The hook is a promise; retention is the invoice

A hook is not a device for getting attention. It is a promise about what the next ten minutes of someone's life will feel like, and the first session is where the bill for that promise arrives. Most studios pay the creative team on the promise and blame the product team for the invoice, which is why the two rarely sit in the same review meeting.

The claim I want to defend goes further than the usual plea for creative and product to talk more. Score a hook on how quickly and how completely the game pays it, and a hook that wins on click-through while deferring its payoff stops looking like a good hook with a retention problem. It looks like a loan taken out against a cohort you have not met yet. Plenty of experienced buyers will disagree, because a cheap install is a cheap install and the network will find payers regardless. That argument used to be better than it is now.

Why the auction rewards the promise, not the payment

Ad networks optimise on the signals they can see quickly. Even on campaigns bidding to a purchase or retention event, the model learns first from installs and early events, because those arrive in volume while the deeper events trickle in. A hook that produces a high response rate gets more impressions, a lower effective CPI, and a faster learning phase. Nothing in that loop asks whether the game deserved the response.

So the auction carries a structural bias toward hooks that overstate. Not lie, necessarily, but overstate: the puzzle that looks easier than it is, the base-building game dressed up as a one-tap defence sequence, the narrative choice that never appears in the actual game. The more efficiently a hook misdirects, the more the delivery system rewards it, right up until the retention curve catches up.

The mismatch does not show up in CPI. It shows up in D1 and D3, in a place where the person reading the number is usually a product manager rather than the creative strategist who wrote the promise. There is the incentive problem in one sentence: the studio books the cost of the promise to one department and the revenue from it to another.

The second-order effect nobody prices

There is a subtler cost. When a misdirecting hook scales, the network builds its audience model around people who responded to the misdirection. It learns that your game suits people who liked what you showed, not what you made. Every subsequent creative, the honest ones included, then reaches an audience the model already shaped around the wrong promise.

This is why teams sometimes find that a truthful creative underperforms in a campaign where a misleading one has run for months. The honest ad is not weak. The algorithm is showing it to a cohort it assembled for someone else. Resetting that takes a new campaign, a new event schema and weeks of exploration budget. The loan was cheaper to take than to repay.

A promise ledger for the brief

The practical fix is to write the promise down before anyone builds the hook, and to write down where the game pays it.

For each concept in the brief, add three lines. Write the promise in one plain sentence a player would actually recognise, not a positioning line. You will sort chaos into order and it will feel effortless. You will make a choice that changes the story. Next to it, name the payoff moment: the exact point in the first session where the player experiences that sentence for real. Measure the gap in seconds or taps from first launch to that moment.

A decision rule follows from that last number.

If the payoff lands inside the first ninety seconds, before any tutorial gate, scale on CPI with a light D1 check. The invoice clears on the spot.

If the payoff lands after the tutorial, gate scaling on D1 relative to the campaign baseline. The promise is real but deferred, and deferred promises leak players at every screen in between.

If there is no payoff moment, or the honest answer is that it arrives somewhere around level 40, do not run the concept on CPI at all. Price it on D7 ROAS from the first test and accept the higher CPI it will show. The concept may still pay, but at least you will know what you are buying.

A worked illustration

Two hooks for the same merge game, with numbers that are purely illustrative.

Hook A shows the actual merge board with a satisfying five-item chain. Promise: order from clutter, immediately. Payoff: the first merge in the tutorial, roughly twenty seconds in. Suppose it tests at a CPI of 2.40 and a D1 of 42 percent.

Hook B shows a fail-state minigame, a bar you fill by tapping before a timer runs out, that the game does not contain. Promise: tense, fast reflexes. Payoff: none. Suppose it tests at a CPI of 1.60 and a D1 of 26 percent.

On the CPI dashboard, Hook B wins by a third. Cost per D1-retained user is 6.15 for B against 5.71 for A, so B is still close, because the CPI advantage nearly covers the retention hit. This is the point where the cheap-install camp usually declares victory.

Run the ledger one step further. The retained cohort from B came in on a reflex promise and is now playing a merge game. If its D7 is 9 percent against A's 19 percent, cost per D7-retained user is 17.78 for B and 12.63 for A. The invoice arrived a week late, and it was larger than the discount.

The illustration is fictional. The shape is not. The mismatch tax compounds along the retention curve because every day gives the player another chance to notice the promise went unkept, while the discount on CPI arrives only once.

Where to put the sentence

The promise sentence belongs in two places, not one. It goes in the creative brief, so the strategist answers for it. It also goes in the FTUE ticket, so the product team knows what the ad has committed them to and can either pay it faster or push back before the ad ships.

Once both teams look at the same sentence, the argument about whether a hook misleads becomes a much duller and more useful conversation about seconds to payoff. That is a number you can move.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• From ad promise to first session: the funnel is collapsing
• The benchmark report is selling you something