The traditional plan has phases: discovery, design, build, test, launch. It reads well in a proposal because it looks like control. Its defect is structural — you find out whether the thing works at the very end, when there is no budget or time left to act on the answer.
What a slice is
A slice is a thin cut through the whole system that delivers one complete, usable piece of value. Not “the database layer” or “the design system” — those are layers, and a layer on its own does nothing for anyone. A slice for a booking product might be: one course, one date, one real payment, one confirmation email. Narrow, unglamorous, and end-to-end.
Because it’s end-to-end, a slice exercises every risky assumption at once: the integration works or it doesn’t, the payment provider behaves or it doesn’t, the flow makes sense to a real person or it doesn’t. Two weeks in, you know.
What phases hide
Phased plans front-load the comfortable work and postpone every unknown. Six weeks of design and eight of build can pass with everything apparently on track, because “on track” measures activity, not working software. The unknowns are still all there, waiting, in the phase with the least slack.
They also make change expensive by design. When feedback arrives after design is signed off and build is underway, acting on it means unwinding decisions instead of adjusting the next slice. The plan starts to defend itself against what you’ve learned.
What changes for you
- You see working software every couple of weeks. Not a status document — a URL you can use.
- Priorities stay live. Each slice is chosen from what matters now, informed by the last one.
- You can stop. If four slices in you have what you need, you stop. That option has real value and phased plans don’t offer it.
- Bad news arrives early. Uncomfortable, and far cheaper than bad news arriving at launch.
What it asks of you
This is the part usually left out of the sales pitch. Slices need a decision-maker available roughly every two weeks — someone who can look at what shipped and say what’s next. If reviews slip by a month, the method degrades into a phased project with extra meetings.
It also means living with visible incompleteness for a while. The product is real and live and obviously partial for most of the project. Teams who need everything hidden until it’s perfect find this uncomfortable, and that’s worth knowing about yourself upfront.
How we sequence the slices
Riskiest thing first, wherever possible. If a project’s success depends on an integration with a system nobody has tested, that’s slice one — even though it’s less satisfying to look at than a homepage. The purpose of slice one is not to impress; it’s to remove the largest piece of uncertainty while there’s still time and money to respond to what it reveals.
The order is a conversation, not a decree. But if a plan puts the scariest unknown last, ask why.
