Four steps from the first observation to a company that manages without us. One build has already gone the whole way.
- 01
Explore
What grates, for whom, and how often? Most of it falls at this stage.
- 02
Define
We turn the observation into a claim that can be wrong. The most dangerous assumption gets tested first.
- 03
Build
Something that runs, not a presentation about what it could be. It is allowed to break.
- 04
Launch
Its own name, its own site, its own way forward. The goal is that it manages without us.
When an experiment goes from play to serious, meaning customers, revenue and outside interest, it becomes its own company. The lab stays lean and gets to break. What holds up gets a life of its own.
The cards say what the steps are. This is what they mean when you are sitting in the work.
The steps rarely run neatly in a row. You discover mid-build that the definition was wrong and have to go back, and that is the point. The order exists so you know which step you are in, and therefore which question matters right now.
explore
What does it cost them today? A problem that annoys one person once a quarter is not worth building away, however real it is.
define
Which assumptions does the claim rest on, and which is the most dangerous? This is where the work narrows, from a broad opportunity to a question a build can answer.
build
Something you can use, and therefore be disappointed by. It goes fast, and a mistake is cheaper in the lab than after a launch.
launch
The step is taken once something has shown that it holds, not before. From that point nothing is built that needs the lab to survive.
The simplest way to show the process is to follow something that went all the way.
The observation came out of the work: technical SEO is stuck in the code. Changing a title, a meta description or the structured data usually takes a developer, a release and a moment of risk. On a site with thousands of pages that becomes a bottleneck, and most improvements never happen.
The definition became a claim that could be wrong: what search engines read ought to be changeable without rebuilding the site. If that held, the cost was not in the change itself but in everything around it.
The build became a layer that answers at the edge, before the request reaches the server, shaping title, description and structured data directly in the response. The site underneath does not have to do anything extra.
The launch came when it held. Dynamic SEO got its own name, its own site and its own company, and is in private beta today. The same four steps as on the cards above, this time with the answer key.
Groteket has walked the same four steps and stays in the lab as a build in production. It is in its own case.
A lab that keeps everything eventually becomes a warehouse of half-finished things nobody maintains. So the spinout is built into the model, not decided after the fact.
When an experiment goes from play to serious, what holds up is separated from what is still being tested: the risk stays in the lab, the asset moves out. That is what assets up, risk down means in practice.
It also makes the lab more honest. When success moves out anyway, there is no reason to keep something alive just because it has already cost money.
Seen something that grates? That is step one.