How we prioritize

Write this down once and the roadmap argument gets much shorter, because you stop re-litigating the method every time you disagree about an item.

The decision this makes

Which candidates enter Now, and in what order. Everything else follows.

The four inputs

Impact — how much this moves the business if it works. High / Medium / Low, and be willing to say Low out loud.

Effort — S / M / L / XL. Deliberately rough. A precise estimate on an unscoped idea is a false-confidence machine.

Customer requests — how many customers actually asked, counted, not remembered. This is the input that stops the loudest voice in the room from being the input.

Source — where it came from. A strategy bet and a support-driven fix are both legitimate, but they should not be compared on the same axis without saying which is which.

What we do not do

We do not roll these into a single score. A number that says "7.4" hides the disagreement instead of resolving it, and the disagreement is the useful part. Sort by Impact, then scan Effort and Customer requests next to it. If two people still disagree, they now disagree about something specific.

Watch for the classic mismatch: the most-requested item is often not the most valuable one. Dark mode shipped in v1.9 with 47 requests and Low impact. That was still the right call, because it was cheap and it removed a recurring complaint, but nobody should have pretended it was a growth bet.

Three questions before anything enters Now

1. Is it Scoped? If someone still has to decide what done means, it is not ready to be committed, whatever the date says.
2. Does it fit before code freeze? A feature that lands the week of GA is a v2.1 feature that is pretending.
3. Who is the single owner? Not a team. A person.

How we say no

Not by leaving things in Idea forever. Move it to Later / Backlog with a one-line reason, or close it. A backlog full of things that will never happen makes the ones that will harder to see.

If it was promised to a named customer, it lives in the Customer-committed sublist, and deprioritising it is a conversation with that customer rather than a quiet edit.

Cadence

• Weekly: triage the Idea intake section. Every item gets fields and a horizon, or gets closed.
• Per release: at feature freeze, anything not Scoped moves out. No exceptions, because the exception is how the freeze stops meaning anything.
• Per release: after GA, look at what shipped against what was in Now on day one. The gap is the number that improves your next forecast.