Variable Reward

While return visits were treated as a question of notification volume, the answer was always the same one. Nobody's showing up, so send more; still nobody, so rewrite the copy. What sat on the other side of the tap went uncounted.

The table Mia read after tidying the home screen shows that spot. On the page notification taps land on, the scattered recent-documents list and the team activity feed got folded away, leaving two large buttons. Over the following two weeks, the share of notifications that got tapped fell from 41 percent to 26. The screen had not gotten worse; it had gotten identical. Someone who sees the same page three taps in a row does not make a fourth.

Variable reward is the term for what has to be on that page. If the notification is the knock, this is what's behind the door. When it differs, people open it again; when it never does, they learn to ignore the knocking.

Not knowing the outcome in advance is what holds attention. People check a mailbox daily not because there is always a letter but because today's contents are unknown. Knowing exactly what will be in there removes the reason to look.

In a product, that variation tends to come from three places.

1. Rewards from people — who reacted to something you left, and how. Three comments, silence, or a read marker from your manager: different every time.
2. Rewards from accumulation — what has piled up since you left. Documents the team touched yesterday, views compared with last week.
3. Rewards from searching — whether you found what you were after. Document search has this character: sometimes one query finds it, sometimes three do not, and the uncertainty is part of why people keep searching rather than giving up.

What matters in a collaboration tool is that the company does not produce this variation. Colleagues do. An active team makes the home screen different every day on its own, and in a quiet team no layout produces anything new to show. So working on this component starts from how many people are in the account, not from where things sit on the page. A single-person account has no reward to attach, which makes team invite the segment that comes first.

Three things get decided in the design: what varies, which screen shows the variation, and what appears when there is nothing. The last one gets skipped most often. On a day with no reactions at all, an empty list makes the visit feel like a loss. Those days get filled with material the user produced themselves — a document they edited yesterday, a comment they still owe an answer to. The point is that the visit ends with something rather than nothing, even on a day the team was silent.

Inserting artificial randomness is a different thing entirely. A wheel to spin on every login, or a random reward box, works for a few rounds and then gets seen through. In a work tool it happens faster: someone arrived to finish a task, not to test their luck, so variation unrelated to the outcome reads as an obstruction.

It occupies the third position in the hook model, which describes habit formation as a cycle of trigger, action, reward and investment. A trigger brings someone in and the action happens; this position decides whether there is a second turn. The first two can be built well and the cycle still stops here if this one is empty.

It gets handled together with notification design. Inside a notification strategy, which decides who hears what and when, this is the test for which notifications to send at all. Send the ones with something behind the tap, and batch the ones leading to an unchanged screen into a single daily digest. Without that test, notification volume climbs into notification fatigue, the state where users swipe messages away unread.

It runs opposite to a streak, which counts consecutive days. A streak is a reward whose result is known in advance, and its force comes from unwillingness to lose what accumulated. A variable reward gets its force from not knowing. They do different jobs, so neither substitutes for the other.

There are places it clearly does not belong: any screen where the outcome has to be predictable. Pressing save, completing a payment, exporting a file — those must produce the same result every time. Variation there does not read as interest, it reads as anxiety. A user who cannot predict what pressing save does will press it twice and then check the result, which costs more attention than any reward would return.

"We tidied the home screen and notification taps dropped. It became the same page every time."

Nothing changed behind the tap, so the reason to tap again disappeared.

"There's nothing to see behind this notification. Let's batch a day of them into one morning send."

Cutting notifications that carry no reward.

"Single-person accounts have no reward to attach. Invites come first."

The variation is produced by colleagues, not by the company.

Mia reverted the home screen at Pagedeck with one change. The recent-documents list came back open, but ordered by documents the team touched yesterday rather than documents the viewer created. Notification taps went from 26 percent to 44.

In accounts with one member, though, the number did not move. With no documents the team touched yesterday, the page stayed identical day to day. The same design worked in one group and not the other, so for that group the direction became pulling the invite moment forward rather than manufacturing a reward. Nobody tried to fabricate activity for an account that had none.

The most common misreading is that adding randomness is the whole job, since "variable" sounds like it means chance. The force comes from the contents, though, not from the chance wrapped around them. Whether a document collected three comments or none matters to the person; where a spinning wheel landed does not. If a work tool ships a login reward box and the open rate goes from 60 percent in week one to 9 percent in week three, that table is a measure of how fast contentless variation gets ignored.

The other one is that bigger rewards work harder, since size and effect feel proportional. Large rewards move the baseline instead. Receive a big one now and the same size next time is no longer satisfying, while anything smaller registers as a loss. A product handing out ten badges in week one and one badge in week two has made week two read worse than week one. What this component manages is not size but keeping the next visit unknown, which is cheaper to sustain than an escalating series of prizes.