Decide what a valuable user actually does
Install volume tells you that acquisition happened. It does not tell you whether the people acquired reached the value your business needs. Define one downstream outcome before ranking channels: activation, first purchase, a paid subscription, or another business-specific event.
Do not invent a universal “quality score.” A learning app, a subscription tool, and a lending app have different paths to value. Document the event, its denominator, the observation window, and why it matters commercially.
Compare users with the same opportunity to convert
- Align the cohorts. Compare acquisition cohorts at the same age, in comparable markets and app versions. AppsFlyer’s cohort documentation explains grouping users by acquisition date. Yesterday’s cohort cannot yet supply a full seven-day result.
- Follow the funnel. Look at install → activation → valuable event. Locate the first meaningful divergence. A high activation rate and a weak purchase rate call for a different investigation than poor onboarding completion.
- Bring cost back in. Evaluate cost per valuable user using matched scope and spend. A better conversion rate can still be too expensive; a low CPI can hide expensive customers.
- Check business outcomes. Consider net revenue, refunds, retention, and margin where available. Do not substitute unobserved lifetime value for early revenue or call missing retention zero.
- Check alternative explanations. Device mix, geography, onboarding changes, attribution settings, and event loss can all change the apparent quality of a source.
The cheaper channel may acquire more expensive customers
Channel A spends $2,000 for 1,000 installs. At the same cohort age, 20 users have purchased. Channel B spends $3,000 for 1,000 installs, with 60 purchasers.
A has the lower CPI: $2 versus $3. But acquisition cost per purchaser is $100 for A and $50 for B. The ranking reverses when the business outcome replaces the install.
That does not establish that B is more profitable over a lifetime. Purchase value, repeat behavior, refunds, and extra available audience remain unknown. It establishes which question to investigate next.
Improve the weakest part of the acquisition chain
If users fail to activate, examine the promise in the ad against the first app experience. If activated users do not purchase, examine price, eligibility, and the offer before blaming the traffic. If the valuable-user cost is favorable, test additional spend rather than assuming the same economics hold at any scale.
Review at a preselected cohort age. Report both customer volume and cost, preserve immature cohorts as incomplete, and keep attributed outcomes separate from overlapping media-reported conversions.
PUT THIS THINKING TO WORK ON YOUR ACCOUNTS
Optimize for the customer you want—not the cheapest install.
Bring the event that defines a valuable user, its expected time to occur, and your acquisition-cost target. Ask ChatGrowing to compare equally mature cohorts and identify where inexpensive traffic stops becoming valuable customers.
The useful output: A source comparison based on the business outcome, with immature or missing outcomes kept explicit.
We are getting cheaper installs, but not more first-time paying users. Compare sources at the same cohort age, confirm the paying-user definition, and separate acquisition volume from post-install conversion. Tell me which source merits another test, what is still uncertain, and when to review it.
Copy the brief, then install or sign in and paste it into your agent. This link opens setup; it does not run an analysis or access your accounts.
Sources & related reading
This is a ChatGrowing editorial framework, not a client case study or a claim of model training. Official references explain measurement behavior; the investigation sequence and scenarios are our synthesis.