Cognitive biases in marketing are the predictable mental shortcuts people use when they judge prices, risk and other people's choices, such as anchoring, loss aversion, social proof and defaults.
Marketers use them to shape how an offer is presented.
In real A/B tests, the same bias can lift revenue on one store and cut it on another, so treat each one as a hypothesis, not a guaranteed lift.
Below are 25 biases, each paired with a test from the LiftArchive database, including the ones that lost.
For deeper pages on single biases, start at the psychology hub.
Cognitive bias examples from real A/B tests
The clearest examples are the ones where a store changed only how something was presented, not the product or the price.
Here are seven, chosen because they show both sides.
Scarcity can win big or lose big
A golf brand working with Convertibles added a stock meter and a note that the product had sold out twice that year.
Revenue per visitor rose 26.4%, with 368 orders against 297 for the control.
The meter ran on a timer rather than live inventory, which is the kind of thing that can cross into dark-pattern territory.
Clean Commit ran a louder version: a site-wide announcement bar saying stock was low.
Conversion rate fell 17.5% in a short test of 5,161 visitors.
Scarcity worked when it was specific and close to the buy button, and failed when it was a generic banner.
More on this split on our scarcity and urgency page.
Urgency depends on the market
Dave Diederen's checkout countdown timer raised revenue per visitor 3.16% for US brands.
The same team reports that countdown timers in European markets, tested across 20+ brands, cut revenue per visitor 35.61%. Same bias, opposite outcome.
Defaults move behavior, but not always revenue
Preselecting a subscription at Jackson's, a snack brand working with Blend Commerce, lifted the subscription rate 50% and add-to-cart 23%.
When Dave Diederen introduced subscriptions pre-checked by default on another store, subscription share of orders jumped 1,295% but revenue per visitor fell 9.17% upfront.
The default worked; whether it paid off depends on retention. See the default effect page.
Framing: percent beat dollars, and "free" lost
On one product page, showing savings as a percentage lifted revenue per visitor 6.42%.
When the same team tested dollar savings above $100 on the same brand, the result was net negative, with conversion down 1.08%.
The bigger surprise came from Milos: changing a 2-pack at $240 to "Buy 2 Get 1 Free" at $240 gave buyers a third unit at no extra cost, yet revenue per visitor fell 52.38%.
A deal that looks too generous can read as a warning sign. Our framing effect page covers more cases.
Authority signals can be noise
Ruben added a doctor-endorsement badge and saw +17% conversion after one week at 99.6% probability to beat control.
After four weeks, the lift was +1.66% and flat.
Dave Diederen went the other way and hid an "As Seen In" press banner, which raised revenue per visitor 5.63%.
Fewer choices sometimes sell more
Oliver Kenyon removed a 1/2/3-bag bundle selector from a travel bag brand's product page, leaving one bag, one price and one button.
Revenue per visitor rose 7.8% and conversion rose 8.7%.
That fits choice overload, though the research on it is weaker than its fame suggests (see below).
Standing out only works if few things stand out
A "Bestseller" badge on three or four products lost 4.72% conversion.
Moving it to only the single top product raised revenue per visitor 8.89%.
That is the isolation effect: a label stops working once everything has one.
List of cognitive biases used in marketing, each with a test
This list covers the 25 biases marketers use most, grouped roughly by pricing, risk, social influence and choice design.
Every row links to the test behind it.
Results are the headline numbers reported by whoever ran the test; several come from vendor case studies or before/after comparisons, which are noted.
| Bias | What research predicts | Test in our database | Result |
|---|---|---|---|
| Anchoring | The first number seen sets the scale for judging the next one. | Price and compare-at savings moved directly below the title (Dave Diederen) | +6.95% revenue per visitor, +7.39% conversion rate |
| Decoy effect | Adding a weaker option makes a target option look better. | Three plans with yearly marked best offer (Rash ID) | About +20% ARPU; over +50% conversion to paid |
| Framing (percent vs. dollar) | The same saving feels different depending on how it is stated. | Savings shown as a percentage instead of dollars (Dave Diederen) | +6.42% revenue per visitor |
| Bonus-pack framing | "Free" units should make a bundle feel like a better deal. | 2-pack changed to "Buy 2 Get 1 Free" at the same price (Milos) | -52.38% revenue per visitor (lost) |
| Temporal reframing | A price split into small periods feels smaller. | Annual plan shown as a per-month price (Mojo) | +45% Brazil, +26% Mexico, +8% USA in revenue per paywall impression |
| Unit-price framing | A per-unit price makes value easier to judge. | Price per serving on collection cards (Blend Commerce) | +8.25% revenue per visitor (directional) |
| Zero-price effect | Free is treated as a different category, not just a low price. | 7-day free trial on the yearly plan (Stardust) | +131% paywall conversion |
| Loss aversion | Losses weigh more than equal gains, so reducing risk of loss helps. | 30-day money-back guarantee under the checkout button (Dave Diederen) | +18% revenue per visitor, +17% conversion rate |
| Ambiguity aversion | People avoid options with unknown costs. | Removed "Shipping calculated at checkout" on a free-shipping PDP (Dave Diederen) | +14.07% revenue per visitor |
| Scarcity | Limited items seem more valuable. | Stock meter on a golf product page (Convertibles) | +26.4% revenue per visitor |
| Urgency | Deadlines push people to act now. | Countdown timer in checkout (Dave Diederen) | +3.16% revenue per visitor |
| Social proof | People copy what others like them do. | Star rating and review count above the product title (Dave Diederen) | +8.45% revenue per visitor |
| Bandwagon effect | Popular choices feel safer. | "Best Seller" label above the product title (Dave Diederen) | +12.99% revenue per visitor |
| Isolation (von Restorff) effect | The item that stands out gets noticed and remembered. | Bestseller badge on only the #1 product (Dave Diederen) | +8.89% revenue per visitor; the 4-badge version lost 4.72% conversion |
| Authority | Expert endorsement raises trust. | Doctor/expert endorsement badge (Ruben) | +17% conversion after week 1, +1.66% (flat) after 4 weeks |
| Default effect | Most people keep whatever is preselected. | Subscription preselected on the product page (Blend Commerce) | +50% subscription rate |
| Reciprocity | A gift creates a wish to give back. | Free BBQ recipe e-book with bundles (Oliver Kenyon) | +$68,000/month, +$0.74 revenue per session |
| Commitment and consistency | A small first yes makes a bigger yes more likely. | Two-step yes/no popup before the email field (OptiMonk) | +76% email sign-ups (before/after, not a split test) |
| Goal-gradient effect | Effort rises as a reward gets closer. | Free-gift progress gauge in the basket (Kiehl's) | +31% revenue per visitor (no significance reported) |
| Choice overload | Too many options can stall a decision. | Removed the 1/2/3-bag bundle selector (Oliver Kenyon) | +7.8% revenue per visitor, +8.7% conversion rate |
| Primacy (order) effect | The first option listed gets extra weight. | Biggest pack listed first in the bundle selector (Elevate A/B Testing) | Revenue per visitor €3.54 to €3.84 (+8.47%) |
| Affect heuristic | How a choice feels shapes how good it seems. | Confetti and "biggest savings unlocked" on the largest bundle (Dave Diederen) | +6.32% revenue per visitor |
| Salience | People weigh what is in front of them at the moment of choice. | "Add to Cart - Save 50%" button copy (Dave Diederen) | +14.41% revenue per visitor |
| Endowment effect | Things that feel like yours are valued more. | "Your cart is ready" homepage reminder (Visually) | +24.3% conversion rate (vendor case study) |
| Zeigarnik/Ovsiankina effect | Unfinished tasks pull people back. | Saved-basket pop-up for returning visitors (Pets at Home) | Over +3% reaching the basket, over +2.5% reaching order confirmation |
@oliverkenyon
@DaveDiederen@DaveDiederen
Convertibles@DaveDiederen
@CROguy_A few rows involve more than one bias.
The biggest-pack-first test is also anchoring, and the Rash ID test added a third plan and a badge together, so the decoy effect is only one possible explanation.
Pricing tricks without a clean test in our database, like charm pricing, are left out of this list.
Cognitive biases: what research shows vs. what A/B tests show
Lab research tells you a bias exists under certain conditions.
An A/B test tells you whether it moved your metric, with your traffic, this month. You need both, and they often disagree.
Where the research is strong
- Anchoring. In Tversky and Kahneman's 1974 Science paper, people who saw a wheel stop at 65 estimated a median 45% of UN members were African nations; those who saw 10 estimated 25%. Anchoring was one of the effects that replicated in the Many Labs project across 36 samples.
- Defaults. Johnson and Goldstein (2003) showed that European countries with opt-out organ donation had far higher consent rates than opt-in countries.
- Foot-in-the-door. In Freedman and Fraser (1966), 52.8% agreed to a large request after a small one, against 22.2% asked cold.
- Social proof. Goldstein, Cialdini and Griskevicius (2008) found hotel signs saying most guests reused towels beat plain environmental appeals. More on Cialdini's work on the Cialdini principles page.
- Decoys, scarcity, zero price and goal gradient. Each has a well-known original study: Huber, Payne and Puto (1982) on decoys, Worchel, Lee and Adewole (1975) on scarce cookies, Shampanier, Mazar and Ariely (2007) on free products, and Kivetz, Urminsky and Zheng (2006) on coffee-card stamps. Gourville (1998) covers pennies-a-day pricing.
Where it is shaky
- Choice overload. The famous jam study by Iyengar and Lepper (2000) found about 30% bought from 6 jams versus 3% from 24. A meta-analysis of 50 experiments found a mean effect of virtually zero, with large variation between studies.
- Loss aversion. Prospect theory (1979) says losses loom larger than gains. Gal and Rucker (2018) argue the evidence supports a much narrower claim than the universal rule marketers repeat. Details on our loss aversion page.
- Zeigarnik effect. A 2025 meta-analysis found no reliable memory advantage for unfinished tasks. People did tend to go back to interrupted tasks, which is the Ovsiankina effect, and the closer fit for cart reminders.
- Endowment effect. The classic mug experiments (Kahneman, Knetsch and Thaler, 1990) involved owning a real object. A cart reminder is a much weaker sense of ownership, and the DIME result could just as easily be a plain reminder effect.
- Nudges in general. A 2022 PNAS meta-analysis of 200+ studies found a small-to-medium average effect. A re-analysis correcting for publication bias found no evidence of an effect left. Published wins are a biased sample, and so are the wins people post online.
How to use a bias in a test
- Start from a visitor problem, not a bias. "Shoppers can't tell if shipping is free" led to the shipping-copy test that raised revenue per visitor 14.07%.
- Make it specific and honest. The scarcity and badge results above suggest precise, believable claims beat loud, generic ones.
- Run it long enough. The authority badge looked like a 17% win after a week. Run full weeks and decide your sample size in advance.
- Measure the money metric. Defaults and urgency can raise one number and lower revenue per visitor. Track revenue per visitor, not just clicks or subscription share.
- Expect losses. Several tests in the table lost. Our research on losing A/B tests shows how common that is.