If you add up the installs Meta claims, the installs Google claims, and the installs your other ad networks claim, you'll find you "acquired" far more users than actually downloaded your app. Each platform reports its own numbers, and each one happily takes credit for the same person. Attribution tracking in India is simply the work of figuring out which ad actually caused a download or a sale — so you stop paying for the same install three times. Get it wrong and you're running a business on numbers that don't add up. And in India, where you're scaling ad spend on thin margins, that's an expensive way to fly blind.
This is the 2026 guide, in plain language: what a mobile measurement partner actually does, how the three big tools (AppsFlyer, Adjust, and Branch) compare, what changed on iPhones with SKAdNetwork 5, and why the winning answer is a mix of tools rather than any single one.
Each ad platform claims the same conversions, so summing their dashboards over-counts reality. The job of attribution is to find the one truth underneath the competing claims.
Attribution tracking — what an MMP does
An MMP, or mobile measurement partner, is an independent tool that counts your app installs fairly across every ad platform. Think of it as a neutral referee that sits between your app and all the ad networks you run. It does three jobs. First, it de-duplicates installs — that means if Meta and Google both claim the same download, the MMP works out who actually earned it so you only count that install once. Second, it ties post-install events (a signup, a purchase, a subscription) back to the ad that drove them. Third, it stitches the limited, privacy-protected data Apple gives you (SKAdNetwork — more on that below) together with estimated data into one clear picture.
If you're spending real money on getting users across more than one network, an MMP isn't optional. Without one, the numbers each network reports about itself become fiction the moment you add them together. You also can't work out your true cost per install (CPI — the average ad spend to get one download) or your true cost per event, because you simply don't know either number.
AppsFlyer vs Adjust vs Branch — which fits
There's no universal "best." Each one is strongest at a different thing — so match it to how your app actually grows.
| MMP | Leads on | Best for |
|---|---|---|
| AppsFlyer | Integration breadth (10,000+ partners), iOS feature depth | Apps wanting the widest network coverage & deepest iOS tooling |
| Adjust | Simpler setup, transparent flat-rate pricing, raw SKAN data access | Teams who want clean data & predictable cost |
| Branch | Deep linking — web-to-app, referrals, re-engagement | Apps where linking & journeys are a core growth engine |
The practical read: AppsFlyer has the edge on breadth — it plugs into around 10,000+ ad networks and tools versus Adjust's ~2,000 — and on iPhone-specific features. The trade-off: it charges per install, so the bill grows as your volume grows. Adjust wins on being simpler to set up, having flat, predictable pricing, and giving you clean access to the raw Apple data. Branch is the pick when deep linking is central to how you grow. (Deep linking is what sends a user straight to the right spot inside your app instead of a generic home screen — for example, tapping a "share this product" link and landing on that exact product. It powers referrals, web-to-app journeys, and re-engagement.) Choose on how your app actually grows, not on whichever name you hear most.
The best MMP isn't the most popular one. It's the one whose strength matches how your app actually grows.
The iPhone problem — SKAdNetwork 5
iPhones (iOS) are where attribution gets genuinely hard. To protect user privacy, Apple won't let advertisers follow individual people around. Instead, all the measurement runs through Apple's own system, called SKAdNetwork (often shortened to SKAN). The latest version, SKAdNetwork 5 — part of the 2026 iOS update — is the biggest recent improvement. In plain terms, it sends back more detail after a campaign (these reports are called "postbacks"), lets you measure a wider range of in-app actions, gives you a longer window to count a result, and does a better job of crediting ads people saw but didn't tap. Apple is also nudging advertisers toward a newer system called AdAttributionKit for some of this.
It's a real step up from the early SKAN days, but the basic reality hasn't changed: iPhone data is delayed and partly estimated — not a clean, person-by-person trail. You no longer get a straight line from "saw the ad" to "installed" to "bought." So iPhone campaigns have to be planned on cohort-level economics — judging whole groups of users who arrived in the same week or from the same campaign, rather than tracking each one. That's a different playbook from Android.
Android vs iPhone, in practice
Android still lets you measure things in fairly precise, person-level detail (within limits that keep tightening), so you can optimise on specific in-app actions. On iPhone you steer by groups and estimates instead. Same app, two measurement playbooks — and pretending iPhone works like Android is a fast way to misread your numbers.
The GUROB attribution stack — hybrid by default
The single most important truth for 2026: attribution is not a one-tool problem. Lean only on Apple's SKAN data and you're slow and missing detail. Lean only on "probabilistic" methods — educated statistical guesses about who likely installed, rather than a confirmed match — and you take on privacy and compliance risk. Lean only on your MMP dashboard and you leave gaps. The brands that win stitch the pieces together.
- Pick the MMP that matches your growth. Want the widest reach and the best iPhone tooling → AppsFlyer. Want clean data and predictable pricing → Adjust. Built on deep linking → Branch. This is the foundation, not the whole answer.
- Run SKAN properly on iPhone. Set it up to record the actions that actually matter to you — a purchase, a subscription, a key signup — and plan iPhone campaigns on group-level economics, not the person-level numbers you get from Android.
- Cross-check everything against one another. Compare your MMP data, Apple's SKAN reports, the estimated data, and your own backend sales records. Treat the ad platforms' own numbers as a steering wheel for individual campaigns — never as the truth for your books.
- Build a number finance trusts. The end goal is one agreed figure for cost per install and per event that your finance and product teams actually believe. This mixed approach is the baseline now, not an experiment.
This measurement discipline underpins everything in our app marketing work — because optimising toward platform-reported numbers means optimising toward a lie, and at Indian-scale spend on thin margins, that's the difference between profitable growth and expensive guessing. We work on performance, so the real number is the one we're accountable to.
Common attribution mistakes
- Trusting each platform's own numbers. Adding up Meta's and Google's claimed conversions counts the same person twice. Use an independent MMP.
- Skipping an MMP at scale. Running across several networks without de-duplication means you can't know your true cost per install or per event.
- Treating iPhone like Android. iPhone data is delayed and partly estimated. Plan it on groups and Apple's SKAN, not the precise person-level data Android gives you.
- Relying on a single tool. SKAN-only, MMP-only, or guesswork-only all leave gaps. A mix is the standard.
- Choosing an MMP by reputation. The right tool is the one whose strength — reach, simplicity, or deep linking — matches how your app grows.
Frequently asked questions
In closing
Good attribution tracking in India isn't about finding the one perfect tool. It's about refusing to trust each platform's own numbers, and building a setup that combines your MMP, Apple's SKAN data, and estimated data into a single, believable figure. Pick the MMP that matches how your app grows, plan iPhone campaigns by group, and report one blended number your finance team believes. Get that right and every other decision finally rests on real data.
Want us to audit your measurement setup and tell you how far off your reported numbers are from reality? Book the 45-minute private audit (free). More on our app marketing work here.