How to Choose a Mobile Measurement Partner for Your Game, and When to Switch: 2026 Pricing Models, an Evaluation Checklist, RFP Questions, and a Parallel-Run Pilot
A mobile measurement partner (MMP) is the official ledger that decides which channel delivered what for a game studio's marketing budget. Every ad network reports its own results, but when several networks claim the same install, it is the MMP that settles the credit on one source and sends that result back to the networks as postbacks. Choosing an MMP is therefore less like buying a tool and more like deciding what your UA team will count as "performance" for the next few years.
Yet MMP decisions are often made from a single price sheet or one sales demo. As of November 2026, the major MMPs publish free plans, per-conversion pricing, and monthly tiers, but free plans frequently do not assume paid media at all, cap the number of ad partners you can connect, or move raw data and fraud protection into higher plans or add-ons. Recalculate around the features a game studio actually uses, and the picture looks different from the first number on the price sheet.
This article is for studios adopting an MMP for the first time or moving from one MMP to another. It covers how to read public pricing models, an evaluation checklist and RFP questions, what breaks during a switch, and a parallel-run pilot that reduces the risk. How an MMP matches installs and why every dashboard shows a different number is covered in Why Every Dashboard Shows a Different Install Count: A Game Marketer's Guide to Mobile Attribution in 2026. This article takes those concepts as given and focuses on the purchase decision: which MMP to sign with.
What does an MMP do, and what does it not do?
An MMP connects ad touchpoints to installs and in-app events, assigns credit to one source, and delivers that result to both the networks and the studio on the same basis. What an MMP does not answer is whether the user would have installed without the ad.
This boundary matters in a purchase decision because sales materials tend to mix core features with add-ons in one feature list. Splitting what studios expect from an MMP into core, extended, and outside-the-MMP looks like this.
Category | What it covers | Note |
|---|---|---|
Core (baseline for every MMP) | Install and re-engagement attribution, in-app event collection, per-network postbacks, tracking links and deep links, self-attributing network (SAN) integrations | Quality differences here come mostly from integration depth and configuration flexibility |
Extended (often split into plans or add-ons) | Advanced fraud rules, raw data export (APIs, warehouse delivery), ad spend aggregation, ad revenue measurement, SKAN conversion value tools, incrementality (lift tests), MMM | Check whether the price sheet labels these "premium" or "add-on" |
Outside the MMP | Judging a channel's net incremental effect, budget allocation, creative production, in-game economy design | Covered by incrementality experiments, MMM, and internal BI |
Two points from this table are worth keeping. First, core features differ little among the major MMPs; the real differences lie in which plan includes the extended features and how deeply each MMP integrates with the networks you use. Second, even when an MMP sells incrementality or MMM as an add-on, it is a separate tool answering a different question from last-touch attribution. What attribution misses is covered in Media Mix Modeling for Mobile Games: Measuring What Attribution Can No Longer See, and experiment design in How to Design an Incrementality Test for Mobile Game UA: Holdouts, Geo-Lift, Sample Size, and Reading the Result.
How is MMP pricing structured in 2026?
As of November 2026, most MMPs that publish pricing charge per attributed conversion by default, alongside a free plan, monthly tiers, and custom enterprise contracts. What differs from vendor to vendor is what counts as a conversion and whether the free plan supports paid media.
The table below summarizes only what each vendor publishes on its official pricing page. Vendors are listed alphabetically; the order implies no ranking. Enterprise contract rates are not public and are left out.
Vendor | Published plans | Published price and limits | Billing unit and caveats |
|---|---|---|---|
AppsFlyer | Zero (free), Growth (pay-as-you-go), Enterprise (custom) | Zero and Growth both include 12,000 free conversions in the first year (Welcome Package); after that, Growth is $0.07 per conversion | A conversion is an install, re-engagement, or re-attribution driven by a campaign. Organic is not billed. Zero is stated as not suitable for apps with paid activities. Protect360 (fraud), Data Locker, and Raw Data API are labeled premium |
Branch | Basics, Essentials, Enterprise | Prices not published. Basics starts with a free trial | Basics is limited to 3 ad partners, 3 data partner webhooks, and 3 ad partner postbacks. CTV and OTT attribution are Enterprise features |
Kochava | Free App Analytics, Foundation, Enterprise | Free up to 10K conversions per month; Foundation from $500 per month (33K conversions per month); Enterprise from $2,000 per month (200K MAU per month) | The free plan centers on SAN attribution and owned media. Attribution for all integrated partners and media cost aggregation start at Foundation. Unlimited raw data queries and custom ETL and postbacks are Enterprise. MMM is quoted separately |
Singular | Free, Growth, Enterprise | Free up to 15,000 paid conversions; Growth at $0.05 per conversion | The free plan lists cost aggregation, fraud prevention, and SKAN attribution. Automated data exports (ETL) are a Growth add-on. Private cloud is an Enterprise option. The page does not state the time period for the free limit |
What the table makes immediately clear is that "free" means something different at each vendor. One free plan is a first-year conversion allowance, another is a limit that resets monthly, and another restricts the number of ad partners or attribution beyond SANs. For a studio running paid UA, it is more accurate to treat free plans as a way to integrate the SDK early and compare vendors on their paid plans.
Under per-conversion pricing, which events count as billable conversions moves the total more than the unit price does. In a contract where re-engagements and re-attributions count as conversions, scaling a retargeting campaign raises the billable count even if new installs stay flat. As an illustration, applying the public list range of $0.05 to $0.07 per conversion gives $2,500 to $3,500 per month at 50,000 paid conversions, and $10,000 to $14,000 per month at 200,000. Actual contracts vary with volume discounts and plan configuration, so these numbers are only a starting point for sizing the budget.
Monthly paid conversions (assumed) | At $0.05 list | At $0.07 list | What to check |
|---|---|---|---|
10,000 | $500 | $700 | Whether it falls within a free allowance, and whether the free plan supports paid media |
50,000 | $2,500 | $3,500 | Whether fraud protection and raw data are included or sold as add-ons |
200,000 | $10,000 | $14,000 | When enterprise terms apply, whether re-engagements are billed, volume discount tiers |
This calculation leaves out extended features such as fraud protection, raw data export, and incrementality. In the RFP, ask for the base unit price and, separately, the total with every feature your game will actually use switched on. Singular's own MMP selection guide also advises checking how pricing changes with volume and what is excluded from the base price.
How should a game studio evaluate an MMP?
For game studios, it works best to evaluate an MMP starting with Android measurement quality, then fraud protection, data access, the joining of ad spend and revenue data, support for iOS privacy frameworks, and data residency. For most studios, Android is where the bulk of measured purchases and ad revenue sits, and it is also where device-level attribution still anchors campaign decisions.
Criterion | Why it matters | How to check |
|---|---|---|
Play Install Referrer handling | The Google Play Install Referrer API passes the referrer URL, the click timestamp, and the install-begin timestamp to the app. The information is available for 90 days and does not change unless the app is reinstalled. It is the raw material for click-to-install-time fraud detection | Whether the SDK calls the referrer once on first launch, whether it retries on connection failure, and how referrer matching is prioritized against probabilistic matching |
Google Ads integration | Google certifies select MMPs through its App Attribution Partner (AAP) program; the list includes AppsFlyer, Branch, Kochava, Singular, Tenjin, and Airbridge, among others. Partners pass conversions, referrer, and consent status by SDK or server-to-server (S2S) | AAP certification, how EEA user consent status is passed, recommended conversion window alignment with Google |
Network coverage | If the MMP does not connect cleanly to the networks your game runs on, measurement gaps start on day one | Give the vendor your current and planned network list and check integration type (SAN or non-SAN), cost data collection, and postback event mapping one by one |
Fraud protection | AppsFlyer's 2026 fraud report analyzed more than 100 billion installs across 246,000 apps and found that 52% of fraudulent installs came from traffic classified as organic | Real-time blocking (rejection before attribution) versus post-attribution detection, custom rules, reports by rejection reason, and whether fraud protection is included or an add-on |
Raw data access | Calculating cohort LTV and retention by network directly in BI requires row-level data | Retention period for user-level logs, direct warehouse delivery, API rate limits and extra fees |
Spend and revenue aggregation | Seeing ROAS inside the MMP requires network spend and in-app purchase and ad revenue joined on the same cohort | Cost collection granularity per network (campaign, creative), handling of delays and retroactive corrections, whether ad revenue measurement is included |
iOS frameworks | Apple's AdAttributionKit is available from iOS 17.4 and supports both the App Store and alternative marketplaces. It supports re-engagement and three conversion windows and works alongside SKAdNetwork | Collection of SKAN and AdAttributionKit postbacks, conversion value design tools, joining cost with SKAN conversions |
Incrementality and MMM | Correcting channels that last-touch overstates requires separate experiments | Whether it is an add-on, a separate quote, and whether external experiment results can be imported |
Data residency and privacy | Depending on regional regulation and internal policy, data storage location can become a contract term | Choice of data center region, private cloud options, consent management integration |
A short note on iOS. According to Apple's documentation, AdAttributionKit and SKAdNetwork share ad network IDs, and when an app has impressions from both frameworks, the system sorts them together and picks a single winner. Click-through impressions always take precedence over view-through, and the system considers at most six impressions for any conversion. For install conversions, up to five non-winning ad networks also receive a postback. The question for an MMP is whether it collects postbacks from both frameworks and interprets them on the same screen. UA strategy in the post-IDFA environment is covered in Mobile Game User Acquisition in the Post-IDFA Era: What Has Changed and What Works Now.
One more thing to check on Android is installs that come from outside Google Play. The Install Referrer is an API provided by the Google Play Store, so installs from other stores or direct distribution are not measured the same way. If you plan to grow the share of alternative stores, ask separately how the MMP matches installs per store. The shift in Android distribution is covered in Rival Stores Now Live Inside Google Play: The New Map of Android Distribution.
Fraud protection is the item game studios most need to read alongside price. The same AppsFlyer report found that the fraud gap between affiliate traffic and SRNs reached 36x, and PPC Land's coverage of the report says overall Android fraud stayed in a 14 to 15 percent range throughout 2025. If your game runs offerwalls and rewarded media, confirm at the contract level whether fraud protection is part of the base plan and whether blocked installs are excluded from billable conversions. Offerwall fraud types and countermeasures are covered in How to Prevent Offerwall Fraud in Mobile Games: Understanding the Risk Is Where Effectiveness Begins.
What questions belong in an MMP RFP?
The purpose of an MMP RFP is not to tick feature boxes but to get, in the same format from every vendor, how the product actually works with your game's network mix and data flow, and what it costs in total. That means writing questions as "how, on which plan, and at what price" rather than "do you support."
Area | Example RFP questions |
|---|---|
Attribution methodology | What is the priority between deterministic and probabilistic matching (the attribution waterfall)? Can click and view windows be set per network? How are reinstalls and re-engagements deduplicated? |
Android | What is the fallback match when Play Install Referrer collection fails? How are installs from stores other than Google Play measured? How is EEA consent status passed in the Google AAP integration? |
iOS | Do you collect both SKAN and AdAttributionKit postbacks? Can conversion value mappings be shown back as event names? Can we see ROAS that joins SKAN conversions with spend? |
Network integrations | For each network on our list, what is the integration type, the cost data granularity, and the supported postback events? How long does a new network integration take? |
Fraud protection | What are the real-time blocking and post-attribution detection rules, and can we create custom rules? Are blocked installs excluded from billing? Can reports by rejection reason be shared with networks? |
Data access | How long are user-level logs retained? Is continuous delivery to a data warehouse available, and at what extra cost? What are the API rate limits? |
Spend and revenue | How are retroactive cost corrections and campaign renames reflected? Which plan includes ad revenue measurement? |
Incrementality and MMM | Are lift tests or MMM separate products? Can results from externally run experiments be imported and viewed together? |
Data residency | Can we choose the data storage region? On which plan are private cloud or regional separation options available? |
Pricing | What is the definition of a billable conversion (are re-engagements and re-attributions included)? What is the annual total at our expected volume with every feature we need switched on? How do contract length and overage rates work? |
Migration support | Do you support importing existing users' device IDs? How do you prevent double billing during the transition? Is onboarding with a dedicated engineer included? |
Reliability and support | Do you publish uptime history? Is support available in Korean and in our time zone? |
Once RFP answers are in, normalize them into a single table. Because vendors give the same feature different names, fix the rows in your own terminology and transcribe each vendor's answer as "included / higher plan / add-on / not supported." That shows where the total cost differences come from. If you used the same approach to select ad networks, it helps to match the format of the evaluation table in How to Choose a Mobile Advertising Platform for Your Game in 2026: Evaluation Criteria, RFP Questions, and Pilot Design.
What breaks when you switch MMPs?
The most common failures in an MMP switch are existing users being counted as new installs in the new MMP, and non-SAN network campaigns dropping or duplicating between the two MMPs. Both can be prevented with scheduling and process agreed before signing.
The first is re-counting existing users. When an app update with the new MMP's SDK ships, users who already had the game installed look like brand-new devices to the new SDK. In its marketing migration guide, Singular explains that importing the existing user base first prevents an organic spike and protects against being billed twice for the same install. The working rule is to load the list of existing users' device IDs before the new MMP receives live traffic. Skip this step and organic installs appear to spike for the first few weeks of the switch, distorting both the organic-to-paid split and blended CPI for that period.
The second is the timing of network integrations. According to Singular's guide, non-SAN networks cannot support two MMPs on the same campaign, so moving all non-SAN campaigns at once is how you keep the data clean. SAN networks vary: some can run with two MMPs, some cannot, and some only for installs. That is why the center of a migration schedule is the switchover date, on which tracking link replacement and postback configuration changes all happen together.
The third is the break in historical data and reporting. Attribution history from the old MMP generally does not carry over to the new one, so export cohort reports and raw data to your BI before the old contract ends. Without a window that both systems measured, you cannot tell whether a change in performance right after the switch is real or a change in measurement basis.
The fourth is replacing everything that was wired in. Tracking links and deep links, postback URLs registered with networks, the SKAN conversion value schema, in-app event name mappings, ad revenue integrations, and the data sources behind internal dashboards all change. Campaigns that use MMP conversion events as optimization signals, such as Google Ads, can see learning destabilize the moment the event source changes, so avoid structural campaign changes around the switchover date.
Risk | Symptom | Prevention |
|---|---|---|
Re-counting existing users | Organic install spike right after the switch, double billing | Import existing device IDs before live traffic |
Non-SAN networks cannot run two MMPs | Missing or duplicated installs for some networks | Move all non-SAN campaigns on the switchover date, test links in advance |
SAN-specific differences | Install and event attribution timing differs by network | Confirm dual-running support per network in advance and decide the order |
Historical data break | No before-and-after comparison | Export old MMP cohorts and raw data, secure an overlap window |
Optimization signal swap | Learning instability in event-optimized Google, Meta, and other campaigns | Freeze campaign structure around the switchover date, confirm event names and values match |
App version fragmentation | Users on older versions do not have the new SDK | Set the switchover date based on update adoption |
Game studios often overlook the last item. The new SDK reaches only users who install the update, so in a game with slow update adoption, the two SDKs will be looking at different user groups for a while. It is safer to set the switchover date not at contract start but when adoption of the version carrying the new SDK is high enough.
How do you design a parallel-run pilot?
The purpose of an MMP switch pilot is to confirm that the new MMP produces numbers within an explainable range of the old MMP for your networks and event structure. To do that, fix the comparison basis, the switchover date, and the rollback conditions before you start.
A parallel run means shipping both SDKs and measuring the same period on both sides. As noted above, however, non-SAN networks cannot attach two MMPs to the same campaign. During the parallel period, the old MMP remains the official ledger, and the new MMP is used to compare SDK data (organic, in-app events, revenue) and data from SAN networks that support dual running. Non-SAN networks move together on the switchover date.
Phase | Duration (example) | What to do | Pass criteria (example) |
|---|---|---|---|
Preparation | 2 to 4 weeks before start | Narrow candidates via RFP, document event name, revenue value, and SKAN conversion value mappings, export raw data from the old MMP, import existing device IDs, confirm dual-running support per network | Mapping document and per-network migration order finalized |
Parallel measurement | About 4 weeks | Ship a build with both SDKs, keep the old MMP as the official ledger, compare total installs, organic, key in-app events, and revenue daily | On matched windows, differences in total installs and revenue are explainable within a pre-agreed tolerance |
Switchover | Switchover date | Replace non-SAN tracking links and postbacks in one batch, switch SAN networks in sequence, freeze key campaign structure | Install and event postbacks confirmed from every network |
Stabilization | 2 to 4 weeks after the switch | Review gaps between network dashboards and the new MMP, check fraud rejection reports, switch internal dashboard sources | Per-network gaps at a similar level to before the switch |
Wrap-up | After stabilization | Ship a build without the old SDK, archive final data before the old contract ends | Historical data archived and comparison report complete |
It pays to fix the order in which you read differences during parallel measurement. First align attribution windows, time zones, and reinstall handling on both sides; then compare organic and SDK events (numbers that do not depend on networks); and only then look at per-network differences. If network-independent numbers already diverge widely, the problem is more likely SDK implementation or event definitions than integrations. The tolerance should not be a figure the vendor proposes; the studio should set it in advance at a level that would not change its own budget decisions.
You can also narrow the pilot's scope. A studio running several titles can switch one title with a fast update cycle and a simple network mix first, then schedule the rest based on the result, which keeps the risk small.
Preparation on the media side runs in parallel with all of this. Playio is an Android-based community of 5 million gamers who love games, and it connects games with players whose tastes fit, based on genre preferences, play history, and in-game behavior data. Like other networks, Playio campaigns can be set up so that installs and in-game events are verified through the tracking links and postbacks of the MMP the studio already uses. CPI is the default pricing model, and CPE is also supported on in-game actions such as tutorial completion, reaching a level, or hitting a playtime goal, so using an event already defined in your MMP as the success criterion lets you compare Playio against other channel cohorts on the same basis. If you have an MMP switch ahead, we recommend aligning the switchover date and link replacement schedule with your campaign calendar. As a first step, tell us which MMP you use today and the one in-game event you want to use as the success criterion, and we can work out a small test campaign built around that criterion together.
More details are available here. (https://playioadsen.oopy.io/bizdeck)
Key takeaways
An MMP is the official ledger that connects ad touchpoints to installs and in-app events, assigns credit, and returns postbacks to networks; a channel's net incremental effect and budget allocation sit outside it. As of November 2026, most public pricing is per attributed conversion: published list prices are $0.05 per conversion on Singular Growth, $0.07 per conversion on AppsFlyer Growth (after 12,000 free first-year conversions), and on Kochava a free plan of 10K conversions per month, Foundation from $500 per month, and Enterprise from $2,000 per month, while Branch does not publish prices. Free plans often limit paid media support, the number of partners, or attribution beyond SANs, so compare on the annual total with every needed feature switched on and on the definition of a billable conversion (whether re-engagements and re-attributions count). Evaluate in this order: Android Install Referrer handling and Google AAP integration, network coverage, fraud protection (whether it is included and whether blocked installs are excluded from billing), raw data access, spend and revenue joining, SKAN and AdAttributionKit, incrementality, and data residency. When switching, import existing device IDs before live traffic, move non-SAN networks together on the switchover date, and export historical data to secure an overlap window. With a parallel measurement period of about four weeks that compares network-independent numbers first, plus tolerances and rollback conditions set in advance, an MMP switch becomes a predictable project.
For inquiries about Playio's advertising solutions, reach out at: [email protected]
Sources (for data verification; whether to display them at publication to be decided in review):
AppsFlyer, Pricing (Zero, Growth, Enterprise; Zero and Growth include a 12,000 free conversion Welcome Package to be used within the first 12 months; Growth $0.07 per conversion afterward; a conversion is a campaign-driven install, re-engagement, or re-attribution, organic not billed; Zero not suitable for apps with paid activities; Protect360, Data Locker, and API labeled premium): https://www.appsflyer.com/pricing/?view=full
Singular, Pricing (Free up to 15,000 paid conversions at $0 per conversion; Growth $0.05 per conversion; Free lists cost aggregation, fraud prevention, SKAN attribution; automated data exports as a Growth add-on; Enterprise private cloud option and 365-day cohorts): https://www.singular.net/pricing/
Kochava, Pricing (Free App Analytics $0 per month, 10K conversions per month, SAN attribution, SKAN/AAK, fraud prevention; Foundation from $500 per month, 33K conversions per month, attribution for all integrated partners, media cost aggregation, row-level reporting; Enterprise from $2,000 per month, 200K MAU per month, unlimited raw data queries, custom ETL and postbacks; MMM quoted separately): https://www.kochava.com/pricing/
Branch, Pricing (Basics, Essentials, Enterprise, prices not published; Basics free trial; Basics limited to 3 ad partners, 3 data partner webhooks, 3 ad partner postbacks; CTV and OTT on Enterprise): https://www.branch.io/pricing/
Android Developers, Play Install Referrer Library (returns referrer URL, click timestamp, install-begin timestamp; available for 90 days and unchanged unless reinstalled; call once on first launch; reconnect on service disconnection): https://developer.android.com/google/play/installreferrer/library
Google Ads Help, App attribution partners (AAP program certifies select MMPs; list includes AppsFlyer, Branch, Kochava, Singular, Tenjin, Airbridge, among others; SDK or S2S integration passing conversions, referrer, and consent status; capture EEA consent first; align conversion windows): https://support.google.com/google-ads/answer/12961402?hl=en
Apple Developer, AdAttributionKit (iOS and iPadOS 17.4+; postbacks for ads in the App Store and alternative marketplaces; re-engagement conversions; three conversion windows; up to five non-winning networks receive postbacks for installs): https://developer.apple.com/documentation/adattributionkit
Apple Developer, Understanding AdAttributionKit and SKAdNetwork interoperability (compatible ad network IDs; impressions from both frameworks sorted together for one winner; click-through over view-through; at most six impressions considered per conversion): https://developer.apple.com/documentation/adattributionkit/adattributionkit-skadnetwork-interoperability
AppsFlyer, The State of Fraud for Marketers 2026 Edition (over 100 billion installs across more than 246,000 apps; organic accounts for 52% of fraud installs; affiliate-to-SRN fraud gap of 36x; fraud rates largely flat in 2025): https://www.appsflyer.com/resources/reports/mobile-ad-fraud-state
PPC Land, Organic traffic is now mobile advertising's biggest fraud blind spot, 2026-06-10 (citing the AppsFlyer 2026 report; 106.4 billion installs from Q1 2025 to Q1 2026; Android fraud held in a 14 to 15 percent range through 2025): https://ppc.land/organic-traffic-is-now-mobile-advertisings-biggest-fraud-blind-spot/
Singular Blog, MMP: how to choose a Mobile Measurement Partner (and why you need one), 2024-03-14 (evaluation criteria and questions; check how pricing changes with volume and what the base price excludes; uptime history; ease of migration): https://www.singular.net/blog/how-to-choose-mmp/
Singular Blog, Migrating MMPs? How to Successfully Launch Marketing Efforts with Singular, 2021-10-10 (importing the existing user base prevents an organic spike and double billing; set a switchover date; non-SANs cannot support two MMPs on the same campaign, so move them together; SAN dual-running support varies): https://www.singular.net/blog/mmp-migration/
Own calculations: monthly cost at 10K, 50K, and 200K paid conversions using the public list range of $0.05 to $0.07 per conversion; pilot phase durations and pass criteria (illustrative)