SHORT ANSWER
Prioritise against one metric for the quarter, not against a scored list. Ask of each candidate: what evidence says this moves the metric, what is the smallest version that would test it, and what comes off the roadmap to make room. Scoring frameworks like RICE are useful for structuring a conversation, but they launder opinion into numbers if the inputs are guesses.
Answered by Andrew Crossley, Fractional Chief Product Officer · Updated 2026-08-01
Prioritisation is where product strategy either becomes real or evaporates. Most roadmaps are a record of who lobbied hardest.
The output of good prioritisation is a shorter list. If nothing was removed, no prioritisation happened.
Without one metric, prioritisation is a preference argument with a spreadsheet attached.
Customer conversations, data, support reason codes, failed workarounds. 'A customer asked' is a data point, not a case.
For each candidate, define the cheapest version that produces a real signal. Ideas that cannot be reduced usually are not understood yet.
Adding requires removing. A visible parked list makes the trade explicit rather than silent.
Pre-PMF, prefer items that resolve the biggest uncertainty. Post-PMF, prefer compounding leverage.
Weekly reprioritisation destroys throughput and morale.
FROM EXPERIENCE
The most valuable prioritisation decisions in my career have been removals: features with internal sponsors and plausible logic that would have consumed a quarter without moving the number that mattered.
Roadmaps rarely fail because a good idea was missed. They fail because too many reasonable ideas were accepted at once.
As a discussion structure, yes. As a decision machine, no — the inputs are usually estimates dressed as data.
Aggregate them by pattern and revenue at risk. One deal is an anecdote; six deals with the same blocker is a roadmap item.
One primary initiative per team, at most one secondary. Anything more and cycle time collapses.
IN SHORT
Rank AI features on value, evidence, cost to serve and eval difficulty.
Read moreStrategy, roadmap and operating model that survive contact with reality.
Read moreThe seven-stage framework that takes an idea to first revenue.
Read moreSeven stages from first idea to first revenue, with one output each.
Explore the Crossley MethodTHE FRAMEWORK
MORE ANSWERS
You become a product manager without a degree by producing evidence instead of credentials: ship something real, own a metric in an adjacent role, and document decisions publicly. Support, sales, operations and QA are the highest-converting entry routes because they give you customer contact and data. Hiring managers screen for judgement and shipped outcomes; almost none check for a degree at interview stage.
Four skills carry most of the job: customer discovery, prioritisation under uncertainty, written communication, and data literacy. In 2026 add two more: AI-assisted execution, and evaluation design for AI features. Frameworks, roadmapping tools and ceremonies are teachable in weeks; judgement about what not to build is the skill that separates senior PMs from everyone else.
Product managers usually fail for structural reasons, not talent ones: no clear mandate, no owned metric, no direct customer contact, an organisation that rewards shipping over outcomes, and an unwillingness to create conflict by saying no. Four of those five are fixable by the company. The fifth — avoiding conflict — is the personal skill most often missing in PMs who stall at mid-level.
Product managers use AI to compress the artefact half of the job: synthesising interviews, drafting specs and tickets, building clickable prototypes, summarising support and review data, and pressure-testing decisions. The best PMs also use it inside the product, designing evaluation sets and quality metrics for AI features. What they do not delegate is the decision, the customer conversation or the accountability.