Two channels bring you installs at the same cost per install. On paper they look identical, so you split budget evenly and move on. Three months later one channel's users are still opening the app and the other's have vanished, and you have been funding both as if they were the same. This pattern shows up in almost every retention audit: a blended average that quietly hides a two or three times gap between the best and worst sources.
Retention is usually treated as a product metric. Measured by channel, it becomes one of the sharpest tools a performance marketer has for deciding where the next rupee goes.
Why blended retention hides your real channel quality
A single retention number is an average across every source you run, and averages flatter the weak and punish the strong:
- One organic-heavy cohort at 40 per cent Day 7 and one incentivised cohort at 8 per cent can blend to a comfortable-looking 20 per cent that describes neither.
- You scale the cheap channel because its CPI looks good, then wonder why revenue does not follow the install curve.
- The cost is compounding. Every extra rupee into a low-retention source buys users who churn before they pay back, so the damage grows the more you scale.
The fix is not a new metric. It is the same retention rate, split by the source that delivered each cohort. Our guide on how attribution data powers retention marketing covers the segmentation logic in more depth.
What does retention by channel mean?
Retention by channel is app retention rate segmented by the acquisition source that drove each install cohort, so you can compare how well different campaigns retain users.
- The cohort is still grouped by install date, but each install carries its source: network, campaign, ad set, and creative.
- You then read Day 1, Day 7, and Day 30 retention separately for each source.
- The question shifts from "how sticky is our app?" to "how sticky are the users this campaign sent us?"
How to build a source-level retention view
The minimum dashboard cut you need:
- Group installs into cohorts by install date. Day 0 is the install day.
- Split each cohort by source. Start at network level, then drill into campaign and creative once the pattern is clear.
- Read one retention window at a time. Pick Day 7 first; it is early enough to act on and late enough to separate quality.
- Match the method. Compare classic to classic, or cumulative to cumulative, never a mix. The three methods are defined on the Linkrunner retention documentation.
- Wait for maturity. An install only enters the Day 30 cohort once it is 30 days old, so give recent cohorts time before judging longer windows.
Keep the cohort sizes honest. A creative with 30 installs and 50 per cent Day 7 retention is noise, not a winner. Set a minimum sample before you rank sources.
Reading the patterns: which sources usually retain
Across audits, the ordering is consistent even when the exact numbers are not:
- Organic and high-intent search retain highest. These users came looking for you.
- Broad social and interest targeting sit in the middle, with wide variance by creative.
- Incentivised and rewarded traffic retain lowest. Users installed for the reward, not the app.
Scale is what makes this reliable. A public example: Jumbo Gaming tracks over $1M in revenue across 200+ campaigns on Linkrunner, the kind of campaign density where source-level retention differences are large enough to reallocate real budget against.
Late returns matter too. A user who lapses and comes back is a reattribution, and reattribution volume is a useful engagement signal to read per channel. Some sources send users who open once and vanish; others send users who lapse but reactivate. Segmenting reattributions by source shows you which channels deliver users worth winning back, not just users who install.
Turning retention gaps into budget moves
Retention by channel is only worth measuring if it changes what you fund:
- Rank sources by retention, not just CPI. A channel at Rs3 CPI and strong Day 30 retention usually beats one at Rs1 CPI that churns in a week.
- Pair retention with revenue. Retention tells you who stays; revenue per cohort tells you who pays. Together they give you a payback view.
- Reallocate deliberately. Move budget toward high-retention sources in steps, watching whether retention holds as volume grows. Our marketing budget reallocation framework sets thresholds for scaling and cutting.
How to validate this in your MMP
To run this without stitching spreadsheets:
- Confirm every install carries network, campaign, ad set, and creative attribution.
- Check retention is available at each of those levels, not just app-wide.
- Compare the same window and method across sources before drawing conclusions.
Platforms like Linkrunner break retention down by campaign and creative and expose it through the Campaign Reporting API, so a "which channels retain" question becomes a saved view rather than a monthly export. For a revenue-weighted version of the same cut, see our performance marketer's guide to cohort analysis.
Frequently asked questions
How do I measure retention by acquisition channel? Group installs into cohorts by install date, tag each install with its source, then read Day 1, Day 7, and Day 30 retention separately for each network, campaign, or creative. Keep the window and method consistent across sources.
Why do paid and organic users retain differently? Organic users sought your app out, so intent is higher and retention usually follows. Paid users vary by targeting, and incentivised traffic retains worst because the install was motivated by a reward.
Is a low-CPI channel worth it if retention is poor? Often not. A cheap install that churns before it pays back can cost more than a pricier install that stays. Rank channels on retention and revenue, not CPI alone.
What retention window should I compare channels on? Day 7 is the best starting point: early enough to act on, late enough to separate quality. Confirm each cohort has matured before comparing Day 30.
From average to actionable
A single retention number tells you how the app is doing. Retention by channel tells you what to do next. The moment you split cohorts by source, the two or three times gap between your best and worst channels stops hiding inside an average, and budget decisions get a lot less guesswork.
Start with one window, Day 7, split by network, and rank. If you want that view maintained per campaign and creative without manual exports, book a demo with Linkrunner and check your own source-level retention against these patterns.