The Button That Was Costing $300 Million a Year

• A required Login/Register gate before checkout blocked sales at industrial scale.
• Replacing “Register” with “Continue” unlocked 45% more purchasing customers in Spool’s account.
• The flow “worked,” which hid the abandonment cost — labels are product behaviour.
• UI aberration: friction that departs from the expected purchase path for real users.

The form looked harmless.

Two fields: email and password. Two buttons: Login and Register. One link: Forgot Password.

It sat between a full shopping cart and the act of paying.

Designers at a major e-commerce retailer (Spool describes a roughly $25 billion business, under NDA) thought the pattern was normal. Returning customers would sail through. New customers would register once and thank the site later.

They were wrong about both groups.

And for a long time, nobody in the building knew how wrong.

This is Jared M. Spool's famous case study, published via User Interface Engineering / Center Centre. Luke Wroblewski asked him for an example of a form change that moved serious money. Spool's answer was not a prettier checkout. It was a single gate that quietly blocked purchases to the tune of hundreds of millions of dollars a year.

This is not a tale of outages or 500 errors.

The checkout worked.

That is what made the damage invisible.

Shoppers filled their carts and hit Checkout.

Then the wall appeared: log in or register before you can pay.

Internally, the team framed the form as a gift to repeat buyers (a faster path next time). First-timers, they assumed, would shrug at a little extra effort.

Everybody wins, right?

First-time buyers bristled.

In usability tests Spool's team ran with real shopping lists and real money, people hated the demand. One shopper's line became the moral of the story: they were not there to enter a relationship. They just wanted to buy something.

People who clicked Register did it with dread. They imagined marketing spam or worse, without even knowing what registration required. (In reality the site only asked for what a purchase needs: name, addresses, payment.)

Returning customers were not much better off.

Many could not remember which email they had used years earlier, or which password. Guesses failed. Forgot Password only helped if you remembered the right inbox. That is a cruel loop when you have several addresses and years of half-remembered accounts.

Spool's team later analysed the retailer's database. 45% of customers had multiple registrations. Some had as many as ten.

Password-reset requests ran on the order of 160,000 per day in their analysis. 75% of those people never completed the purchase after asking for a reset.

The form meant to speed repeat business was blocking sales at industrial scale.

Even customers who eventually logged in often still had to update stale addresses or cards. The supposed shortcut was not always shorter.

The team had no clue there was a problem, because nothing was "broken" in the engineering sense.

Metrics that only watch successful checkouts miss the people who never become a successful checkout.

The redesign was almost insultingly simple.

Remove Register. Put Continue in its place, with a short message Spool records in full: you do not need to create an account to make purchases on the site; click Continue to proceed to checkout; to make future purchases faster, you can create an account during checkout.

Optional account later. Purchase now.

Results, as Spool published them: purchasing customers up 45%. About $15 million in extra revenue in the first month. Roughly $300 million extra in the first year.

The CEO of the $25 billion retailer left Spool a voicemail that needed no complexity: "Spool! You're the man!"

All they had changed was a button, and the story that button told about what the site demanded before it would take money.

It is tempting to file this under "clever microcopy."

That undersells what went wrong.

The bug was organisational blindness to behaviour that does not throw exceptions. Forced registration was a policy choice wearing a UI. It survived because people inside believed registration was fine, because the happy path still converted, and because lost sales do not page anyone at 3 a.m.

Silent abandonment looks like Tuesday.

Spool writes that the designers of the site had no clue there was even a problem. That may be the scariest sentence in the whole case study.

Watching real people try to buy (with their own confusion, their own email habits, their own distrust of "Register") showed what dashboards flattened.

The flow worked for the minority who remembered credentials and accepted the relationship. It punished everyone else, including loyal customers who had bought before and simply could not get back in.

Duplicate accounts were not a data-hygiene curiosity. They were evidence the product had trained people to fail at login and invent a new identity instead.

Labels are product behaviour.

"Register" is not a synonym for "continue." It is a commitment. "Login" is not a courtesy for returners when returners cannot remember which self they were last time. "Forgot Password" is not a safety net when three-quarters of the people who use it never finish buying.

And the most expensive defects are often the ones that never look like defects.

No stack trace. No red build. Just a button that quietly taxed hundreds of millions of dollars a year until someone sat beside a shopper and listened.

Spool never named the retailer. Neither should anyone else pretend to. "It was Amazon / Best Buy" retellings are noise. The published record is deliberately anonymous. The NDA is part of the story's discipline.

What travels is the pattern.

A UI that "works." A business that assumes the friction is fine. Real human behaviour writing off a fortune in silence, until someone measured what the screen was costing, not merely whether the page returned 200.

https://articles.centercentre.com/three_hund_million_button/

https://medium.com/uie-brain-sparks/the-300-million-button-76b566ae5f73