The Zeigarnik effect is the claim that people remember unfinished tasks better than finished ones, first reported by Bluma Zeigarnik in 1927. A 2025 meta-analysis of the original research method found no reliable memory advantage for interrupted tasks, but it did confirm a related effect: people resume interrupted tasks about two-thirds of the time.
For conversion work, that means an open loop is a reason to come back and finish, not a guaranteed lift.
This page covers the three ideas behind most progress bars, checklists and "spend $12 more" cart messages: the Zeigarnik effect, the goal-gradient effect and the endowed progress effect.
For each one we separate what the research found from what real A/B tests in our database show.
What the Zeigarnik effect is, and how well it holds up
In Zeigarnik's experiments, participants worked through a series of small tasks.
Some were interrupted before they could finish. Afterward, participants were asked to recall the tasks, and Zeigarnik reported that they recalled more of the interrupted ones.
The idea came from Kurt Lewin's theory that an intention creates a tension that lasts until the goal is met.
A year later, Maria Ovsiankina used a similar setup and found that people tended to go back to interrupted tasks on their own when given the chance.
That second finding became known as the Ovsiankina effect.
The memory claim has not held up well.
Ghibellini and Meier's 2025 meta-analysis in Humanities and Social Sciences Communications pooled studies that used the interrupted-task paradigm.
The weighted ratio of interrupted to completed tasks recalled was 0.99 across 37 publications (excluding Zeigarnik's own), which means people recalled both about equally.
The resumption claim did hold up. Across 20 publications (excluding Ovsiankina's original), the weighted resumption rate was 66.79%, well above the 50% you would expect by chance.
The authors conclude that the Ovsiankina effect is a general tendency, while the Zeigarnik effect lacks universal validity.
Practical reading: "unfinished things stick in memory" is weak. "People who started something tend to want to finish it" is better supported, and it is the version that matters for forms, carts and onboarding.
So when a CRO article says the Zeigarnik effect will make visitors remember your abandoned cart, treat it as a hypothesis.
The useful mechanism is closer to commitment and resumption, which overlaps with the consistency principle: a task that is visibly half-done feels easier to finish.
The goal-gradient effect: effort speeds up near the finish
The goal-gradient effect says people put in more effort the closer they get to a goal.
Clark Hull proposed it in 1932 in "The goal-gradient hypothesis and maze learning" (Psychological Review 39: 25-43), based on rats running faster as they neared food.
The best-known consumer test is Kivetz, Urminsky and Zheng (2006) in the Journal of Marketing Research.
Members of a "buy ten coffees, get one free" program bought coffee more often as they got closer to the reward: on average, time between purchases fell by 20%, or 0.7 days, over the course of the card.
The same paper found an "illusionary progress" effect.
Customers given a 12-stamp card with 2 bonus stamps already filled in finished the 10 required purchases faster than customers given a plain 10-stamp card.
The real distance to the reward was identical; only the feeling of progress differed.
Nunes and Drèze (2006) in the Journal of Consumer Research called this the endowed progress effect.
In a car wash field experiment, they turned an 8-stamp card into a 10-stamp card with 2 stamps already given, which raised the share of people who completed the card and shortened the time it took.
They report that the effect depends on the feeling of being partway through, not on a desire to avoid wasting the free stamps, and that it is moderated by how the head start is justified.
What the goal gradient looks like in A/B tests
In ecommerce, the closest application is a cart message or bar showing how far the shopper is from free shipping or a free gift.
Our database has several tests of this, and they mostly point in the same direction, with one important exception.
- Kiehl's added a free-gift progress gauge on the basket page and reported +31% revenue, +25% purchases and almost +5% basket size. The case study gives no sample size or significance, and the revenue lift is large relative to the basket-size change.
- Convertibles tested one progress line ("You've got 4 Free Gifts! Add 1 more slime to get more") at the top of a slime brand's cart drawer, worth +$50,099/month in revenue. Richer versions with thumbnails and upsell headlines did not beat the simple line, and a free-shipping-led version lost to the free-gift framing.
- Oliver Kenyon added tiered spend rewards in the cart (free shipping plus bonus items) and reported +$57k/month.
The exception is useful. Clean Commit added free-shipping progress to a mini cart and saw conversion rate fall 5.9% and orders fall 7.0%, while AOV rose 9.2% and revenue per visitor rose 3.2%.
The result was mixed, with the original showing a 73.7% probability of being better on conversion rate.
That pattern fits the theory. A goal gradient pushes people who are close to the threshold to add more, but a visible gap can also tell people far from it that they are "not there yet."
Whether the net effect is positive depends on where your typical cart sits relative to the threshold.
Our free shipping threshold hub collects every test on this.
Convertibles
Clean CommitProgress bar psychology: what research and tests actually show
Progress bars are meant to work through both effects above: they show that the task is underway and how close the end is.
The research on them is more mixed than most design advice suggests.
The largest review is Villar, Callegaro and Yang (2013), a meta-analysis of experiments on progress indicators in web surveys (Social Science Computer Review 31: 744-762).
A progress indicator that moved at a constant speed did not significantly reduce drop-off.
Speed mattered. Indicators that moved fast at first and slowed later reduced drop-off.
Indicators that started slow and sped up increased it.
In studies where a small incentive was promised, a constant progress indicator actually increased drop-off.
The lesson is that a progress bar carries information, and that information can be bad news.
A bar that shows "step 1 of 12" at 8% can discourage people as easily as it motivates them.
Progress bars in forms and onboarding tests
Lead-gen forms are where our database has the cleanest progress-bar tests.
- Tinkoff Bank tested two progress bars on a four-step credit card application, each showing the current step and an approval likelihood that rose as people filled it in. A green bar that filled up lifted completed applications 12.8%; a yellow bar with a progress line lifted them 6.9%. The test ran 39 days on more than 190,000 visitors.
- HomEquity Bank added a progress tracker to a multi-step inquiry form and reported +35% conversions. A later test, a tracker that explained what happens after submitting, added +20% conversion rate on top.
- Reassured replaced a long single-page quote form with one question per step and a progress bar, and reported +31.23% form submissions over about five weeks. The progress bar and the multi-step layout changed together, so the test cannot separate them.
Notice what the winning Tinkoff bar did: it tied progress to something the applicant wanted, a rising chance of approval, rather than just counting steps.
The HomEquity follow-up points the same way. Telling people what comes next reduces uncertainty about the end of the task.
In SaaS onboarding, the equivalent is a setup checklist.
AURUM added a floating onboarding checklist with a "First Steps" progress indicator to its trial and reported +187% activation, though it was a controlled rollout rather than a strict A/B split.
And progress bars do not rescue the rest of a funnel.
Patrick McKenzie's Bingo Card Creator redesign, which added a step progress bar to the app, lifted trial sign-ups from 20.9% to 23.1% on AdWords pages, but purchases were flat (50 vs 46).
Research vs A/B tests at a glance
| Idea | What research shows | What our A/B tests show |
|---|---|---|
| Zeigarnik effect (memory) | Not reliable: interrupted and completed tasks recalled about equally (ratio 0.99) | No test isolates memory; don't rely on it |
| Ovsiankina effect (resumption) | Interrupted tasks resumed about 67% of the time | Fits form and checklist wins, but tests measure completion, not resumption |
| Goal gradient | Purchase intervals shrink about 20% near a reward | Cart reward bars mostly win on revenue; one test lost conversion rate while AOV rose |
| Endowed progress | Pre-filled stamps raised completion in field experiments | Few direct tests; checklists that start partly done are worth testing |
| Progress bars | Constant-speed bars don't reduce drop-off; fast-to-slow helps, slow-to-fast hurts | Form progress bars won at Tinkoff and HomEquity; often bundled with other changes |
How to test these ideas without fooling yourself
Start with honest progress. Pre-filling a step the user really completed, such as account creation counting as step 1, is endowed progress.
Inventing progress that isn't real, or a bar that freezes at 90%, drifts into dark patterns and erodes trust once people notice.
Put easy steps first. The survey research suggests early movement matters, so front-load short questions and leave harder ones, like phone number or payment, for later.
Our multi-step forms hub covers step order in more detail.
Show distance only when it is short. A free-shipping bar works best when most carts are within reach of the threshold.
If the median cart is far below it, test hiding the bar until the shopper is close, and watch conversion rate as well as AOV.
Measure the outcome that pays. A progress bar can raise step completion and still leave purchases or paid conversions flat, as the Bingo Card Creator test showed.
Pick a primary metric downstream of the bar before the test starts.
Expect smaller effects than the headlines. Many of the biggest numbers here come from vendor case studies or bundled redesigns.
Treat them as reasons to run your own test, not as a forecast.
You can browse more tests by page and element, or read how related biases hold up on our psychology hub.