Buying became faster than deciding

What ecommerce friction actually means

Friction, in ecommerce, is everything standing between an intention and a completed purchase: steps, form fields, information you have to supply, accounts you have to create, authentication, navigation, waiting, and any moment that interrupts what you were doing. Frictionless checkout is the design goal of reducing as much of that effort as possible.

Most of that work was straightforwardly good. Retyping a card number on every visit added effort that many buyers were glad to lose. When someone has decided what they want, every extra screen is a tax on their time, and thirty years of removing that tax made the web better to use.

It is worth noticing what the word assumes, though. Friction is measured against an intention that is treated as already settled. The interface is the obstacle; the wanting is a given. That assumption holds for most purchases. The rest of this page is about the ones where it does not.

How execution friction disappeared

The first generation of online checkout was a form. You typed a card number, an expiry date, a billing address and a shipping address, you read them back, and you submitted them. Nothing was kept, so the next purchase started from an empty page.

Storage changed that. Once a retailer could keep an address and a card on file behind an account, a repeat purchase stopped being data entry and became confirmation. Amazon then patented a way to compress the confirmation itself, filing in 1997 for a method of ordering in response to a single action.

Two other paths opened around the same time. Express checkout shortened the flow for customers who had already identified themselves; Barnes & Noble had shipped exactly such a feature, called Express Lane, by 1999. Guest checkout went the other way and removed the account requirement altogether, which mattered because account creation is still one of the reasons shoppers give for abandoning a cart.

Then payment left the retailer entirely. Apple Pay reached US customers on 20 October 2014, authorised with Touch ID on iPhone 6 and, from the same launch, inside apps. The card stopped being something you produced and became something the device already held. Face recognition later replaced the fingerprint, and a payment came to be confirmed the way a phone is unlocked.

What 1-Click actually changed

The patent is worth reading rather than recalling. Amazon filed US Patent 5,960,411 on 12 September 1997; it was granted on 28 September 1999 and expired in September 2017. Claim 1 describes ordering an item “in response to only a single action being performed”, where the request carries an identifier of the purchaser and the server retrieves “additional information previously stored” for that purchaser, so that the item is ordered “without using a shopping cart ordering model”.

Read closely, that is narrower than the legend. It is one-action ordering for a customer who is already identified and whose details are already on file. It presumes the stored-details era rather than creating it, and what it claimed was a particular implementation of the last step, not speed in general.

Its legal history was also less settled than the story suggests. Amazon sued Barnes & Noble over Express Lane on 21 October 1999 and won a preliminary injunction, but the Federal Circuit vacated it on 14 February 2001, holding that Barnes & Noble had raised substantial questions about the patent's validity.

So the tidy version — checkout stayed slow because one company owned the fast way, and became fast when the patent expired — does not hold. Express checkout existed at a competitor within weeks of the patent issuing and kept running after the litigation. What 1-Click changed was narrower and more interesting: it made one particular implementation of single-action ordering a proprietary asset, and everyone else arrived at comparable speed by another route.

When payment became infrastructure

Stored cards, wallets and biometric confirmation are usually listed as three improvements. They are better understood as one move: payment stopped being an event the buyer performs and became a capability the device provides.

Here is our reading of what that changed, offered as interpretation rather than fact. A signature, a typed card number, even a password, are all actions performed only in the context of paying. A fingerprint or a glance is not. It is the same gesture used to open a phone dozens of times a day, and it carries no built-in sense of occasion. The confirmation step survives, but the part of it that used to say you are spending money now has thinned considerably.

What instalment payment changes

Buy now, pay later is often described as removing a step. It does not. The steps were already gone. What instalment framing changes is which number is most prominent while the decision is being made: a £120 purchase can present itself as four payments of £30, and the figure on the button is not the figure owed. That is a claim about salience, not about harm — the total is disclosed, and for a planned purchase the framing is simply a financing choice.

How these products should be classified is genuinely unsettled. In the United States, the Consumer Financial Protection Bureau withdrew its 2024 BNPL interpretive rule on 12 May 2025, as part of a broader withdrawal of guidance documents; that remains the position stated on its BNPL compliance page, last updated 1 July 2025. Treatment differs by jurisdiction and continues to move, so any summary here is a snapshot with a date on it.

Meanwhile, decision pressure evolved too

Everything above answers one question: how easily can someone act? A second discipline grew up alongside it answering a different one: why should they act now? These are not the same problem, and they are rarely owned by the same team.

In our fixed retailer panel, we observe mechanisms including countdown timers, stock messages, shipping deadlines, limited-time offers, flash-sale framing, reserved-cart notices and signals about other shoppers' activity. That is a description of what appears on the pages we watch, on the days we watch them. It is not a measurement of how common these mechanisms are across ecommerce, and we are not claiming any trend from it.

What matters for this argument is only that the techniques exist, are deliberate, and work in the opposite direction to deliberation. A countdown does not make a purchase easier to complete. It makes waiting feel like a cost.

The overshoot

Take the two developments together and a question appears that neither raises on its own.

When an intention is stable — the printer ink ran out, the flight is booked, the size was wrong — fast execution is usually useful, and unnecessary delay adds little. The system was designed for that case and serves it well.

The interesting case is the other one: an intention still being formed, inside an interface that can complete it in one gesture while simultaneously giving reasons to complete it in the next thirty seconds. Nothing there is dishonest. Each part is doing its job. But the time in which a person decides has been compressed from two directions at once, and the combined effect can fall between the metrics each system is optimized for.

We are not claiming that checkout optimisation causes impulse buying. We have not measured that, and we would not assert it from a timeline. The narrower claim is the one worth sitting with: two systems optimised independently can overshoot together, and the thing they overshoot is the interval in which someone works out what they actually want. That is what velocity bias names.

What the abandonment data does and does not show

Baymard Institute aggregates 50 studies published between 2006 and 2025 and puts the average documented cart abandonment rate at 70.22 percent. Among the reasons shoppers give, extra costs such as shipping and fees lead at 40 percent, followed by slow delivery at 20 percent, not trusting the site with card details at 19 percent, a required account at 18 percent, and a checkout that is too long or complicated at 17 percent.

The figure we find most relevant sits slightly to one side: 42 percent of US online shoppers say they have abandoned a cart because they were just browsing or not ready to buy, and Baymard describes that segment as largely unavoidable.

That is a description of ordinary behaviour, not a failure. A large share of carts belong to people who have not finished deciding, and adding an item is part of how they think. It tells us nothing about what those shoppers would have bought under a different checkout, and we will not pretend otherwise. It does suggest the interval between interest and commitment is a real and populated part of shopping, rather than a defect to be engineered away.

Not all friction is the same

The word covers two different things, and most disagreements about it come from mixing them up.

Imposed friction is applied to you by someone else, usually as a side effect of bad design: a form that rejects a valid address, a mandatory account, an unnecessary step, an error at the final screen. It has no defender. It is what the last thirty years were right to attack.

Chosen friction is applied by you, to a category of decisions you selected, in advance. A deliberate wait. A confirmation you asked for. A 24-hour rule kept by hand, or the same idea automated by one of the apps built for it. The difference is not how many seconds it costs; it is who decided that those seconds were worth spending, and whether it can be switched off.

It only pays when it is aimed narrowly. A pre-commit window on a category you named is a different proposition from a delay on everything you buy, and the friction dividend is the test of whether the seconds bought anything. It is also why spending interception happens before a payment: the interval it protects exists only before the money moves.

And the outcomes have to stay open. A pause that is built to produce one answer is not protecting a decision, it is making it. Walking away, keeping it in mind for a week, and buying it right now all have to remain available, or the mechanism is just a different kind of pressure.

Where is friction still useful?

The history of ecommerce is largely a history of making action cheaper, easier and faster. Mostly that was progress, and the parts of it that removed pointless effort need no defence.

But once acting becomes nearly effortless, a different design question appears, and it is not the one the industry has been answering. Not how do we remove the next step, but where, in a system this fast, should a person still have time to decide? That question has no settled answer. It is not obvious who should answer it: the retailer whose conversion rate depends on the reply, the platform that owns the payment gesture, the regulator arriving years later, or the person themselves.

Griotta is one experiment in answering it from the last of those positions — a pause the person installs, aimed at a category they choose, with the outcome left open. It is one answer among several possible ones, and the question is considerably more durable than any particular attempt at it.

Sources

• US Patent 5,960,411 — filing date, grant date, anticipated expiration, and the text of claim 1 quoted above.
• Amazon.com, Inc. v. Barnesandnoble.com, Inc., 239 F.3d 1343 (Fed. Cir. 2001) — the Express Lane feature, the 21 October 1999 suit, and the 14 February 2001 decision vacating the preliminary injunction.
• Apple Newsroom, 16 October 2014 — Apple Pay available in the US from 20 October 2014, using Touch ID, in stores and within apps.
• Baymard Institute — 70.22 percent average documented cart abandonment across 50 studies (2006–2025), the reported reasons, and the 42 percent browsing figure.
• Consumer Financial Protection Bureau — withdrawal of the 2024 BNPL interpretive rule on 12 May 2025; page last updated 1 July 2025.
• Mechanism observations are our own, from a fixed retailer panel captured on a schedule. They describe what we saw on those pages on those days, and support no claim about ecommerce as a whole.