ALPHA KIT

Quick Start: one feature, three slices, 20 minutes

Apply SLICE to one real piece of work in five steps. If it takes more than 20 minutes, tell us: that is a finding we want. Use the three templates.

1. Pick one roadmap item (3 minutes). Something you plan to build in the next two weeks, for example "user accounts". Classify it: a Bet if its value is uncertain, an Obligation if it is required (security, legal, a contract), an Improvement if the benefit is clear and small.
2. For a Bet, write the decision criterion (3 minutes). One hypothesis, what you believe today and why, and a criterion you can count at your volume: "continue if at least 5 of the first 20 users do X" beats "conversion rises 2%". Set a review date.
3. Cut three slices (5 minutes). Each slice is the smallest piece someone can check on its own, for example "email and password sign-in", "password reset", "sign-in with Google". Give each a planning mode: 0 if obvious, 1 if bounded, E if a question about value or price blocks it, 2 for architecture, 3 for money, logins or sensitive data.
4. Estimate each slice (6 minutes). Use the calculator's quick estimate: how long building takes with AI, how risky it is, what is still undecided, and who you are waiting on. Copy the label it gives you, such as "Email sign-in · Target · F1 · 3.5–9.5 working days", onto a Slice card together with the acceptance evidence.
5. Name your bottleneck and your weekly check (3 minutes). Where does work wait longest today? Assume review if you don't know. Start no new slice while more than about a week of work waits there. Put a 20-minute weekly check in your calendar.

From then on, write down the day each slice starts and the day it is accepted. Most of this can come from your issue tracker or Git history. After seven similar slices, the calculator's full forecast can give you a measured range instead of a judgment.

Do not convert sizes into hours, multiply an old estimate by an "AI factor", or quote percentiles before you have measured history.