You guys liked the numbers I shared in the last Idea Byte around weekly paid plans dominating annual. So, here’s another one from RevenueCat’s latest data deep dive…

Once an annual subscriber churns, only ~5% ever come back. Ever.

Not 5% per campaign. 5% total, in the 12 months after they leave. It ranges from 3% to 8% depending on category, and interestingly, the size of the discount you offer doesn't matter at all.

So, the winback playbook so ofen copied from B2B SaaS of “wait for the subscription to lapse, then fire off a 40%-off re-engagement sequence” is mostly writing emails to nobody.

But emails to nobody aren't the whole story obviously. The same dataset I’ve been combing through in pursuit of providing CPR to my own paid app, shows exactly when subscribers decide to leave, and it's much, much earlier than when they actually have their plan expire. That gap is the real winback window, and almost nobody is running plays inside it.

Because most people aren’t worried about winning back their current customers. But guys, we should be.

One scope note before we go: this edition is about subscription apps, and the sharpest data is on popular annual plans. If you run weekly-first pricing (covered last week), your churn shape is different, but the four winback moments below still apply.

Quick note: RevenueCat is a running partner of The Diff— fitting, it's actually their monster dataset that I was reviewing for today’s post, and they have an epic winback product that makes this super actionable.

A few friends who write big newsletters on Substack asked me this, and a few of you asked me this. So, why did I leave a platform that was bringing me growth in the thousands a month (via recommendations) to start from 0 growth again here? And why would I do it again?

Substack is great, no shade. They did great things for me. But for what I want, which is to grow a brand and build a business, it’s not the best place. Substack is betting on being a social network, beehiiv is betting on being the all-in-one system to do everything your newsletter needs. Which is a creator-first stance that I wanted…

If you’ve ever considered starting a newsletter, I’d start here. And if you’re on the fence, reply to this email and I’m honestly happy to chat through it and help.

🥷 Stop running winbacks on people who already left. Run them on people customers who are still with you.

The mechanism that makes this work hides inside a definitional gap that most of us just never instrument:

For an annual subscriber, cancelling and churning are two fundamentally different events, often months apart. A user who hits "cancel auto-renew" in week 2 keeps full access until month 12. They're not gone. They're a current customer who has told you, explicitly, that you're currently losing them—with up to 11 months left for you to change their mind.

That's the real winback window.

Post-expiry, you're competing with every other app for a re-install, and the data caps you at ~5%. Not a good theater to try resurrect people from.

But pre-expiry, you're an app already on their home screen with a flipped toggle (unless they delete the app but they usually don’t).

One of these is a growth channel. And you know which.

What the data says: 35% of all annual-plan cancellations happen in the first month. And I’ve been that stat..I’ve downloaded apps on annual and hit cancel early just to be done with it.

Cancellations generally then flatten to 3–10% per month through the middle of the year, before spiking again in Month 12 when renewal reminders land. And the trend is getting worse: ~72% of annual subscribers now cancel within their first year, up from ~56% the year before.

Check this chart out showing annual subscriber cancellations in Y1

Ok, so here’s what I take from all this red at the bottom: people don’t "forget" they subscribed. Renewal reminders, subscription-management screens, and a decade of subscription fatigue killed the lazy annuity. You don't have 11 months to earn the renewal. You have about 30 days to earn the toggle staying on.

The battle for Year 2 actually starts in Week 1. Which intuitively makes sense as product people, yet it’s always new customers we peruse vs the winning back of existing ones.

This is the cleanest way I've found to think about it. Four states, four different plays:

You A/B test your install paywall obsessively and then send your cancel flow to Apple's default settings sheet. But the cancel moment is the highest-intent screen in your entire app: the user is guaranteed to be making a purchase decision in the next 10 seconds.

The save offers that work are matched to the reason:

An exit survey is more so about taking input and using it as the targeting layer vs just slapping an exit survey because your boss wants to know why people cancel.

One guardrail from the data though. In their Part 1 report, Blinkist shared that targeting users who cancel within minutes of subscribing barely converted on any offer. Some cancels are accidental sign-ups or instant regret. No flow saves those, so don't judge your save rate against them.

If you're building anything AI-powered, this is your problem on hard mode: AI apps generate 41% more revenue per payer but churn ~36% faster than non-AI apps. Users will absolutely pay to try the wrapper the magic. Keeping them is the whole game. Acquisition got harder and retention got harder at the same time, which means the winback window isn't a nice-to-have on your roadmap—it is the roadmap.

Here’s the retained subscribers view after 12 months, by AI vs. non-AI

And per Part 1 of the report, the median app MRR growth is 5.3% while the top 10% grew 306%. The gap between apps that operate retention and apps that hope for it is big.

…but beehiiv's Summer Release Event on July 16 is worth clearing your calendar for. Apparently they are dropping some insane new updates for newsletter creators, and for a company that ships at crazy velocity, I’m excited.

Five companies, five unique winback moments. Consumer & B2B. None of them are the old "please come back with this discount” email after expiry. Ramp is my favorite one here

Duolingo's growth team famously rebuilt around retention mechanics— streaks, streak repair, leaderboards, and aggressively-optimized notifications—and tracked "resurrected users" as a first-class node in their growth model. The result, documented in former CPO Jorge Mazal's breakdown, was years of compounding DAU growth without a winback email in sight. The user who misses three days gets pulled back by the product (your streak is about to die!) long before they become churned.

Takeaway: The cheapest winback is the one your product runs automatically, before the cancel decision forms.

Ramp's entire product is a running winback campaign.

Their brand promise is "save an average of 5%," and then the product proves it continuously: a real-time savings dashboard tallying exactly what you've saved, alerts when it spots duplicate subscriptions or wasted spend, even line items like "our policy agent blocked X out-of-policy transactions, saving you $Y." By renewal time, the question "is this worth it?" has been answered weekly, in dollars, for 12 months. There's nothing for a CFO to deliberate.

Grammarly and Wispr Flow run the consumer-grade version of the same play: that weekly insights email "you were more productive than 94% of users, here's your word count, you saved this much time typing, etc" exists purely to make the value visible before you ever ask yourself why you're paying.

The mechanism they nail: users don't always churn because the value stopped, they churn because they stopped noticing it. A "you've done X / you've saved Y" surface is the cheapest churn prevention you can ship, because it fights the forgetting, not the leaving.

Takeaway: Don't make users do the ROI math at renewal time. Show the receipt every week so renewal isn't a decision.

Try to cancel Audible and you won't hit a "are you sure?" guilt screen. You get offered a pause (keep your account, billing stops for a few months) and a reminder that your unused credits disappear if you leave but stay if you don't. They're matching the most common cancel reason ("I'm not using it right now") with an option that fits it exactly.

Takeaway: Despite the data saying only 5% ever come back, at the moment of decision, people cancelling are usually feeling "not right now," vs "never again." Sell the pause.

The NYT's cancellation flow is the canonical save-offer machine: try to leave and you're met with a steeply discounted rate to stay, an experience documented by approximately everyone on the internet who has ever tried to cancel. Whatever you think of the friction, the underlying read is that the cancel moment is a price-sensitivity reveal, and a subscriber retained at a lower price massively out-earns the ~5% chance of reacquiring them later. And people are more likely to take the discount at the moment of cancel, than to come and resubscribe again with one later. It’s the endowment effect.

Takeaway: A discounted "stay" beats a discounted "come back" — same margin hit, wildly different conversion

In 2020, Netflix started emailing subscribers who hadn't watched anything in a year and asking if they wanted to keep paying, and auto-cancelling anyone who didn't respond. They voluntarily gave up zombie revenue (under 0.5% of members) to protect trust.

The play: a user who leaves feeling respected is a user you can actually win back when the next must-watch show drops. A user who feels trapped leaves a one-star review on the way out.

Takeaway: Retention dark patterns spend the exact trust a future winback needs.

So, where do we go from here?

Don't respond to the 5% wall by cranking the discount. The data says offer size doesn't lift post-churn reactivation, but aggressive discounts do leak to active subscribers, who learn that cancelling is the loyalty program. You'll convert almost none of the lost souls, but train your healthiest cohort to churn on schedule. The fix for churn is earlier not cheaper.

Ok, that’s the idea and the data backs it. And RevenueCat really is the natural place to run this thing: their pitch is to "catch users before they cancel, and turn churn into a growth lever", which is the backbone of everything you just read. The four winback moments map straight onto their stack:

Start with the first one.

That whole loop is a few hours in Customer Center, and it's free to start.

The latest update of my app—new features, new design, new growth ideas implement from recent editions—just hit the app store. I also built a new market site for it and I’m super happy. I’d love for you to check it out: www.getlittlemoments.com

Why am I building this?

Most people can’t remember what happened last Tuesday. Life is made if little moments that slip away and days blur together. And journaling often feels like pressure, a chore, and those blank pages or open prompts are off putting.

Yet, I take photos all the time and my camera roll holds stuff that just ends up being forgotten about.

Little Moments is about one moment a day. Pick a photo or video. Say or type a few words to caption it and add some detail. 60 seconds max per moment. And all saved moments go to your private moment Capsule. We turn those into movies of your life, generate thoughtful weekly chapters for you, and even point out connections and patterns across your moments like a friend following along.

The goal is noticing the little things. The big, bad, or boring, that make up real life. Noticing what matters (a smile from your partner, a word from your kid, a beer with a friend), and ending up with a real keepsake your future you will be grateful for having.

The core mechanic is free, and I’d love for you to try it out. See how it feels after 1 week. If you use it for 7 days, don’t like it, and tell me why—I’ll give you a free 30m call to chat about whatever you want. Product, growth, etc.

The new feature I’m most happy with in terms of UX—Magic Fill. Designed to help you “backfill” a bunch of moments really quickly and caption them in one voice flow.