What's wrong with the way startups do discovery

She also had eight weeks less runway, and a team that had spent six weeks watching other people think about their product instead of building it.

This is not a story about a bad agency. The agency did exactly what it had been trained to do, by an industry that has confused process with progress. The discovery phase, as practised by most product agencies and most in-house product teams, is one of the most expensive forms of theatre in modern software.

Here is what is wrong with it.

Discovery has become an end in itself

The original idea behind a discovery phase was reasonable. Before you build something, understand the problem. Talk to users. Look at competitors. Don’t waste engineering time on the wrong thing.

Somewhere along the way, this got industrialised. Discovery became a deliverable. Then a methodology. Then a fixed-price six-week engagement with its own templates, its own facilitation cards, and its own Lucidchart conventions. The artefacts multiplied — empathy maps, opportunity solution trees, jobs-to-be-done canvases, value proposition canvases, lean canvases, business model canvases, canvases for the canvases. Each one promising to make decisions easier. Each one making the next decision a little harder.

A real discovery happens in the first two weeks of a project, talking to five users, looking at three competitors, and writing down what you learned. Anything longer is usually not discovery. It is the team avoiding the moment where they have to commit to a decision.

The output is almost never used

Pull up the deck from a discovery phase six months later and watch what happens. The personas are forgotten. The journey map has been overtaken by what users actually do. The “How might we” questions have been answered by the build, or rendered irrelevant by it. The opportunity tree has one branch left, because that’s the one the team chose to pursue.

The deliverables of discovery have a half-life of about eight weeks. After that, they are documentation of what people thought before they knew anything. The only thing that survives is the decisions the discovery led to, and those rarely needed 96 pages to support them. They needed a one-page memo and a willingness to be wrong.

Talking to users is not the same as understanding them

Most discovery phases involve user interviews. This sounds like a good idea. It is, in principle. In practice, the interviews tend to be done by people who are not the people building the product, written up by other people who are also not building it, and presented to the team in summary form weeks later.

By the time the insight reaches the engineer or designer, it has been translated three times. It has lost its texture. It has been smoothed into bullet points. The actual signal — the moment a user did something the team didn’t expect — is gone.

The teams who really understand their users are the ones where the people building the product talk to users themselves, every two weeks, forever. Not in a research sprint. Not in a discovery phase. As an ongoing habit, woven into the build. Discovery is not a phase. It is a practice.

Six weeks of discovery is six weeks of not deciding

Here is the thing nobody says out loud. Most discovery phases are not gathering information. They are deferring commitment.

A founder who hires an agency for six weeks of discovery has, in many cases, already made the decision they need to make. They know what to build first. They have known for weeks. The discovery phase is permission to feel certain. It is the act of getting other people to agree before you commit, so that if it goes wrong, the responsibility is shared.

This is human and understandable. Building things is frightening. Committing to a direction is an act of nerve. The discovery phase is a way of borrowing nerve from the process.

But the cost is enormous. Six weeks of runway. The momentum lost in a team that hasn’t shipped anything. The competitor who used those six weeks to ship a v1 and start learning. The investor who watches the burn and wonders what the team has been doing. None of that is recovered when the deck is delivered.

What discovery should actually look like

A good discovery is short, opinionated, and run by the people who will build the thing.

It looks like this: two weeks. Five user conversations done by the founders themselves, not by a research consultant. A competitive teardown that fits on two sides of A4. A clear point of view about what to build first, written as a memo, defended in a ninety-minute readout, decided in the room. By the end, the team has a direction, a scope, and the names of the three things they are deliberately not building.

It does not produce a 96-page deck. It does not produce twelve personas. It does not produce a journey map with 47 pain points. It produces a decision, and the team’s permission to commit to it.

Anything longer is usually the team buying time. Anything shorter is usually skipping the conversations that matter.

Why agencies sell long discoveries

Because they are easy to sell. A six-week discovery is a clean piece of scope. It has a fixed price. It has tangible deliverables. It does not require the agency to commit to building the actual product, which is the part where things get hard.

A two-week discovery, followed by a commitment to build, is harder to sell because it requires the agency to be confident in their own delivery. It requires them to skin some game. It requires them to accept that the discovery is a service to the build, not a product in itself.

This is also why so many discovery phases never lead to a build. The agency has been paid. The deck has been delivered. The relationship has done its job. The founder, six weeks poorer and no closer to a shipped product, goes looking for someone who will actually build it.

The point

Discovery is real work. Understanding users is real work. Looking at competitors is real work. Picking a direction is real work. None of it requires six weeks of process theatre.

If you are about to commission a discovery phase, ask three questions:

Does it fit on two sides of A4? If not, it is too long.

Are the people doing the discovery the same people who will build the thing? If not, the insight will not survive the handoff.

Does it end with a decision the team is committed to? If not, it has not done its job.

Most discovery phases, judged honestly against those three questions, do not pass.

The good news is that the work itself is not the problem. Founders who skip discovery entirely build the wrong thing. Founders who buy six weeks of process buy a deck. Founders who do two weeks of focused work, talk to real users themselves, and commit to a direction — they ship.

That’s the bit nobody is selling.