Duolingo sorts every person who ever opened its app into a group, based on how long ago they last used it. One group is called “resurrected users”: people who come back after 30 days or more away.
When the app’s growth slowed, its data team built a simulation to find out which group was worth working on. In the model, each group came back 2% more often, and the result showed how many extra daily users that would add. The findings hold three lessons for any affiliate trying to keep their traffic alive.
1. 
Feed the living first
The group with the biggest effect was the regulars: people who used the app today and on other days that week. A 2% gain there had about five times the effect of the next-best group.
The reason is compounding. A regular who stays active returns to the same group the next day, so each small gain repeats and builds on itself, while a resurrected user is a single comeback that only keeps paying if they stay.
For an affiliate paid on ongoing activity, that’s the difference between a temporary spike and a stable base. The key metric is how many referred users are still active a month later.
2. 🧟 Wake the sleepers in stages
Duolingo separates its quiet users by how long they have been away: up to 6 days, 7 to 29 days, or 30 days and more. Each group has its own chance of coming back, so Duolingo tracks each one separately.
A single “inactive” segment hides that difference. Someone who went quiet two weeks ago and someone who left six months ago are different problems, and a win-back message averaged across both can look like a failure when it worked on one of them.
The same simulation also sets a limit. Regulars mattered six times more for daily users than for monthly ones, because monthly totals depend more on new and returning people.
3. 🎃 Haunt them gently
Duolingo’s former chief product officer tells a story that Groupon’s CEO once told Duolingo’s CEO. Groupon sent one email a day. A test with a second daily email improved the numbers it tracked, so Groupon kept adding, up to five a day. Then the channel lost most of its power. People who had opted out stayed out, even after the test ended.
Duolingo took one rule from it: protect the channel. Its team was free to change the timing, wording and pictures of its notifications, but sending more of them needed a strong reason and the CEO’s approval.
A subscriber list, push audience or Telegram group can wear out the same way, and a test that lifts this week’s clicks may be the one that shrinks the channel.
All three findings point the same way for an affiliate: the first visit is only the start of what a piece of traffic is worth. What it earns afterwards depends on whether people keep coming back, and that’s decided as much by the brand behind the link as by the content in front of it.
That’s the thinking behind Altair Affiliates: we focus on timeless brands and stay beside you every day of the year, peak season included. If you’d like your traffic to land on brands people return to, Altair Affiliates is your answer.
