The subscription lifecycle, from trial to win-back
Growth without waiting on devDeep dive9 min readMiniwall team
On this page
A subscription isn't one decision, it's about eight, spread over weeks or years, and each one has a different person making it. The visitor weighing a price is not the customer deciding whether to renew, and neither is the person whose card just failed. If a small team has to pick, we'd work backwards: fix the failed-payment screen first, add a cancel survey second, and leave the win-back offer for last. The rest of this post is the map behind that opinion, and where it breaks.
Nothing here is measured. It's a way of organising the work, written by people who build screens, and you should expect your own order to differ.
The map
| Stage | What is being decided | A screen that can help | One metric |
|---|---|---|---|
| 1. First paywall view | "Is this worth money?" | A plan screen after a useful moment | Views that end in a tap on a plan, per view |
| 2. Trial | "Will I get value before it ends?" | Usually nothing | Trials where the first key action happens |
| 3. Trial end | "Do I keep paying?" | A reminder that doesn't block the page | Trials that become paid, by cohort |
| 4. First value | "Was this the right choice?" | A short welcome that points to one action | Paid users who finish that action in week one |
| 5. Renewal | "Do I still want this?" | A heads-up before an annual charge | Renewals that are followed by a cancel or a dispute |
| 6. Payment trouble | "Is my card the problem?" | A calm "update your card" notice | Failed payments that are fixed before access ends |
| 7. Cancel | "Why am I leaving?" | One question, then the cancel button | Cancel screens opened, answered, and completed |
| 8. Return | "Is it worth another try?" | A welcome back, shown to the right people | Returning visitors who resubscribe, against a holdout |
Swipe sideways to see every column.
Stage 7 and stage 8 are the same person at two moments, and so are 3 and 5. We'll take them in order.
1. The first paywall view
The person is deciding whether the product is worth paying for before they know what it does for them. The screen that helps is the one that shows up after something useful has happened, or when they hit a limit they can see. We wrote about when to ask in when to show an upgrade prompt, so we'll skip the reasoning.
What shouldn't be there: a price on the sign-up screen for a product that has a free tier, and a prompt that interrupts someone halfway through a task.
The number to watch is how many views lead to a tap on a plan. Read it next to how many people saw the screen at all, because a screen nobody reaches looks like a screen nobody likes.
2. The trial
In the middle of a trial the person is trying to find out whether the product works for them. The best screen here is no screen. Every popup in week one is a tax on the only thing the trial is for.
The exception is something that helps them get to the first useful result: a short onboarding flow that asks a question and starts them in the right place. That is a different job from selling.
What shouldn't be there: a second plan screen, a countdown, a "your trial is ending!" banner on day two.
Watch whether the key action happens during the trial: the first export, the first project, whatever your product is for. Trials where it never happens are mostly lost already, and a reminder at the end won't change that.
3. The end of the trial
Now the person is choosing whether to pay. They need to know when it ends, what they keep, and what it costs. A small card or a bottom sheet that leaves the page usable does that without getting in the way; our trial-ending reminder is built that way.
Two rules we'd hold to. State the date and the price plainly. And let them close it.
What shouldn't be there: a full-screen block with no way to dismiss it, or wording that hides that a charge follows.
One catch. The screen only knows what your site tells it. Days left in the trial are a fact in your own system, which you pass in as an attribute that is checked in the browser and never sent to Miniwall [2]. If your site doesn't pass it, the reminder can't be targeted.
The metric is the share of trials that become paid, split by when people started. Count the people who saw the reminder separately from the people who didn't, or you can't tell whether it helped.
4. First value after paying
This is the stage most teams skip. The person has paid, and the usual response is silence. A short screen that confirms the purchase and points to the one thing worth doing first is cheap, and it's a good place to say what just unlocked.
Your code decides when the purchase has succeeded, then opens the screen. Miniwall doesn't confirm payment [1]. We'd unlock the account from your payment provider's server-side notification and not from the browser alone, which is what the docs recommend as well.
What shouldn't be there: a second sales pitch, an upsell to a higher plan on the confirmation, a request for a review.
Watch how many new paying users do the key action in the first week. It's the earliest sign that someone is likely to stay.
5. Renewal
Most people don't think about a subscription between payments. For monthly plans a charge is a minor event. For an annual plan it's a surprise to anyone who forgot, and I'd expect a surprise charge to end in a dispute or a refund request more often than a reminder would. That's a guess, not a statistic.
A heads-up a few days before an annual renewal says what will be charged and offers three paths: keep, change plan, cancel. The renewal reminder does that. The "keep" button just closes it.
Your backend has to know the renewal date. Miniwall doesn't. And it doesn't send email, so if some of your annual customers rarely open the product, a screen won't reach them. Keep an email for those.
What shouldn't be there: a reminder that makes cancelling harder to find than the reminder itself.
The metric is what happens after the notice: how many renew, how many cancel, how many dispute. A drop in disputes is the thing you'd hope for, but with small numbers you'll see only a few, so treat it as a hint.
6. Payment trouble
A failed payment is rarely a decision. The person didn't choose to leave; a card expired. Treating them like a cancelling customer, with a survey or a discount, is the wrong move.
What they need is one calm message: the payment didn't go through, here's how long access lasts, here's the button that takes you to update the card. The failed payment notice is that screen. Your code opens it when your billing data says the payment failed, and the button sends the person to your own billing page. Miniwall can't see payment status and doesn't handle cards [1].
Timing matters more than wording. Show it when the person is next in the product, not on every page, and cap it. Don't block the product on the first failure. Whether and when to retry the charge is a matter for your payment provider's settings; read their documentation for how it behaves, because we haven't covered it here.
What shouldn't be there: a cancel survey, a win-back discount, or alarming language about "losing everything".
Watch how many failed payments are fixed before access changes. If you can only get one number from this stage, that's it.
7. Cancel
Here the person has decided, so the screen has about five seconds. One question about why, a few fixed answers, an optional box. We wrote the long version in cancel survey questions, and the cancel survey template is the starting point.
Two things never change. The cancel button stays visible and honest, and a broken survey must not become a broken cancellation. Your code does the cancelling, not the screen.
What shouldn't be there: three "are you sure?" steps, or a discount for every answer. Match the reply to the reason, and some reasons deserve only a thank-you.
Watch three numbers together: how many opened the cancel screen, how many answered, and how many completed the cancellation. A survey that raises the first number but not the third may be a trap, and it should make you uncomfortable.
8. Return
Someone who cancelled and came back is telling you something. The question is whether an offer helps, or whether it just gives money to people who would have returned anyway. A win-back offer can be shown only to visitors your site marks as lapsed.
The way to find out is a holdout: a share of matching visitors who see nothing, compared with those who see the offer. Decide what counts as success first, and read the result the way we describe in reading experiment results without fooling yourself. Revenue per visitor will tell you more than conversion alone, as paywall test metrics shows, and if your traffic is thin, A/B testing with little traffic is the sober read.
What shouldn't be there: an offer that shows to current subscribers, or one that appears on every visit.
Where to start with a small team
This is the part that's opinion all the way down. Our order:
- Payment trouble. The people are already willing to pay, and the fix is one screen. It has the clearest "what did I save" of any stage.
- Cancel. It costs one screen and teaches you what the other stages should say.
- First value after paying. Cheap, and it improves what the next stages measure.
- Trial end, if you have trials. Renewal, if you sell annual plans.
- Return, last. It needs a lapsed audience, enough traffic to hold some out, and a reason to believe an offer is the answer.
We put the first paywall view outside the list on purpose. It's where everyone starts, and it's where the most advice already exists.
If you can't tell where to begin, look at where people disappear: the share who start a trial and never pay, the share who fail a payment and vanish, the share who cancel without a word. Fix the biggest hole, not the earliest stage.
When this map doesn't fit
A one-time purchase has stages 1 and 4 and not much else. Skip the rest.
If you sell to companies, the person who signs isn't the person who uses it, and renewal is a conversation with procurement, not a screen. Email and account managers do the work of stages 3 to 8.
If cancelling or paying happens on a page you don't control, such as a store's own sheet, you can't show a screen at that point. You can still show one when the person next opens your product.
And if you have a few dozen subscribers, write to them. You'll learn more from three replies than from a funnel.
