Nudge theory says you can change what people choose by changing how the choice is presented, without banning options or changing prices.
Richard Thaler and Cass Sunstein define a nudge as "any aspect of the choice architecture that alters people's behavior in a predictable way without forbidding any options or significantly changing their economic incentives" (Behavioral Economics Guide).
On a website, the most powerful nudge is the default: the option that is already selected when the page loads.
That power cuts both ways. In our database, preselecting a subscription raised subscription rate by 50% for one store, while a pre-checked subscription cut revenue per visitor by 9.17% for another.
This page covers what the research shows, what real A/B tests show, and how to tell a helpful default from one that backfires.
Choice architecture: every page already has one
Choice architecture is the way options are arranged before someone decides.
There is no neutral version. Every product page has a first variant, every paywall has a plan that sits in the middle, and every cart either has a checkbox ticked or it doesn't.
Thaler and Sunstein add a condition: to count as a nudge, the intervention must be easy and cheap to avoid.
A preselected size that the shopper can change in one tap is a nudge.
A pre-ticked insurance fee buried under the subtotal is closer to a trick, and in some places it is illegal (more on that below).
For CRO work, choice architecture usually shows up in four places:
- Variant selectors: which size, color, or quantity is selected on load. See variant selector tests.
- Purchase type: one-time purchase versus subscribe-and-save. See subscription tests.
- Add-ons: shipping protection, warranties, and gift wrap that are on or off by default. See upsell tests.
- Plans and trials: which plan is highlighted on a pricing page or paywall, and whether a free trial is included automatically.
Defaults sit inside a wider set of ideas from behavioral economics, alongside anchoring, loss aversion, and the endowment effect, all covered in our psychology section.
Default bias: why people stick with what's selected
Default bias is the tendency to accept a preselected option rather than switch. Researchers give three main reasons it happens:
- Effort: changing the option takes a decision and a click. Doing nothing is free.
- Endorsement: people read the default as a recommendation from whoever set it up.
- Endowment: the default feels like the status quo, so leaving it feels like giving something up.
What the research shows
The best-known study is Eric Johnson and Daniel Goldstein's "Do Defaults Save Lives?" in Science (2003).
In their online experiment, 42% agreed to be organ donors when they had to opt in, 82% stayed donors when they had to opt out, and 79% chose to donate when forced to make a choice with no default (AMA Journal of Ethics summary).
The same paper found opt-out European countries had far higher consent rates than opt-in ones (Science).
Brigitte Madrian and Dennis Shea studied a company that switched its 401(k) plan to automatic enrollment.
Nothing else about the plan's economics changed, yet participation rose sharply, and many new hires kept the company's default contribution rate and fund even though few earlier employees had picked that combination (NBER working paper, published in the Quarterly Journal of Economics in 2001).
The authors attribute part of that to employees reading the default as advice.
A 2019 meta-analysis by Jachimowicz, Duncan, Weber, and Johnson pooled 58 default studies with 73,675 participants.
It found an average effect of d = 0.68. Effects varied a lot: most studies found positive effects, several found no significant effect, two found negative effects, and defaults worked better in consumer decisions than environmental ones (Columbia Business School).
Where the evidence is weaker than it looks
Nudging as a whole has a publication-bias problem.
A 2022 PNAS meta-analysis by Mertens and colleagues covered more than 200 studies and reported an average choice-architecture effect of about d = 0.45, with "decision structure" nudges such as defaults doing better than information-based ones (PNAS via PMC).
Maier and colleagues then reanalyzed the same data with a correction for publication bias and concluded there was no evidence for an overall nudging effect (PNAS via PMC).
That doesn't prove defaults never work. It means the average published nudge is inflated, so you should not treat any single research number as the lift your store will get.
Two more caveats matter for CRO. Most default studies measure whether people keep an option, not whether they buy at all.
And the classic studies involve choices people rarely think about, like retirement plans.
A shopper paying with their own card is paying close attention, which is where defaults can feel pushy.
Default effect: what A/B tests show
The default effect is the measurable change in outcomes caused by which option is preselected.
Our database has dozens of tests that changed a default.
They fall into two clear groups: defaults that help the shopper finish a choice they were already making, and defaults that add a cost or commitment the shopper didn't ask for.
Blend CommerceBlend Commerce
Visually
SplitBase
Convertibles
@DaveDiederenDefaults that won: removing a decision
Stone Creek Coffee used to show grind type as an empty placeholder that shoppers had to fill.
Blend Commerce preselected Whole Bean, still changeable, and saw +4.54% conversion rate, +1.63% revenue per visitor, and +25.74% subscription orders per visitor, with AOV down 2.78%.
Revant Optics defaulted every frame to a gray lens.
Visually tested preselecting each frame's best-selling lens color instead and reported +12.2% revenue (vendor case study; sample size not stated).
The default became a recommendation, which matches the endorsement explanation from the research.
Dr. Squatch's soap pages had no quantity field.
SplitBase added one with 2 bars as the default quantity and reported up to +54% revenue per user among soap buyers, a best figure across several iterations.
Subscription defaults: big opt-in gains, mixed revenue
For Jackson's, Blend Commerce preselected the subscription option instead of one-time purchase.
The store reported +50% subscription rate, +23% add-to-cart rate, and +10% AOV, with new visitors converting 16% better.
Huron's buy box was reordered so the subscription came preselected, run with Visually.
Subscription opt-in rose 28.9% with no change in conversion rate.
The counterexample comes from Dave Diederen, who introduced subscriptions with the option pre-checked.
Day-one numbers fell: -2.86% conversion rate, -9.17% revenue per visitor, -6.50% AOV.
Subscription share of orders rose 1,295%. Dave's point is that asking "buy this and get it again automatically" is a bigger commitment, so the test has to be judged on lifetime value, not first-order revenue.
In apps, defaults around trials show the same tension.
Linearity Curve made its free trial an opt-in rather than a default and got +10% trial conversion rate, though with fewer trial starts and no net revenue figure. See the free trial tests for more.
Defaults that lost: pre-ticked add-ons
Pre-ticked fees are where defaults most often backfire.
Dave Diederen tested hiding a pre-ticked shipping protection checkbox in the cart: +6.39% conversion rate, +3.74% revenue per visitor, and +2.33% profit per visitor, with AOV down 2.49%.
Convertibles ran a 3-way test for an electronics accessories brand, replacing a shipping protection widget with a single toggle set to off.
Conversion went from 1.53% to 1.90% (+23.7%) and revenue per visitor rose 16.4%, at 97% probability to beat control on conversion and 82% on revenue.
AOV fell 6%, partly from fewer protection fees.
Oliver Kenyon saw the same pattern when they unticked pre-ticked accidental damage cover in the cart and sold it explicitly instead. The variant won.
There is a legal reason to avoid these defaults too.
In the EU, Article 22 of the Consumer Rights Directive says that if a trader infers consent to an extra payment "by using default options which the consumer is required to reject," the consumer is entitled to a refund (EUR-Lex).
Our page on dark patterns covers where the line sits.
A default can also be a greener or better-value choice
Kronans Apotek's checkout preselected free delivery. Conversionista tested preselecting a greener delivery method with several delivery-price combinations.
The best variant produced +11% net profit, +2% revenue, and +25% green deliveries.
The default steered choice without hiding the alternatives.
How to test a default without fooling yourself
| Default type | Usual outcome in our tests | Metric to judge it on |
|---|---|---|
| Best-seller or most common variant preselected | Often wins | Conversion rate and revenue per visitor |
| Higher quantity preselected | Can win, watch returns | Revenue per visitor, units per order |
| Subscription preselected | Raises opt-in, can cut day-one revenue | Subscriber LTV, churn, revenue per visitor over months |
| Paid add-on pre-ticked | Often loses, legal risk in the EU | Conversion rate and profit per visitor |
| Trial included by default | More starts, sometimes weaker trial conversion | Paid conversions per visitor, not trial starts |
A few rules follow from the tests above:
- Measure the full funnel. A default almost always moves the share of people who keep it. The question is whether more people finish buying and whether they stay. Pick a primary metric like revenue or profit per visitor before launch.
- Wait for downstream data on subscriptions. Opt-in rate is known in days. Churn and refunds take months. Plan the readout around subscription metrics, not the first week.
- Default to what most people would pick anyway. The Stone Creek and Revant wins preselected the popular option. Defaults that match intent save effort. Defaults that fight intent create doubt.
- Keep the switch obvious. If changing the default takes more than one tap, you are no longer nudging.
- Don't expect the research effect size. Publication bias inflates the average. Your test is the evidence that counts.
Defaults also interact with how many options you show.
If a page has too many choices, a sensible default can rescue it, but trimming options may work better.
You can also browse every test in the database.