Roadmap Prioritization Coach

Prioritization fails socially before it fails analytically: the list is secretly three lists, scores get tuned until the pet project wins, and "no" never gets written down. This skill runs a forced-ranking pass that makes the tradeoffs explicit — anchored scoring, a capacity cutline, and a written not-doing list — so the argument happens once, in the open, instead of forever, in hallways.

When to use this skill

• Quarterly or cycle planning when candidates outnumber capacity three to one
• A stakeholder escalation — "why isn't X on the roadmap?" — deserves an answer better than vibes
• Two teams share a dependency and their priorities silently conflict
• The roadmap is a wish list with no cutline, and delivery keeps "slipping" work that was never actually resourced

Workflow

1. Get everything into one list. Features, debt paydown, research spikes, compliance work: one list, one owner per item, one sentence per item stating the claimed benefit and for whom. Separate lists are how debt and research quietly starve.
2. Define the scoring dimensions and anchor them. Reach: how many users or accounts this touches per quarter. Impact on the target metric, anchored — a three moves the dashboard visibly, a one is plausible but indirect. Confidence: an evidence-based discount, where tested beats researched beats informed guess. Cost: team-weeks, estimated by the people who would build it and nobody else.
3. Score in one sitting, out loud. The debate the scoring provokes is the actual product; the numbers are a forcing function. Record disagreements on scores rather than averaging them into silence.
4. Compute, sort — then treat the ranking as a witness, not a judge. Ask where the sorted list contradicts strategy, double-funds one bet, or starves a strategic theme. Overrides are allowed and healthy; each one gets its reasoning written next to the item it moved.
5. Draw the capacity cutline. Sum team-weeks from the top until the cycle's capacity is consumed — after subtracting the standing tax of support, bugs, and operations, which is rarely under 20 percent when measured. Below the line is not "lower priority"; it is not happening this cycle, and the plan should say so in those words.
6. Write the not-doing list. For each notable item below the line: what it was, why it lost, and what evidence would change the call. Publish it with the roadmap — it is the half that prevents relitigation.
7. Schedule the re-run. Priorities are a cache of current evidence. Set the invalidation date — next cycle, or a named event like a launch or a funding change — rather than waiting for the cache to fail loudly in a meeting.

Output format

Four artifacts: the scored table (item, owner, reach, impact, confidence, cost, score, override note); the funded list above the cutline with the capacity arithmetic shown; the not-doing list with reasons; and a changelog from last cycle — what moved and which evidence moved it.

Guardrails

• Scores are never edited after the sort without a written override note; silent retuning is how the exercise dies
• Cost estimates come from builders; benefit claims arrive with evidence or take the confidence discount
• If everything scores high, the anchors are broken — recalibrate against the two most extreme items and rescore
• Debt and research compete in the same list under the same rules, or they will never win a slot
• A roadmap without a not-doing list is a wish list wearing a lanyard

Worked example: one scored row

"Bulk import from spreadsheets" — owner: growth squad. Reach: 4 (touches every new account's first week). Impact: 3 (activation is the target metric; import failure is the top drop-off). Confidence: 0.8 (funnel data plus twelve support threads). Cost: 6 team-weeks. Score: (4 x 3 x 0.8) / 6 = 1.6. Override note: none needed — it cleared the cutline on arithmetic alone.