The case against the discovery phase

The discovery phase, in most agency engagements, is a polite way of saying "we're not ready to build, and we don't want to admit it."

Real discovery (figuring out what users want, what's technically feasible, what's worth building) happens in code, not in a Notion doc. The fastest way to learn what should be built is to build something and put it in front of someone. A prototype made in three days reveals more than three weeks of user interviews. The user can't tell you what they want; they can only tell you whether the thing in front of them is wrong.

We don't run discovery sprints. We start by shipping something: a clickable prototype, deployed to a real URL, week one. It will be wrong in places. That's the point. The wrong version surfaces the right version faster than any amount of upfront thinking.

This isn't a rejection of strategy. It's a rejection of the idea that strategy is a separate activity that happens before building. The two are the same thing, separated by an artificial wall that exists because old agency contracts billed for "strategy" as a separate line item. The wall isn't useful any more. Strategy lives in the small decisions you make in code: this navigation, that hierarchy, this default. You can't reason about those decisions abstractly; you reason about them by looking at them on a screen.

There are projects where real upfront research is necessary. Medical devices. Regulated industries. Anything where the cost of being wrong in production is catastrophic. For everything else (which is most early-stage products), the prototype is the research.

We get asked sometimes whether we're skipping the "thinking" step. We're not. We're doing the thinking with our hands, in the medium the product will eventually live in. That tends to surface the real questions faster than asking them in the abstract.

What replaces discovery, in our process: a 90-minute conversation in week zero, a written one-pager that says what the product is and is not, and a clickable prototype seven days later. The one-pager keeps us honest about scope. The prototype keeps us honest about feasibility. Together they do everything a discovery phase pretends to do, in a week.