Hook Model

For a long time the answer to "why aren't people coming back" was a list of features. A retention report lands, a meeting gets scheduled, and proposals arrive in sequence: send notifications, hand out points, redesign the home screen. None of them is wrong on its own, and nobody could explain how any of them connected to the others.

So the same meeting reconvened every three months. Last quarter we added notifications, the number rose briefly and then settled back, so this quarter let's try something else. The ideas were always new, but since nobody wrote down why the previous attempt failed, a failed idea would often reappear six months later under a different name.

The Hook Model lays a sequence over that room. It treats return visits not as one decision but as a loop of four slots, and it asks a team to write down what each of those four slots currently contains in their own product. Empty slots become visible, and it becomes obvious that "add notifications" and "redesign the home screen" are proposals aimed at completely different parts of the loop.

Start by writing down the loop already turning in your product. Fill the four slots before generating new ideas, then look for the one that's blank.

1. Write one full turn. In a collaborative document tool: a notification says a colleague commented (trigger), you tap in and reply (action), an emoji lands on your reply within seconds (reward), and along the way one more document of yours is sitting in the team folder (investment). If those four sentences won't write, there is no loop there yet.
2. Split triggers into external and internal. Notifications and email arrive from outside; remembering before a meeting that the doc needs cleaning up and opening the app yourself comes from inside. Count what share of sessions arrive through a notification and you'll see how much the product leans on external triggers. When that share is high, turning notifications off turns visits off with them.
3. Measure the friction in the action. Literally count the taps between opening the notification and posting the reply. If the session has expired or the app opens on its home screen instead of the document, a perfect trigger dies right here.
4. Check whether the reward varies. The same screen every time gives people less reason to look; three comments today and none yesterday makes them want to check. Variable reward is the name of this slot.
5. Look at whether the investment slot manufactures the next trigger. This is where the model separates itself from other frameworks. More of your documents in the folder raises the odds someone comments on one, and that comment is the next turn's trigger. With nothing accumulating, four filled slots still produce a wheel that turns once and stops.
6. Only ship the proposals that fill an empty slot. If all four are populated and the numbers still won't move, the problem lives outside this model.

It fits products that only pay off with frequent use. Messaging, expense tracking, collaborative documents — anything where opening it three times a week is normal — and filling the four slots overlaps almost entirely with making the product better.

It does not fit products that are meant to be used once or twice a year. Attach a streak to a tax filing tool or a moving-quote service and you manufacture users who open the app to protect a number, producing visits that never turn into revenue while muddying the metrics. An effort to raise return visits in a product like that is usually the downstream symptom of choosing the wrong metric.

Collaboration tools carry an extra limit. The variation in the reward isn't produced by the product; it's produced by colleagues. An active team generates it automatically, and in a quiet team no amount of screen design creates something new to look at. The lever there isn't the reward slot at all — it's getting more people invited into the team.

The heaviest limit is that the model never asks whether the habit is worth having. The four slots fit a useful routine and a compulsion equally well. So one question comes before filling them in: if a user understood this loop completely, would they still want to keep riding it? A no gets billed later as notification fatigue.

"Run this through the hook model and the investment slot is empty. Nothing accumulates the more you use it."

Narrowing a retention problem down to one of the four slots.

"We're living on external triggers. Turn notifications off and weekly visits halve."

A diagnosis that internal triggers haven't formed yet.

"Before we run the hook model, let's check whether anyone has a reason to open this weekly."

Pushing back in case low frequency is simply correct for this product.

At Pagedeck, Ravi brought the model out during a stretch where weekly return visits had been flat for three months. The meeting already had a proposal to add more notification types. When Ravi wrote the four slots on the board, trigger, action, and reward were all populated and investment was blank. People who read a document, left a comment, and closed the tab were leaving nothing of their own behind.

So the change had nothing to do with notifications. The team built a personal workspace that collected recently read documents and let people pin the ones they used often, and the accumulated list became the reason for the next visit. In a quarter where not one new notification type shipped, weekly return visits moved for the first time.

The most common misreading treats the model as a method for sending more notifications. Trigger comes first among the four slots, and notifications are the easiest thing in that slot to change, so the association hardens fast. But a notification is one external trigger, and what the model is actually aiming at is the state where users come back without one. Keep expanding the trigger slot and visits rise for a quarter, after which notification opt-outs and churn rise together.

Believing that filling all four slots produces a habit comes from the same place. The model is drawn cleanly enough that completing it looks like finishing something. In practice the loop only spins on top of value the product already delivers. Bolt four slots onto a product nobody needs and the wheel turns once under force and never comes around again. The order runs the other way: confirm the value first, then use the model on the part where that value has to be delivered repeatedly.