Somewhere in your Stripe dashboard is a dispute marked fraudulent, filed by a customer who wasn't lying. You pull up the record and everything checks out: their real email, their real card, two logins during the trial, a session that lasted twenty minutes. And yet the bank's paperwork says they never authorized the charge. Reading it feels like being accused of theft by someone you can see on your own security camera. Here's the uncomfortable part: in their memory, the accusation is true. They don't remember agreeing to be charged — because the part of memory that was supposed to hold that agreement is the least reliable one humans have.

Free trial chargebacks are among the most preventable disputes in all of commerce. The prevention is a single email. And most merchants are quietly afraid to send it.

A dispute that was filed the moment they signed up

Walk through the timeline from the customer's side. The signup took forty seconds. They wanted the thing behind the paywall — the export, the report, the first episode — and the trial was the tax they paid to get it. In that moment they made a silent deal with themselves: I'll cancel before it charges me. That's not consent to pay. That's consent deferred, handed off to a future self.

Fourteen or thirty days later, a charge posts under a billing descriptor they half-recognize, for an amount they never consciously registered, on a date they never wrote down. Their banking app puts a Dispute this charge button one tap away. "I don't recognize this" is the option closest to what they're feeling, so that's the one they press. This is why so many trial-conversion disputes arrive coded as fraud rather than as a cancellation issue: the customer isn't describing what happened, they're describing what they remember. And what they remember is nothing.

The memory failure with a name

Psychologists distinguish between retrospective memory — recalling things that already happened — and prospective memory: remembering to do something at a future moment. Remembering your childhood address is retrospective. Remembering to take the bread out of the oven is prospective. Prospective memory is the flakier of the two, and it fails in a predictable pattern.

Research on prospective memory draws a further line between event-based intentions (do X when Y happens — "give Sarah the book when I see her") and time-based intentions (do X by a certain date). Time-based intentions fail more often, for a simple mechanical reason: there's no cue. Nothing in the environment fires to remind you. Your customer's plan — "cancel before the 14th" — is a time-based prospective memory task with zero external triggers, competing against every other demand in their life for two straight weeks. It was always going to lose.

Then the charge lands, and something worse than forgetting happens. The memory of consenting has decayed, but the sting of losing money is immediate and vivid. The honest reconstruction their brain produces is I never agreed to this — which, to them, feels indistinguishable from fraud. Your dispute problem, in other words, is not a dishonesty problem. It's a cue problem. And cue problems can be engineered away.

The card networks already wrote your playbook

This pattern got bad enough, industry-wide, that Visa rewrote its rules for trial and introductory offers in 2020. Merchants running free trials that roll into paid billing are expected to capture the cardholder's express consent at enrollment — with the amount, the billing date, and the terms stated plainly, not buried — to send an electronic reminder before the first real charge, and to offer a simple online way to cancel, no phone call required. Mastercard imposed similar requirements on trial merchants even earlier.

It's tempting to read these as compliance hoops. Read them again as cognitive engineering: consent stated clearly enough to encode a real memory, a reminder that converts a doomed time-based intention into a cued, event-based one, and a cancellation path easy enough that the cued customer actually acts. The networks didn't write a bureaucratic checklist. They wrote a treatment plan for prospective memory failure — and every step of it produces a record you can later hand to a dispute reviewer.

The email you're afraid to send

Here's where most merchants flinch. A reminder email three days before the trial converts feels like inviting cancellations — deliberately waking up customers who would otherwise slide silently into paying. Some of them will cancel. That's true, and it's worth saying out loud.

But look at what that revenue actually was. A customer who only pays because they forgot to leave isn't a subscriber; they're a refund request with a delay on it. When they notice — and they always notice, on the first statement or the third — they exit through the most expensive door available: a support ticket demanding a refund, or a dispute that costs you the month's revenue, the dispute fee, an evening of evidence-gathering, and a permanent tick on the dispute rate your processor watches. The reminder email doesn't destroy revenue. It filters out revenue that was going to leave anyway, and lets it leave cheaply.

And the email is doing a second job the whole time. Every send is a logged, timestamped record that this cardholder was told the exact amount, the exact date, and given a working cancel link — and chose not to click it. The thing that prevents most of your trial disputes is the same thing that wins the rest.

What wins the dispute when it comes anyway

Some will still come. When one does, the case you're making is simple: this wasn't unauthorized — they enrolled, they were reminded, they used it, and they didn't cancel. Four exhibits carry that argument:

  1. The consent snapshot. What the signup screen actually said, with the price and billing date visible near the button, plus the timestamped acceptance record — date, time, IP — from your checkout.
  2. The reminder log. The delivery record from your email provider showing the pre-charge notice went to their address, when, and what it said.
  3. Usage during the trial. Logins, downloads, projects created — activity tied to the same email and account, proving a real person was inside the product.
  4. The open door. Evidence that cancellation was one click away in every email you sent, and the account shows no cancellation attempt.

That's the whole story a reviewer needs. Tight, dated, and hard to argue with.

Your next moves

  • Wire up the pre-charge reminder today. Stripe fires a customer.subscription.trial_will_end webhook three days before a trial converts — connect it to an email stating the exact amount, the exact charge date, and a one-click cancel link. (Stripe Billing can also send trial-ending reminders natively; turning that on takes minutes.)
  • Put the price next to the button. Change your CTA from "Start free trial" to "Start free trial — then $29/mo starting July 27." Consent stated at the moment of action encodes a memory; consent in a terms page does not.
  • Remove the login wall from cancellation. Put a Stripe customer portal link in every billing email so cancel is one click, not a password-reset spiral that curdles into a dispute.
  • Check your billing descriptor. If the name on the customer's statement doesn't match the name on your website, fix it in your Stripe settings before the next conversion cycle.
  • Archive your consent artifacts. Screenshot the full signup flow — pricing visible — today, and again after every redesign. When a dispute arrives months later, you need what the customer saw then, not what your site looks like now.

Even with all of this in place, some trial disputes will land — and when one does, you get seven days on Stripe to assemble the consent record, the reminder log, and the usage trail into a response a bank reviewer will accept. That's exactly the window Argeback exists for: it ingests your disputes the moment they're filed, drafts an evidence-backed response from your records, and files it before the deadline — from your phone, before the forgetting problem becomes yours too. See how it works at argeback.lumenlabs.works.