The Zeigarnik Effect: Why an Unfinished Game Follows You to the Store

The finding is usually told as a story about waiters, and we have told it that way ourselves. It is worth stating what the effect actually claims, because the popular version drifts.

It is not that people remember unfinished things. It is that interrupted tasks stay mentally active in a way finished ones do not. The distinction matters enormously. A task you chose to stop doing closes almost as cleanly as one you completed — the mind treats abandonment as a kind of ending. A task that was cut short while you were still engaged in it does not close at all. It sits there, using a small amount of attention, quietly available.

So the mechanism is not about leaving things undone. It is about who did the leaving.

That single correction changes what it tells you to build. It is not a licence to stop the ad early. It is a fairly narrow instruction about the state a player should be in when the ad ends.

Here is the part of this business that gets almost no attention, and it is where most campaigns lose the people they paid for.

A player finishes an ad interested. Genuinely interested — not politely, not accidentally. They tap through. And then a store page loads, and there are screenshots that look like a different product, and a rating, and a description, and possibly an age gate or a payment confirmation or a sign-in. Somewhere in there the feeling that made them tap dissipates, and they close the tab.

Nothing went wrong in the ad. The ad worked. What failed is that interest is perishable, and the walk from an ad to an installed app is long enough for ordinary interest to expire on the way.

An open loop is the only thing we know of that reliably survives that walk. Not because it is stronger than interest, but because it is a different kind of thing: interest fades on its own, while an interrupted task keeps asking to be resumed until something closes it.

The two markers worth arguing about are 0:18 and 0:22, and they are in that order for a reason. Completion comes first, then the glimpse of what is next. Reverse them and the player is interrupted while failing rather than while intending, which produces a quite different and much less useful feeling.

The same cut produces two opposite reactions depending on one variable: whether the player believes the next part was reachable.

If they were about to do something they knew how to do, the interruption leaves an intention. If they were stuck, or losing, or had no idea what came next, the interruption leaves relief — and relief closes the loop immediately, because there is nothing the person wants to resume.

This is why difficulty is such a bad tool here. The instinct is that a player who almost succeeded will want to try again, so make it hard. What actually happens is that a player who could not see the path out experiences the ending as an escape.

An open loop is a promise with a specific shape. It says: there is a next part, and it is the next part of this.

Which means the product has to be the place the loop closes. If the ad interrupted a match-three board and the installed app opens on a base-building screen, the loop does not close — it is abandoned, and abandonment does not merely fail to convert. The player has been told something inaccurate about what they were getting, and they now know it.

This is the mechanism behind a pattern anyone who buys installs will recognise: strong click-through, decent install rate, and users who are gone within a day. The loop did its job all the way to the download. It could not survive first contact with a product that had nothing to do with it.

The end card matters here more than it usually gets credit for. It is the last moment the loop is still intact, and its job is not to summarise the product. It is to make continuing look like the obvious next action — same art, same mechanic, same thing they were in the middle of, and as few steps as possible between that feeling and the install.

Most of this comes down to where the ad stops, which is usually decided late and by whoever is cutting the file to length. It deserves more thought than that.

This is the most abusable idea in this category, and it would be dishonest to present it without saying so.

If an open loop drives action, then a loop engineered never to close drives more of it. Ads built so the player is always one move away and never allowed to arrive. Products where the interrupted thing does not exist and the loop simply stays open, unresolved, indefinitely. The mechanism works either way — it does not check whether the promise is real.

The argument against this is not only ethical, though it is that. It is that the loop is a debt. Every open loop is a commitment to close it somewhere, and a product that takes the install without closing it has borrowed attention it cannot repay. That shows up as day-one churn, refund requests and reviews written by people who feel they were tricked, because they were.

The test we would apply: if you explained the interruption to the player afterwards — we stopped there on purpose, because the next bit is in the app — would they feel misled? If the next bit really is in the app, no. They would probably think it was fair. If it is not, the explanation is a confession.

Some care about how far this should be pushed.

The original research was done on laboratory puzzles and recall tests, not on advertising, and the effect has had a mixed replication history — it is real but smaller and more condition-dependent than its popular retelling suggests. Anyone quoting a percentage uplift from it is quoting something that does not exist.

It is also only one of several things happening at the end of an ad, and probably not the largest. Whether the product is appealing, whether the store page matches what was shown, whether the person had a reason to want this kind of thing at all — each of those outweighs the shape of the interruption.

What the effect gives you is narrower: a reason to think carefully about the last four seconds, which are usually decided carelessly. The difference between ending on a success and ending on an intention is free to implement, and one of them leaves something behind.