The half-finished thing your users can't stop thinking about

There’s a story every psychology student hears about a waiter. He can hold a dozen complicated orders in his head with no notes, recite them back perfectly, deliver every plate to the right seat — and then the moment the bill is paid, the whole table evaporates from his memory. Ask him ten minutes later what those people ate and he genuinely can’t tell you. The paid bill closed the loop, and the brain, efficient as ever, let go.

Whether or not that exact scene happened, it’s the seed of one of the more useful findings in psychology. In the 1920s Bluma Zeigarnik ran experiments where people were given a series of small tasks and interrupted partway through some of them. Later she asked what they remembered. People recalled the interrupted tasks far better than the ones they’d been allowed to finish — by a wide margin in her original work. An open task keeps a kind of low-grade tension humming in memory. A completed one is allowed to fade.

Think about what that means. Your mind treats an unfinished thing not as a neutral fact but as a slightly uncomfortable open state — a tab that won’t close, quietly drawing attention until it’s resolved. That discomfort is not a bug. It’s the mechanism that makes you come back.

Now walk through your product with that lens on, and a whole shelf of familiar patterns stops being decoration and starts being applied Zeigarnik.

The progress bar isn’t a status readout — it’s an open loop you’re refusing to let close. “Profile 60% complete” works because the missing 40% is now an itch. The bar didn’t inform the user of anything they cared about; it manufactured an unfinished task and put it in front of them. A close cousin makes it even stronger: the goal-gradient effect — the finding that motivation intensifies the nearer a goal gets. People push harder in the last stretch than the first. Which is why a bar that shows you’re almost there pulls harder than one showing you’ve barely begun.

“Resume where you left off” is the same trick, aimed at re-entry. A blank home screen asks the user to start something. A screen that shows the draft they abandoned, the level they were on, the cart they didn’t check out, hands them a loop that’s already open — and open loops want to close. The related Ovsiankina effect is the direct evidence here: interrupt people mid-task and, given the chance, they spontaneously go back to finish it, even without being asked. Your “continue” button is just making that easy.

Pre-fill the first step and you steal progress the user hasn’t earned. The endowed-progress effect is one of my favorite findings in this whole area: give people a loyalty card that needs ten stamps and start them at zero, versus a card that needs twelve but arrives with two stamps already on it, and the second group — identical work remaining — finishes far more often. The pre-stamped card converts “start a task” into “finish a task I’ve already begun,” and the brain treats those completely differently. An onboarding checklist where step one is already ticked (“account created ✓”) is the same move.

There’s an ethical edge here worth naming, because the wedge cuts both ways. Zeigarnik tension is discomfort. A ruthless product can farm it — leave a hundred loops half-open to keep someone anxiously returning, a notification badge that never quite reaches zero. That’s using the mechanism against the person. The honest version uses it to reduce friction on things the user actually wants to finish: nudging them past the boring setup toward the value, helping them pick back up the work that was theirs to begin with. Same neuroscience; the difference is entirely in whether the loop you’re leaving open is one the user is glad to close.

The reframe I’d hand a first-time builder: stop thinking of your product as a set of finished screens to inform people, and start thinking of it as a set of loops — some open, some closed. Every open loop is a small pull on attention. Your job isn’t to eliminate that tension. It’s to point it at the one thing you’d genuinely want the user to come back and complete.