Monday morning, three messages. Sales needs an integration to close a deal before the quarter ends. A large customer has escalated a complaint about a screen that has not changed since launch. The CEO has seen a feature on a competitor's website and wants to know when you will have one. Each request is reasonable. Each one, on its own terms, is urgent. Together they would fill the next three months.
So you open the spreadsheet, score them for reach, impact, confidence and effort, and discover that sales, support and the CEO each weigh those inputs differently. The scoring meeting becomes the argument it was meant to prevent, and the request with the most senior sponsor wins anyway.
Two ways prioritisation fails
Product managers under this kind of pressure tend to swing between two failure modes. The first is prioritising by decibel: whoever shouts loudest, or sits highest, gets their feature. The second is analysis paralysis: more scoring, more data, more meetings, and still no decision anyone will defend. Neither moves the product in a coherent direction.
Both come from the same gap. Requests are hard to push back on when there is no clear direction to push back with. Stakeholders who have not agreed a strategy ask for what makes sense for them now: sales for what closes accounts before the period ends, customer service for the fixes that cut its costs, marketing for something worth a campaign. The product suffers death by a thousand cuts, and the product manager becomes a feature factory worker. In time a line manager steps in to set the direction, with less knowledge of the domain than you have.
A scoring framework does not close that gap. RICE, which Sean McBride described for Intercom, is a sound way to rank items that serve the same goal. It cannot tell you what the goal is. Without one, it launders loudness into numbers.
You can only say no so many times before it becomes career limiting. The strategy has to do the saying for you.
Seven steps, in order
- 01Agree the strategy before you rank anythingUnderstand leadership's strategy, even if it is light, then sell your product strategy to them. Without an agreed destination, every request is judged on who asked and how loudly.
- 02Split development capacity into three bucketsOne for roadmap work, one for technical debt, one for tactical requests from sales, marketing and customer service. Agree the split with the CEO and CTO before the requests arrive.
- 03Rank the roadmap bucket by the strategy aloneEverything in the largest bucket is prioritised strictly by how well it serves the strategy. Who asked for it does not count.
- 04Hand the technical-debt bucket to your most senior engineerBrief them on the strategy, then let them choose what to reinforce, so the parts of the product that will carry the next stage are ready for it.
- 05Let tactical requests compete only with each otherA request from sales competes with other tactical requests, not with the roadmap. That stops you comparing things whose value and time horizons cannot be compared.
- 06Set the mix by the product's stageBefore launch, technical debt should be small and tactical work close to zero. After launch, set roadmap capacity aside for tactical requests.
- 07Say "not right away", and show what goes firstPresent the strategy before the roadmap, every time. Then the answer to a request is rarely "no". It is "not right away": these other items get us there faster.
1. Agree the strategy before you rank anything
The first job comes before any ranking: understand the strategy your leadership is following, and accept that it may be light. Then sell your product strategy to them as the product's answer to theirs. Where the company strategy is thin, you may find yours absorbed into it. Either way, once senior stakeholders have understood and accepted it, you have something to judge every request against that is not your own opinion. If you have no product strategy yet, start there: our guide on how to build a product strategy sets out the steps.
2. Split development capacity into three buckets
An agreed strategy will not remove every distraction on its own, which is why it helps to think in buckets. Take your development capacity and split it three ways, agreeing the shares with the CEO and CTO before the requests arrive, not while they are being argued over.
| Bucket | Share of capacity | Who prioritises, and how | What it holds |
|---|---|---|---|
| Roadmap execution | The largest share | The product manager, strictly by how well each item serves the strategy | The work that moves the product towards its vision |
| Technical debt | Agreed with the CEO or CTO; in Cyril’s experience rarely below 15%, up to 40% for a very distressed product | The most senior engineer, once briefed on the strategy | Reinforcing the parts of the product that will be load bearing in the next stage |
| Tactical requests | What remains; close to zero before launch | The product manager, comparing each request only with the others in this bucket | Asks from sales, marketing and customer service that cannot wait for the roadmap |
The buckets solve the problem the spreadsheet could not: comparing things whose value and time horizons are different in kind. A deal this quarter, a sturdier platform next year and a step towards the vision are no longer fighting over one number.
3. Rank the roadmap bucket by the strategy alone
The largest bucket is roadmap execution, and everything in it is prioritised strictly by how well it serves the strategy. Not by who asked, not by how long it has waited, and not by how easy it is. A request from the CEO that serves the strategy well goes near the top; one that does not, waits, and the reason is one the CEO has already agreed to. Scoring models can help here, because inside this bucket everything is aiming at the same goal.
4. Hand the technical-debt bucket to your most senior engineer
At each development cycle, or at regular intervals, a fixed share of effort goes to technical debt. The product manager probably should not prioritise that work at all. Brief the most senior engineer on where the strategy is taking the product, and let them choose, so they reinforce the parts that will be load bearing in the next stage. How big the share should be depends on the state of the stack. In Cyril's experience it should not fall below 15%, and a very distressed product may need up to 40%.
5. Let tactical requests compete only with each other
The third bucket takes the tactical demands from sales, marketing and customer service. Here, who asked does matter: a fix for a large customer about to leave may reasonably beat a nice-to-have for a prospect. What changes is the competition. A request from sales no longer competes with the roadmap; it competes with other tactical requests, and when the bucket is full, sales can help choose which of its own asks goes first.
6. Set the mix by the product's stage
The shares are not fixed for life; they follow the product's lifecycle. Before launch, you would hope technical debt is minimal, and tactical work is close to zero because there are no live customers asking for anything. After launch, everything changes, and you need to set part of the roadmap aside for tactical requests before they take it by force. Review the split when the product changes stage, not when a stakeholder is unhappy with it.
7. Say "not right away", and show what goes first
Present the vision and strategy before the roadmap, every time, so the discussion that follows is about solving the same problem. If the roadmap is challenged because the strategy is, stop and settle the strategy; there is no point arguing about items. With the strategy accepted, pushing back on an idea stops being "this is a bad idea". It becomes "these other ideas will get us to our destination quicker". You are not saying no. You are saying not right away, because something already planned will have a stronger impact.
Why this matters more now
This was always hard. Now that engineers build with AI, the pace has gone up sharply: requests arrive faster, and more of them can be built before anyone asks whether they should be. "Let us launch and see" sounds cheap when code is cheap, but features nobody uses still need maintaining, still clutter the product and still pull it in several directions at once. Faster building makes the filter more valuable, not less.
AI can also give you time back. Keep asking of each task: is this a job for me, or something an AI could take care of? Grouping a few hundred requests by the problem behind them is a good example. Defending the ranking to your CEO is not.
A worked example: from strategy to a prioritised roadmap in Scaffold
Scaffold walks product managers through the research and analysis a strong product decision needs, on their own product, and each activity declares what it builds on. For prioritisation, three of them matter most:
- Stakeholder Map (Power/Interest) maps stakeholders by power and interest to guide how you engage each one. Its sections are Power / Interest Grid, Engagement Plan, Engagement Strategy.
- Strategic Options lays out three to five mutually exclusive strategic directions, compares them on the same evidence, and records the one you choose. Among its sections are "What This Rules Out" and "Reopen This Decision If", so the direction the roadmap bucket is ranked against is written down, along with what it excludes.
- Roadmap generates a prioritised product roadmap, returned as a Roadmap Timeline and a Priority Board. It builds on 9 earlier activities, Strategic Options and the Strategy Narrative ("The Trip") among them, so the ranking draws on the strategy work done before it.
Two limits, said plainly. Scaffold owns the judgement, never the backlog: requests, tickets and sprints stay in whichever tools your team already uses, and the buckets are yours to agree with your CEO and CTO. And it does not make the call for you. You are the one who has to explain the ranking, so you have to believe it.
When this is not your call
In a very early startup whose CEO runs the strategy hands-on, the product manager's job is to execute that strategy faithfully, not to set up a competing one. The buckets still help, because they make the trade-offs visible to the CEO. But the strategy that ranks the roadmap bucket is the CEO's.
In short
- Prioritising by decibel and analysis paralysis share a cause: no agreed strategy.
- Understand leadership's strategy, then sell yours to them.
- Split capacity into roadmap, technical debt and tactical buckets, agreed in advance.
- Rank the roadmap by the strategy alone; let the senior engineer rank the debt; let tactical requests compete with each other.
- Shift the mix as the product moves from pre-launch to live customers.
- Present the strategy before every roadmap, and say "not right away" instead of no.
Look at the last three requests you turned down. Could you have said which bucket each belonged in, and what it would have displaced? If not, agree the buckets before the next one arrives.