Persuasion psychology is the study of the predictable mental shortcuts, such as social proof, scarcity, loss aversion, defaults and anchoring, that shape what people choose when they can't weigh every option carefully.
For a website, these shortcuts are a source of test ideas, not a guarantee: the same countdown timer that lifts revenue on one store has cut it by more than a third on another.
Use the research to decide what to try, then let an A/B test decide what to keep.
This page is the hub for our psychology section.
It covers what lab research says about each major bias, what the real A/B tests in our database say, and how to turn one into the other without fooling yourself.
Every claim about a website change links to a test, with the result and the person or team who ran it.
Consumer psychology: what the research actually shows
Consumer psychology looks at how people search, compare, decide and buy.
Two bodies of work dominate how marketers use it: behavioral economics, which grew from Daniel Kahneman and Amos Tversky's research on judgment, and the persuasion research collected by Robert Cialdini.
Kahneman and Tversky's 1979 paper "Prospect Theory: An Analysis of Decision under Risk" in Econometrica argued that people judge outcomes as gains or losses from a reference point, and that losses loom larger than gains of the same size.
That idea underpins most of what CRO calls loss aversion, framing and anchoring.
Cialdini's book Influence grouped persuasion into six principles: reciprocity, commitment and consistency, social proof, liking, authority and scarcity.
The 2021 expanded edition added a seventh, unity, the pull of a shared identity.
Our Cialdini principles guide maps each one to tests, and the cognitive biases index covers the wider list.
Neither body of work was built for websites.
Prospect theory was tested mostly with hypothetical gambles, and many persuasion studies used students, hotel guests or shoppers in a single store.
That doesn't make them wrong. It means each finding is a hypothesis about your visitors until you test it on them.
The classic studies behind the biases marketers use
Most of the effects you see on CRO checklists trace back to a small number of famous studies.
Some have held up well. Others are weaker than their reputation.
| Effect | Classic study | What later research found |
|---|---|---|
| Social proof (descriptive norms) | Hotel signs saying most guests reuse towels beat a plain environmental appeal (Goldstein, Cialdini and Griskevicius, 2008) | Two replications in German hotels found the norm messages were not more effective than the standard message (Bohner and Schlüter, 2014) |
| Scarcity | People rated cookies from a jar of two higher than from a jar of ten, and higher still when the supply had just dropped (Worchel, Lee and Adewole, 1975) | A small lab study of value ratings, not purchases. Treat it as a starting idea, not proof that low-stock badges sell |
| Loss aversion | Losses loom larger than gains (Kahneman and Tversky, 1979) | Contested. Gal and Rucker (2018) argued it has been overgeneralized into a universal law; other researchers have defended it |
| Choice overload | 3% of shoppers who stopped at a 24-jam display bought, against 30% at a 6-jam display (Iyengar and Lepper, 2000) | A meta-analysis of 50 experiments found a mean effect of virtually zero, with large variation between studies (Scheibehenne, Greifeneder and Todd, 2010) |
| Defaults | Opt-out countries had far more people in the organ donor pool than opt-in countries (Johnson and Goldstein, 2003) | Holds up well on average: a meta-analysis of 58 studies found d = 0.68, though several found no effect and two found negative effects (Jachimowicz et al., 2019) |
| Decoy effect | Adding an option that is clearly worse than one choice raised the share picking that choice (Huber, Payne and Puto, 1982) | Widely reproduced in lab settings; how strongly it works on real pricing pages is a test question |
The replication problem
Psychology has a well-documented reliability problem. When the Open Science Collaboration repeated 100 published studies, only 36% of the replications had statistically significant results, against 97% of the originals, and the effects were about half as large on average.
Nudges, the applied version of these biases, show the same pattern.
A 2022 PNAS meta-analysis of more than 200 studies reported an average effect of d = 0.45.
A reanalysis in the same journal argued that after correcting for publication bias, no evidence for an overall nudge effect remained, though some categories, such as defaults, held up better than others.
The practical takeaway is simple. A famous study tells you a bias exists somewhere, under some conditions.
It doesn't tell you the size of the effect on your checkout, with your customers, at your price point.
The psychology of conversion: what real A/B tests show
The psychology of conversion is where lab findings meet live traffic.
Our database holds 1,724 test results from ecommerce stores, SaaS companies, lead-gen sites and apps, and a large share of them are psychology tests in practice: badges, timers, stock messages, guarantees, defaults and price framing.
Read the totals with care. Of the 1,724 results, 1,447 are listed as wins, because practitioners publish winners far more often than losers.
At Microsoft, Ronny Kohavi's team found that only about one-third of well-designed experiments improved the metric they targeted.
Assume your own win rate for any bias will be closer to that than to our database's.
Our research on losing A/B tests covers this in more depth.
Social proof: strong on average, not automatic
Tests tagged social proof in our database include 87 wins, 9 flat results and 6 losses.
Simple popularity labels do well. Dave Diederen added a "Best Seller" label directly above the product title and measured +12.99% revenue per visitor.
Scarcity of the label itself seems to matter.
In another of Dave Diederen's tests, putting the Bestseller badge on only the single #1 product on a collection page, instead of on three or four, gave +8.89% revenue per visitor.
But more social proof isn't always the answer.
Dave Diederen also tested a product page layout focused on benefits instead of social proof, and it won by +43.1% revenue per visitor.
Clean Commit's test of "New" and "Best Seller" badges on the main product image came back flat.
See the social proof and bandwagon page for the full pattern.
Scarcity and urgency: the widest spread of results
No psychology tactic in our database swings both ways as hard as urgency.
The same element, a checkout countdown timer, produced +3.16% revenue per visitor in one of Dave Diederen's tests and -7.1% revenue per visitor in another.
Market matters too. Dave Diederen's countdown timer for a European market lost by -35.61% revenue per visitor.
Kasper Dolk went the other way and removed a countdown timer from checkout, which raised revenue per visitor by 26.18%.
Stock messages show the same split. Visually's test of a subtle "X left" stock line below the CTA for Odd Muse lifted conversion rate by 22%.
Clean Commit's low-stock announcement bar on product pages cut conversion rate by 17.5%.
And Dave Diederen found that showing "In Stock" instead of "Almost Out of Stock" raised revenue per visitor by 16.78%.
The common thread is believability. Subtle, specific, true scarcity tends to help; loud or generic pressure can backfire.
The scarcity and urgency page breaks down when each version wins.
@DaveDiederen
@kasper_dolk
Visually@DaveDiederen@DaveDiederen
Clean CommitLoss aversion and risk reversal
Guarantees are the most common risk-reversal test in our database: 45 wins, 5 flat and 3 losses among tests tagged guarantee.
Dave Diederen added a "30-day money-back guarantee" under the checkout button and saw +18% revenue per visitor.
Still, DRIP Agency's test of a more prominent money-back guarantee lowered purchases by 9.2%.
One plausible reading is that a very loud guarantee can make shoppers think about the product failing.
That's a hypothesis, not a finding, and the next test is how you'd check it.
Savings framing is the other loss-aversion lever. Putting the savings into the button itself, "Add to Cart - Save 50%", gave +14.41% revenue per visitor for Dave Diederen.
Yet Conversion Fanatics found that hiding bulk savings on quantity options raised conversions by 25.1%.
Defaults: powerful in both directions
Defaults are one of the better-supported effects in research, and the tests agree they move behavior.
Whether that movement is good for revenue depends on what you default people into.
Blend Commerce preselected subscription on a product page and raised the subscription rate by 50%.
Dave Diederen introduced subscriptions with the option pre-checked and saw revenue per visitor fall 9.17%.
SplitBase's test of a quantity selector defaulting to two soaps for Dr. Squatch gave up to +54% revenue per user.
The reverse can also win. Linearity Curve made its free trial an opt-in instead of a default and raised trial conversion rate by 10%.
More on this in the default effect guide.
Anchoring and choice: order and number of options
Anchoring says the first number people see shapes how they judge the rest.
Elevate_ab tested showing the most expensive bundle first for Repeat Undies and measured +8.47% revenue per visitor. Our anchoring page collects similar price-order tests.
Choice overload is weaker in the meta-analysis than in the jam study, and the tests show why: it depends which choices you remove.
DRIP Agency showed 75% fewer products on a collection page and gained 14% revenue.
But MarketingExperiments found that removing a subscription-length choice cut conversion by about 40%.
The paradox of choice page covers when fewer options help.
Authority and reciprocity
Authority cues borrow trust from an expert. Ultraviolet Digital placed an "Expert opinion" block above a booking form for Kaya Skin Clinic, and conversion went from 4% to 9.5%.
MECLABS Institute replaced "Contact Sales" with "Speak to an Expert" plus a demo CTA and saw +291% clicks.
Authority results also show how easy it is to read too much into early data.
Ruben added a doctor and expert endorsement badge and saw +17% conversion rate after one week at a 99.6% probability, yet the test is recorded as flat.
A one-week number is a signal to keep testing, not a result.
Reciprocity, giving something first, usually shows up as free gifts.
Those can work, but Oliver Kenyon found that lowering the price and removing the free gift ($82 vs $85 plus a gift) raised conversion rate by 46%.
For that audience, a lower price beat a bonus.
What the research says vs. what our tests say
| Bias | Research verdict | A/B test verdict in our database |
|---|---|---|
| Social proof | Real, but the famous towel result didn't replicate in Germany | Mostly wins; can lose to clearer benefits |
| Scarcity and urgency | Supported mainly by small lab studies of perceived value | Widest spread: big wins and big losses for the same element |
| Loss aversion | Contested as a universal rule | Guarantees and savings framing usually win; loud versions can lose |
| Defaults | Among the most robust effects | Reliably move behavior; revenue impact depends on what you default to |
| Choice overload | Average effect near zero across studies | Cutting clutter can win; cutting a needed choice can lose badly |
Behavioral psychology marketing: how to use it without fooling yourself
Behavioral psychology marketing is the practice of designing pages, offers and messages around how people actually decide.
Done well, it means fewer guesses and better test ideas.
Done badly, it means adding a timer because a blog said timers work.
Start with the problem, not the bias
A useful diagnostic is BJ Fogg's model, B = MAP: a behavior happens when motivation, ability and a prompt come together at the same moment.
If one is missing, the behavior doesn't happen.
The Fogg Behavior Model page applies it to checkout and sign-up flows.
Use it to pick the right lever:
- Low motivation (visitors don't see why they should buy): social proof, authority, benefits and value framing.
- Low ability (the task feels hard or risky): defaults, fewer form fields, clearer options, guarantees.
- Weak prompt (nothing tells them to act now): clearer CTAs, honest deadlines, real stock levels.
Session recordings, surveys and analytics tell you which of the three is missing.
A bias chosen without that evidence is a coin flip.
Match the bias to the visitor
The same visitor doesn't need the same push every time.
A first-time visitor lacks trust, so social proof and guarantees have more work to do.
A returning visitor already knows the brand and may respond more to a reminder or a deadline.
Endless Gain tested this directly for Fragrance Direct, showing social proof messages to new visitors and lifting conversion rate by over 8%.
Optimizely ran a similar test for a retailer, showing bestsellers instead of featured picks to new visitors, and click-through rose 23.01%.
If your testing tool supports it, check results by new and returning visitors and by device before you call a psychology test.
An average that looks flat can hide a win in one group and a loss in another.
Treat any segment finding as a new hypothesis to confirm, since slicing data many ways makes chance wins likely.
Write the hypothesis in plain terms
A good psychology hypothesis names the bias, the place, the audience and the metric.
For example: "First-time visitors hesitate on the product page because they can't tell which variant is most popular.
Labeling the top seller will raise revenue per visitor."
That structure forces you to state why the bias should work here.
It also makes a losing result useful, because you learn something specific about your customers rather than "timers don't work."
Measure the right outcome
Many psychology tactics raise a click or a start without raising money.
BoldVoice tested a hard paywall with no close button: trial starts went up, but refunds rose by more than the gain.
Pressure that converts the wrong people shows up later in refunds, churn and support tickets.
Track revenue per visitor or a downstream metric alongside the conversion rate.
Run the test long enough to reach a planned sample size, because small psychology tweaks usually produce small effects that need a lot of traffic to detect.
A worked example shows the scale. Say your product page converts at 3% and you expect a trust badge to lift that by 5% relative, to 3.15%.
With a two-sided test at 95% confidence and 80% power, the standard formula is:
n per variant = (1.96 + 0.84)² × [p1(1 − p1) + p2(1 − p2)] / (p2 − p1)²
Plugging in gives 7.84 × (0.0291 + 0.0305) / 0.0015², or roughly 208,000 visitors per variant.
If your page gets 20,000 visitors a month, that test can't finish in a reasonable time.
Test bigger, bolder changes instead, or test on a higher-traffic page.
Keep it honest
Fake countdown timers, invented stock counts and pre-ticked add-ons can win a short test and still cost you trust, chargebacks and legal risk.
The dark patterns page lists the tactics to avoid.
Honest versions can win on their own. Oliver Kenyon tested authentic scarcity messaging on product pages and reported +$46,900 per month.
Real scarcity, stated plainly, gives you the benefit without the downside.
Common mistakes
- Treating a bias as a law. Every effect in the table above has tests that went the other way.
- Stacking tactics. A timer, a stock warning, a popup and a badge on one page add noise and pressure. Test one idea at a time where traffic allows.
- Copying another store's winner. The European timer loss shows that audience and market change the result.
- Calling a test early. A lift that appears in the first few days often shrinks with more data.
- Ignoring the losers. Published results lean toward wins; your own losses are the most useful data you have.
To see how a specific bias has performed, browse the full test database by page type and element, or go to the bias pages linked throughout this guide.