A phase gate is a promise that the next stage will be better informed than this one. In practice it is a promise that the next stage will inherit decisions nobody can now explain.
We ran gated projects for three years. Discovery, then design, then build, each ending in a document and a sign off. The documents were good. The projects were slow, and the slowness was never in the work — it was in the handovers.
What a gate actually costs
Every gate is a translation. Research becomes a deck, the deck becomes a design, the design becomes a ticket. Each translation loses the reasoning and keeps the conclusion, so by the time someone is building, they are implementing decisions whose context has evaporated.
The expensive part is not the translation. It is that a gate makes the decision expensive to revisit. Once discovery is signed off, discovering something new in week six is a change request rather than a Tuesday.
A gate does not stop you learning late. It only makes learning late expensive.
The loop instead
One cycle, run weekly, with the same people in it from the first day to the last: decide what question matters most this week, build the smallest thing that answers it, put it in front of someone, decide again.
Discovery does not end. It gets thinner as the questions get smaller. By week six we are still learning, but the things we learn cost an afternoon instead of a renegotiation.
What we kept
Two artefacts survived the change, because they earn their keep:
- A decision log. Not a deck — a running list of what we chose, when, and what would make us choose differently. It is the only document anyone reads twice.
- A cut list. Everything we deliberately did not build, with the reason. It becomes the roadmap at handover, and it is consistently the most valued thing we deliver.
The deck was never the deliverable. The working thing was. Removing the gates just stopped us pretending otherwise.
