Win them back on the slope, not at the bottom

A lapsed user is not a coin flip. When someone stops opening your product, it’s tempting to picture a binary: they’re gone, or they’re not. But the thing that actually decides whether they come back is decaying in the background on a shape we’ve known for over a century — and if you know the shape, you can aim your win-back at the one window where it works.

In 1885 Hermann Ebbinghaus did something slightly deranged and very useful: he memorized lists of nonsense syllables, then tested himself as time passed, on himself, alone, for years. What he found became the forgetting curve — memory drops steeply at first, then the fall slows and flattens into a long, shallow tail. Most of what you lose, you lose fast. What survives the first drop tends to stick around.

He also found something builders should tattoo somewhere: savings. Relearning a list he’d “forgotten” was far faster than learning it the first time. The trace wasn’t gone; it was below the waterline, and it discounted the cost of getting it back.

Two cautions before we build on this, because the wedge here is accuracy. Ebbinghaus tested one person on deliberately meaningless material, which is close to the worst case for retention. A careful 2015 replication (Murre and Dros) broadly reproduced the shape, but meaningful, emotionally weighted, real-world memories don’t decay on the same schedule — some barely decay at all. Treat the forgetting curve as a shape and an intuition, not a formula you can plug a user ID into.

Here’s the translation. The “memory” that decides re-engagement isn’t a fact — it’s the felt reason to come back: the specific value your product delivered, still warm enough to pull them. That feeling decays too, steep-then-flat. And most re-engagement is timed against the wrong clock.

The common approach is a fixed cadence — nudge on day 3, day 7, day 14 — or the calendar (the fresh-start Monday email). Fixed cadences ignore the individual’s decay entirely. Fire too early and you’re pestering someone whose reason to return is still vivid; you spend a notification to annoy a warm user. Fire too late and you’ve let the value memory flatten to near-zero — now you’re not re-engaging anyone, you’re re-acquiring a cold stranger, at cold-acquisition cost.

The lever is to aim for the slope, not the bottom. There’s a window on the descending part of the curve where a nudge lands well: enough has faded that a reminder carries a little novelty and genuine usefulness, but not so much that the person has rebuilt their life around your absence. That window is a property of the user and the value, not of your email calendar — which is why the best win-back triggers are behavioral (this user’s own decay pattern) rather than a global day-number.

Then spend your savings. A returning user who is dumped back into onboarding-from-zero is being charged the full first-learning cost a second time — the most expensive possible welcome. Savings is your built-in discount: land them exactly where they left, state restored, the half-finished thing still half-finished, the setup they did still done. Relearning-is-cheaper only pays off if your product lets them relearn instead of forcing them to start over.

This is a different lever from three neighbors it’s easy to confuse it with. The fresh-start effect times a nudge to a calendar landmark (a Monday, a birthday, January 1) — it’s about the when in the user’s calendar. The spacing effect is about how to teach across sessions so a lesson sticks. Cue-based reminders (prospective memory) hang a nudge on an event in the user’s world. This piece is narrower and more specific: read the shape and timing of one user’s own forgetting of your value, and re-teach on the slope.

The science, to look up: Hermann Ebbinghaus, Über das Gedächtnis (1885) — the forgetting curve and the savings method of measuring relearning; Murre & Dros (2015), a modern replication of Ebbinghaus that broadly reproduced the curve’s shape. Hedge: single-subject, nonsense-syllable material; meaningful real-world memories decay differently — the curve is a shape and an intuition, not a formula.