Influencer Marketing for Mobile Games: Tracking In-App Purchases and Player Value, Not Just Installs

Why an install alone tells you nothing

A creator campaign for a mobile game can generate thousands of installs at a low cost per install (CPI) and still be a bad investment — or the reverse: generate a handful of expensive installs that turn out to be the most profitable channel in the whole mix. The reason is that CPI only measures how cheaply you got someone to download the game — not whether those players ever open it again, how long they stay, or how much they spend. Two creator campaigns can post identical CPI and be wildly different businesses, because one attracts players who vanish after a single session while the other attracts players still playing and buying in-app purchases 90 days later.

What the industry actually manages by is a mobile measurement partner (MMP) — a third-party platform such as AppsFlyer, Adjust, Singular or Kochava that connects an install to what the player does afterwards: opening the game again, completing in-app purchases (IAP), or viewing ads if the game is ad-monetized. An MMP eliminates double-counting across channels and gives one consolidated view of which channels actually bring in valuable players — creator links included.

Two different worlds: iOS and Android

Attribution for mobile games works fundamentally differently depending on the platform — a distinction many brands and creators miss when comparing numbers across the two operating systems.

The consequence: a “D30 ROAS of 3.0” on Android and a “D30 ROAS of 3.0” on iOS aren't necessarily calculated the same way or with the same certainty — the iOS figure is built on aggregated, delayed SKAN data, while the Android figure is typically built on more direct, device-level data. Compare the two with that caveat in mind.

SKAdNetwork in practice: what it shows — and what it never shows

Per Apple's own framework (summarized by several independent attribution guides, since this research couldn't fetch Apple's developer documentation directly), SKAdNetwork reports installs and subsequent events through so-called conversion values and postbacks — never through individual user profiles:

• Up to three postbacks under the current SKAN 4 version, spread across three windows after install: 0–2 days, 3–7 days and 8–35 days.
• The first postback carries a “fine” conversion value (up to 64 possible values), allowing relatively detailed early behaviour to be encoded into that one value.
• The second and third postbacks carry only a “coarse” conversion value (typically low/medium/high — four buckets), to further protect user privacy the more time has passed.
• There's a delay of at least roughly a day before the first postback is sent, and the delay is deliberately randomized to prevent data from being traced back to a single user.
• The three postbacks are not linked by a shared identifier — so any real cohort analysis across the three has to be modeled by the MMP, not read directly out of Apple's data alone.

That makes SKAN fundamentally different from the tracking the Academy otherwise covers for ecommerce — see how influencer tracking actually works for cookie-based tracking. SKAN never identifies which specific user installed, let alone which specific creator actually drove a given install — only aggregated, campaign-level patterns.

D7, D30 and D90 ROAS: why cohorts, not one number

Because mobile-game revenue (in-app purchases, ad revenue) typically builds up over weeks and months — not in one single purchase the way ecommerce usually does — the industry measures ROAS as a cohort over time, not as one fixed number:

The point isn't that D90 is always “more correct” than D7 — it's that the two numbers answer different questions. D7 tells you whether a channel is even worth following further. D30 and D90 tell you whether it's actually making money.

In-game creator codes: a different tracking mechanic

Beyond link-based MMP tracking, an entirely different mechanic exists that several major game publishers use for creator partnerships: a code the player enters themselves inside the game after installing — not a link clicked before install. Epic Games' own Support-A-Creator program (read directly from Epic's own legal terms, legal.epicgames.com/fortnite/eula/sac, checked 27 August 2026) is the best-documented example: a creator is assigned a Creator Code as well as an attribution link. Per Epic's own terms, a user becomes an “Attributed User” when they either click the link or enter the code on a device where the user is signed into their Epic Account — and subsequent purchases from that user count toward the creator's payout. Epic's own terms disclose a payout threshold of $100 USD before amounts are paid out, but don't state a precise percentage or per-purchase rate in the terms themselves — that detail is left to the creator portal. Secondary sources across several fan and guide sites often cite a figure around “$5 per 10,000 V-Bucks spent,” but that isn't confirmed directly by Epic's own terms in this research and should therefore be read as reported, not officially confirmed.

The decisive difference from MMP-based tracking: a code entered in-game is self-reported and deterministic — there's no ambiguity about which creator gets credit, because the player types the code themselves. In exchange, it doesn't measure the earlier part of the funnel at all (how many people saw the content and clicked but never installed) — that part still requires an MMP. The two mechanics solve two different problems and shouldn't be confused with, or treated as a substitute for, one another.

Decision framework

IF you need to judge a creator campaign a few days after launch → use D7 ROAS as an early, directional signal — not a final verdict, unless the number is catastrophically low relative to the game's typical D7-to-D90 multiplier.

IF the game's monetization is IAP-heavy with a long conversion tail (strategy and RPG titles are typical examples) → weight D30 and D90 figures more heavily before making a final call on the channel.

IF the campaign runs on iOS with a low tracking opt-in rate → expect coarse, delayed SKAN data, and build that delay (up to 35 days for the final postback) into your timeline for when you can draw any conclusion at all.

IF a creator promotes via a link outside the app itself (a YouTube description, a TikTok bio) → confirm deep linking is correctly configured with your MMP before the campaign goes live — otherwise the click-to-install match rate can quietly collapse, and the creator's real effect disappears into “organic” traffic.

IF you're instead running an in-app creator code → treat it as self-reported, deterministic data — not a substitute for MMP-based cohort ROAS, since the code entry happens after install and therefore doesn't measure the earlier part of the ad funnel.

Worked example: why D7 alone can mislead (hypothetical)

The figures below are hypothetical and for illustration only — this is not a real Make Influence customer case, and none of the numbers are benchmarks for any specific game genre.

Two creator campaigns each cost $5,000 as a flat fee and generate identical revenue after 7 days: $2,000 each — a D7 ROAS of 0.4 for both. Looked at in isolation on D7, the two campaigns look equally weak and equally bad.

By day 30 the picture looks completely different: Campaign A kept generating revenue from the same players and reaches a D30 ROAS of 1.2, while Campaign B largely stalled and only reaches 0.44. Had the decision to continue or stop the channel been made on the D7 number alone, the two campaigns would have looked equally weak — and a premature “kill both” decision would have shut down what was actually the stronger channel.

Common mistakes

• Killing a campaign after 1–3 days based on D1 or D3 revenue alone in a game with a long monetization tail.
• Comparing an iOS SKAN-based ROAS directly with an Android deterministic ROAS as if they were calculated on the same basis.
• Assuming a low-CPI campaign is automatically a good campaign without checking D30 or D90 ROAS — cheap installs can be worthless players.
• Not getting deep linking configured correctly for creator links, so genuinely creator-driven installs get misattributed as organic.
• Treating an in-app creator code's redemption count as a full substitute for MMP tracking, even though the code says nothing about how many people saw the content without installing.

Make Influence's operational perspective

Make Influence doesn't run MMP-based UA campaigns for mobile games ourselves — our model is built for ecommerce, with cookie- and discount-code-based tracking and commission on tracked sales, not install and in-app-purchase tracking. But the underlying discipline the mobile-game industry practices with D7/D30/D90 cohorts is exactly the same one we recommend in a different context: a single number measured too early — whether that's day-1 CPI or ROAS without accounting for gross margin — hides more than it shows. We always recommend measuring a campaign's real effect over a time window that matches how long the customer (or the player) actually takes to show their full value — see also attribution windows in influencer marketing for the same principle applied to ecommerce.

FAQ

What is an MMP?

A mobile measurement partner is a third-party platform (e.g. AppsFlyer, Adjust, Singular, Kochava) that connects an app install to what the user does afterwards — across all your marketing channels, so you avoid double-counting and get one consolidated view of which channels actually work.

Can SKAdNetwork show me which specific creator drove an install?

No. SKAN reports aggregated, campaign-level data with no device or user identity. You can see that a campaign generated a certain conversion-value distribution — not which individual user or creator was behind a given install.

How long should I wait before judging a mobile-game UA campaign?

It depends on the game's monetization model. Use D7 as an early, directional signal, but wait until D30 — and for IAP-heavy titles ideally D90 — before making a final decision to scale or kill a channel.

Is Android attribution more reliable than iOS?

Android typically delivers more direct, deterministic data via Google Play's Install Referrer API, while iOS post-consent-requirement leans on SKAN's aggregated, delayed data. That doesn't make Android numbers “more correct” in themselves, but the two platforms shouldn't be compared as if they measure on the same basis.

Can an in-app creator code replace MMP tracking?

No. A creator code gives reliable, self-reported credit for purchases after install, but it doesn't capture the earlier part of the funnel — how many people saw the content and clicked without installing. The two mechanics solve different problems.

What's the difference between CPI and D30 ROAS?

CPI (cost per install) only measures how cheaply you got an install. D30 ROAS measures whether the installed players actually generated enough revenue to cover the cost 30 days later. A low CPI says nothing about player value — only D30/D90 ROAS does.