Content
Try for free

Cohort Analysis for Mobile Apps: A Practical Guide

Cohort analysis for mobile apps means grouping users by a shared starting event, such as install date, trial start date, or the ad campaign that brought them in. You then track each group over the same time windows. Instead of one blended retention number, you see whether users who joined in March behave differently from users who joined in June, and whether users from one channel stay longer than users from another.

For subscription apps, the useful version of this is a cohort table where rows are start periods and columns are weeks or months since start. Cells show retention, paid conversion, renewal rate, or revenue per user. A flat blended average hides the difference between a good cohort and a bad one. A cohort view shows when something broke, which channel is worth scaling, and whether a product or paywall change helped.

What a cohort chart shows and how to read it

A cohort chart has two axes. Down the side are cohorts, usually one row per week or month of acquisition. Across the top is time since the cohort started: day 1, day 7, day 30, or month 1, month 2, and so on. Each cell is the share of that cohort still active, still subscribed, or still paying at that point.

Read it in two directions:

  • Across a row to see the shape of one cohort's decay. Most apps lose the largest share of users early, then the curve flattens. The level where it flattens matters more than the first-week drop.
  • Down a column to compare cohorts at the same age. If the month 2 column gets worse in every newer row, something changed in acquisition, product, or pricing.

Keep cohort size visible next to each row. A cohort of a few dozen users will swing wildly, and a good-looking percentage on a tiny group is not evidence of anything.

Which cohorts to build for a subscription app

Start with the cohort definitions that match decisions you actually make.

  • Acquisition date. The default. Weekly cohorts work for apps with steady traffic, monthly cohorts for smaller ones.
  • Acquisition source. Meta, TikTok, Snapchat, search, organic, and creators behave differently. Cohorts by channel and by campaign tell you where to put budget.
  • Plan or price. Weekly versus annual plans, or two price points in an A/B test, produce very different renewal curves.
  • Onboarding or paywall variant. If you test flows, keep each variant as its own cohort so you can compare paid retention and not only first purchase conversion.
  • Geography and language. Users in different countries often differ in conversion, refunds, and renewals.
  • Trial start versus paid start. A trial cohort and a paid cohort measure different things. Do not mix them.

Pick the cohort that matches the question. If you are deciding where to spend, cut by source. If you are deciding whether a new paywall worked, cut by variant.

Metrics to track inside each cohort

Retention is the starting point, but for a subscription business the revenue-side metrics matter more.

  1. Activation: the share of installs or signups that complete the key first action, such as a first workout, first lesson, or first generated result.
  2. Trial-to-paid conversion: the share of trial starters that become paying subscribers.
  3. Subscriber retention: the share of paying users still subscribed after each renewal period. This is different from app-open retention, and both are worth watching.
  4. Refund and chargeback rate: a cohort can look strong on revenue and still be weak after refunds.
  5. Cumulative revenue per user: total revenue from the cohort divided by its original size, plotted over time. This is the curve you compare against acquisition cost to see payback.
  6. Cost per acquisition by cohort: only useful when you can tie spend to the same cohort definition as revenue.

The payback view is the one founders and UA leads use most. Plot cumulative revenue per user for each cohort and compare it with the cost to acquire that cohort. If newer cohorts pay back slower, you see it months before it shows up in a blended profit number.

Cohort analysis when users buy on the web first

In a web2app funnel, the user sees an ad, goes through a landing page or quiz, gets a personalized plan, hits a paywall, pays on the web, then downloads the app and logs in. This changes cohort analysis in two ways.

First, there is a cohort stage before the install. You can measure quiz starts, quiz completions, paywall views, and payments, all grouped by campaign and creative. That gives you early cohort signals within days, well before an install-based retention curve matures.

Second, purchase data lives in a different place than in-app behavior. The payment happens on the web, so revenue sits in your payment provider and funnel data, while usage sits in your product analytics. To build one cohort table, you need a shared user identifier carried through login, and your subscription tool needs to know about the web purchase. Mismatched IDs are the most common reason web2app funnel cohorts fail to reconcile.

web2wave is a no-code builder for web2app funnels, and it is one way to get the front half of this data. It captures UTM and attribution parameters per user, lets you run A/B experiments on funnel variants, and connects to subscription and analytics tools. The integrations page lists the supported connections, including RevenueCat, Adapty, Qonversion, Superwall, Apphud, Amplitude, PostHog, and Mixpanel. For the funnel itself, see the guide to the quiz builder.

As a public data point, MyMental AI GURU generated over $100K in revenue with web2wave, with about 55% of subscribers remaining active, refunds under 5% of revenue, and paying users from 40+ countries. Metrics like active subscriber share and refund share are the kind of numbers you can only judge properly per cohort, because a single blended figure hides how each month's acquisition performed.

Ways to run cohort analysis: a comparison

There is no single right tool. Teams usually combine two or three. Here is how the common options compare on the factors that matter.

Best use

  • Spreadsheet or SQL warehouse: custom revenue and payback models that join ad spend, payments, and usage
  • Subscription tool (RevenueCat, Adapty, Qonversion, Apphud): subscriber retention, renewals, and revenue by plan
  • Product analytics (Amplitude, Mixpanel, PostHog): behavioral retention and feature-level cohorts
  • Funnel builder analytics (web2wave): pre-install cohorts by campaign, creative, and funnel variant

Setup effort

  • Spreadsheet or SQL warehouse: highest, since you build the data pipeline and the definitions yourself
  • Subscription tool: low, because purchase events come from the SDK and store or web payments
  • Product analytics: medium, because you need a clean event plan and consistent user IDs
  • Funnel builder analytics: low for funnel data, since it is collected inside the funnel

Blind spots

  • Spreadsheet or SQL warehouse: only as accurate as the data you load into it
  • Subscription tool: limited view of what happened before the purchase
  • Product analytics: usually weak on payment and refund detail unless you send it
  • Funnel builder analytics: does not see in-app behavior after login unless you connect it to your analytics tool

Plan availability on web2wave

  • Basic, Starter, Pro: funnels, attribution parameters, and integrations, with A/B testing from Pro
  • Advanced: adds ROAS and LTV analytics, along with AI localization, email sequences, and payment orchestration

A common setup is to let the funnel builder and ad platforms handle everything up to payment, the subscription tool handle renewals and revenue, and a warehouse or spreadsheet join them for payback by cohort.

How to set up cohort analysis step by step

  1. Define one user ID. Decide which identifier follows the user from the funnel through payment, install, and login. Without this, nothing else reconciles.
  2. Pass attribution through the funnel. Store UTM parameters, campaign, ad set, and creative on the user record at the first touch, not only at purchase.
  3. Choose cohort windows. Weekly cohorts for fast-moving paid campaigns, monthly for slower ones. Use the same windows in every report.
  4. Send purchase and renewal events to one place. Your subscription tool is the usual source of truth. Make sure web payments are included, not only store purchases.
  5. Build the three core views. Subscriber retention by acquisition cohort, cumulative revenue per user by cohort, and the same two views split by channel.
  6. Review on a schedule. A weekly look at young cohorts and a monthly look at older ones is enough for most teams.
  7. Tie changes to cohort boundaries. When you ship a new paywall, price, or onboarding flow, note the date, so a shift in later cohorts has an explanation.

If you run funnels in several markets, keep language and country as cohort dimensions. Teams that localize funnels, such as Leya AI, which scaled its funnel roughly tenfold with localization, need per-market cohorts to see which translated versions hold up after the first payment.

Common mistakes to avoid

  • Judging cohorts too early. A cohort that is two weeks old has not shown its renewal behavior. Compare cohorts at equal age only.
  • Mixing payment types. Trial starts, paid starts, and web purchases are separate events. Combining them distorts conversion and retention.
  • Ignoring refunds. Revenue per user should be net of refunds when you compare channels.
  • Reading tiny cohorts as signal. Always show cohort size.
  • Changing definitions midstream. If you redefine what counts as active, rebuild the history, or old and new rows will not be comparable.
  • Optimizing for install retention only. App opens are not revenue. Track paying subscriber retention.

FAQ

What is the difference between cohort analysis and retention analysis?

Retention is a metric. Cohort analysis is the method of grouping users and comparing that metric across groups. You can run cohort analysis on retention, conversion, refunds, or revenue.

How big should a cohort be?

Big enough that one or two users do not move the percentage visibly. There is no universal number. Show cohort size next to every row, and merge weekly cohorts into monthly ones if your volume is low.

Should I use weekly or monthly cohorts?

Use weekly cohorts when you spend on paid traffic and change creatives often, because you need fast feedback. Use monthly cohorts when volume is low or when you are looking at long-term renewals.

How do I run cohort analysis if users pay on the web before installing?

Carry one user ID from the funnel through payment into the app login, store attribution data when the user first enters the funnel, and make sure your subscription tool receives the web purchase. Then you can group by campaign or funnel variant and follow the same users through renewals. For legal and platform considerations when selling outside the app store, see the web2app funnel compliance guide.

Which cohort metric matters most for a subscription app?

Cumulative net revenue per user, compared against acquisition cost for the same cohort. It combines conversion, renewals, and refunds into one payback curve, which is what budget decisions depend on.

Want to see how funnel-level cohorts and attribution look in practice? Book a demo or start free.

Share this Article

Build your web funnel with web2wave

Start using web funnels: Launch faster, convert better, and keep more revenue with no app store limitations