They only remember what was about them

Two onboardings can show the identical feature and land in completely different places. The first one narrates: here’s the dashboard, here’s the filter, here’s where the reports live. The second one asks a question — what are you trying to get done this week? — and then shows the same feature answering that specific thing. A day later, the first onboarding has evaporated. The second one is still there. The difference isn’t that the second one was cleverer. It’s that it was encoded against a structure the brain never lets go of: the self.

There’s a clean old finding under this. In the 1970s, Rogers, Kuiper and Kirker gave people words to judge under different instructions — is this word in capital letters, does it mean the same as another word, does this word describe you? When they later asked which words people remembered, the self-judged words won, and not by a little. Deciding whether “honest” or “impatient” fit them personally made those words far stickier than deciding what they meant. The self, it turns out, is the most elaborated, most densely cross-linked structure you own. Anything you route through it gets hooked into a thousand existing threads, and things with many hooks are things you can find again later. Brain-imaging work has since tied this self-referential judging to the cortical midline — the medial prefrontal region that lights up when you think about you — though the exact machinery is still argued, so hold the mechanism loosely and the behavioral effect firmly.

Now look at how most onboarding is actually written. It’s a tour. It describes the product — which is to say, it’s a set of facts about you, the builder, and the thing you made. For the user, that’s shallow, semantic encoding of someone else’s world. It competes with everything else fighting for a slot in a stranger’s memory, and it mostly loses. The self-reference effect says the cheapest upgrade available to you is to change the subject of the sentence. Stop making the first run about the product. Make it about them — their goal, their data, their word for the problem — and then let the product show up as the answer to a question they just posed.

The payoff compounds past memory. When the user types their own goal instead of reading yours, you also pick up the generation effect — we remember what we produce better than what we’re shown — and you’ve quietly planted the first seed of endowment: their words, their target, their data are now inside the thing, and people overvalue what has a piece of them in it. One question does three jobs. It makes the feature memorable, it makes it theirs, and it pre-attaches the feature to a real intention instead of a hypothetical one.

Two honest cautions, because this is easy to cargo-cult. First, Hi {{FirstName}} is not the self-reference effect. Token-swapping a name into a template is a trick the brain discounts almost instantly, because no self-processing happened — the user didn’t judge anything about themselves, they just saw their name pasted in. The effect needs real processing: a choice, a rating, a sentence they had to compose about their own situation. Second, there’s a line where “about you” curdles into “we’ve been watching you.” Reflecting back something the user told you feels like being understood; reflecting back something you inferred about them feels like surveillance. Stay on the side of the things they handed you.

The rule of thumb is small and it survives most of your design decisions: before you explain a feature, find the question about the user that the feature answers, and ask that first.

The science, to look up: the self-reference effect — Rogers, Kuiper & Kirker (1977); the Symons & Johnson (1997) meta-analysis; self-referential processing and the cortical midline / medial prefrontal cortex (Northoff and colleagues). Related: the generation effect (Slamecka & Graf, 1978).