Choice architecture: guiding consumer decisions

There is no neutral way to present a choice

Something has to be listed first. Something has to be pre-selected or not. A price has to be shown as a discount or as a figure. Every one of those decisions changes behaviour at the margin, and there is no arrangement that changes nothing.

Richard Thaler and Cass Sunstein named this in 'Nudge' in 2008, and Thaler was awarded the Nobel Memorial Prize in Economic Sciences in 2017 for the broader body of work. Their term choice architecture describes the design of the environment in which a decision is made.

My position follows from the first paragraph. Because there is no neutral option, 'we do not use choice architecture' is not available to you. The only real question is whose interest the arrangement serves, and that question now has a legal answer in the UK as well as an ethical one.

The tools, properly named

The clearest taxonomy comes from Eric Johnson and colleagues in Beyond nudges: tools of a choice architecture, published in Marketing Letters in 2012. They split the toolkit in 2.

Tools that structure the choice task: how many options you show, how they are grouped, what is pre-selected, what order they appear in, and what happens if the person does nothing.

Tools that describe the options: how each one is labelled, which attributes are shown, and how the same fact is framed. Amos Tversky and Daniel Kahneman established in 1981 that logically equivalent descriptions of the same outcome produce different choices, which is why a discount expressed as a percentage and the same discount expressed in pounds do not perform identically.

Both sets are in use on your website right now whether you designed them or not.

What defaults do, and what they do not

Defaults are the strongest tool in the set and the most oversold. Comparing 2 famous cases shows why.

UK pension auto-enrolment worked. The DWP's official statistics on workplace pension participation record a general decline between 2009 and 2012, from 58% of eligible employees to a low of 55%, or 10.7 million people. Auto-enrolment was introduced in 2012. By 2019 participation had reached 88%, or 19.2 million eligible employees.

Organ donation is the example everyone quotes, and it is weaker than its reputation. Eric Johnson and Daniel Goldstein's 2003 paper Do defaults save lives showed very large differences in recorded consent between opt-in and opt-out countries. What followed is less encouraging. Wales moved to a soft opt-out system in December 2015, and a decade later the anticipated increase in consent rates has not materialised, with researchers pointing to the role of families in the final decision and to infrastructure rather than the default itself.

The difference between those cases is the useful part. A default works when the only barrier is effort. It works much less well when there is a real objection behind the inaction, or when somebody else can override the choice later.

Applied to a business, that means a default is legitimate and effective for the delivery option, the plan most customers should be on, or the sensible starting configuration. It will not hold when it commits someone to a payment they have not decided to make, because the objection is real and they will reverse it, usually through their bank.

Choice overload, corrected

The article this replaces stated that too many options paralyse buyers. That is the popular version and it does not survive the evidence.

A meta-analysis of 99 observations involving 7,202 participants found that choice overload is heavily moderated by the complexity of the choice, the difficulty of the task, how certain people are about their preferences and what they are trying to achieve. The mean effect across studies is close to zero, with strong effects appearing under particular conditions rather than as a rule.

The practical version is that cutting options is not automatically an improvement. What reliably helps is giving people a basis for choosing. Six options with a sentence explaining who each suits is an easier decision than 3 with no guidance, which is the same conclusion I reach in designing decision interfaces.

Where the line is, and where UK law now draws it

This is the part the original article handled as an abstract question about autonomy. It is no longer abstract.

The Competition and Markets Authority published research on online choice architecture in April 2022, including an evidence review setting out a taxonomy of 21 practices and the harms attached to them. The unfair commercial practices provisions of the Digital Markets, Competition and Consumers Act 2024 then applied to practices from 6 April 2025, with a schedule of practices considered unfair in all circumstances.

The test I use is whether the technique still works once the buyer can see it. A default that saves effort survives being explained. A pre-ticked box that adds a cost does not.

The technique	The version that helps	The version that harms	Where the line falls
Defaults	Pre-selecting the option most customers should have, and saying so	Pre-selecting a paid add-on the buyer did not ask for	Would you be comfortable labelling the default and why it is the default
Price framing	Showing the total, then explaining what makes it go up or down	Revealing fees only at the last step, which is drip pricing	Whether the buyer knows the full price before they invest effort
Scarcity and urgency	Saying that 3 slots remain because 3 slots remain	A countdown that resets on reload, or permanent 'ending soon'	Whether the statement is true at the moment it is made
Ordering	Leading with what most people need	Burying the cheapest suitable option below the fold	Whether the ranking is explained or disguised
Social proof	Real reviews, shown with their source	Commissioned or fabricated reviews, now explicitly banned	Whether you could show where each review came from
Friction	Making a risky action deliberate	Making cancellation harder than signing up	Whether leaving is as easy as arriving

The last row is the one I would check first in any business. Sign-up and cancellation asymmetry is the most common version of this in small businesses, and it is usually accidental rather than cynical: the sign-up was designed and the cancellation was not.

I am describing the shape of the rules rather than giving legal advice. If you use any of the practices in the middle column, read the CMA guidance directly.

What to do with this

Choice architecture is not a growth tactic to add. It is an audit of decisions you have already made by default.

Take your pricing page and answer 4 questions. What is pre-selected, and would you defend it out loud. Does the buyer see the total before they invest time. If you show scarcity, is it true right now. Can someone leave as easily as they arrived.

Then take one thing you found and fix it. In my experience the pre-selected item and the cancellation route are where the answers are least comfortable, and both are usually the result of nobody deciding rather than somebody deciding badly.