Consumer Apps: Everything Is Decided at the Second Use

The consumer software graveyard is not full of bad products. It is full of competent products that people installed, opened once, and never opened again.

That is the actual failure point, and it is not where attention goes. Enormous effort goes into acquisition — getting the download, the sign-up, the first session. Comparatively little goes into the step immediately after, which is where almost everything is decided.

The question is not whether someone will try your product. Trying is cheap. The question is whether anything happens in that first session that makes returning feel natural rather than effortful.

Why the second use is the hard one

A person opening something for the first time has no habit attached to it, no accumulated value inside it, and no memory of it being useful. Every subsequent use has to be chosen, against competing options that already have all three.

Two things reliably produce a second visit.

It attached to something the person already does. Not a new routine — an existing one. A meal, a commute, a weekly payment, a conversation with a specific person. Products that slot into an existing rhythm inherit that rhythm's regularity. Products that require someone to construct a new habit from nothing are asking for something people are historically bad at.

Value accumulated inside it. Data, history, configuration, connections. After a few uses, leaving means losing something. This is a switching cost built by the user rather than imposed on them, and it is why the fifth session is far more predictive than the first.

Both are visible early if you measure the right thing. Downloads measure marketing spend. Second-week return measures whether the product solved something. The gap between those two numbers is where most product post-mortems should start and rarely do.

The distribution problem

There is a second structural difficulty that has nothing to do with the product.

Getting a consumer product in front of people has become expensive and concentrated. Most discovery flows through a small number of app stores, search systems, and feeds, each of which sets the terms. Paid acquisition prices are set by whoever is willing to pay most for the same attention, which means a well-funded competitor can raise the cost of your customers without doing anything about your product.

This is the aggregation position seen from below: the platform owns the relationship with demand, and everyone building on top competes for access on the platform's terms.

The consequences are unpleasant and worth stating plainly.

Paid acquisition rarely builds a business alone. If each customer costs more to acquire than they eventually contribute, scale makes the problem worse rather than better. A great many consumer companies have grown quickly on this basis and stopped abruptly.

The realistic paths are narrow. Something inherently shareable, so users bring other users. Something with a genuine network effect, where existing users make it better. Or an unusually low cost of serving each user, so thin monetisation works.

Being a feature is a real risk. A product doing one useful thing well is exposed to a platform absorbing that thing, at which point distribution disappears entirely.

The metric problem

There is a persistent tension in this industry that deserves stating directly rather than as an aside.

The things that are easy to measure — sessions, time spent, notifications acted on — are not the same as whether the product improved anyone's day. Time spent is measurable and improvable. Whether the time was well spent is neither.

A team optimising the measurable number for long enough will find techniques that increase it without increasing value: intermittent rewards, artificial urgency, notifications engineered around uncertainty, friction placed deliberately in the exit path. None of this requires anyone to intend harm. It is Goodhart's Law with a product team attached, and it happens by default unless something actively counteracts it.

The counterweights that work are unglamorous: measuring long-run retention rather than short-run engagement, since manipulative techniques tend to produce a spike followed by abandonment; asking directly whether users are glad they used it, which is imprecise and better than nothing; and treating uninstalls and cancellations as information rather than as leakage to be reduced by making cancellation harder.

The honest position is that the incentives point the wrong way and the counterweights are weak, which is why the same patterns keep appearing across unrelated companies. It is a structural feature of monetising attention, not a coincidence of who happens to be building these products.