campaign, and the header shows whether the MMP is connected. - Follow those users into the app. App analytics › Users › Retention and the version and channel reports show whether a campaign's users come back, and All-platform analytics › Behavior › Behavior analysis runs funnels on the same events. The Installs tab of All-platform analytics › Business › Channel ROI groups the postback installs by source next to the spend you import.
- Keep names aligned. Record the shared events in All-platform analytics › Data › Tracking plan so the purchase the MMP credits and the purchase TapCub counts are the same event. To track daily monetisation per active user alongside, see ARPDAU.
See also: Mobile Marketing Strategy: Choose the Number Before the Channels.
See also: What is a postback? Server-to-server postbacks in ads and app attribution.
See also: What is click-to-install time (CTIT)? How short CTIT flags install fraud.
Sources
- Apple Developer, SKAdNetwork.
- Apple Developer, AdAttributionKit.
- Apple Developer, App Tracking Transparency.
- Android Developers, Play Install Referrer.
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Customer Experience Consultant
Writes about live chat, visitor conversations, and self-service, sharing practical ways to make customer communication clearer and more effective.