How to Decide What to Design First: A Prioritisation Framework
A discovery ends with thirty findings on a wall. All of them are real, most of them are evidenced, and nobody in the room can agree which three the team should work on next. What settles it is usually not the evidence. It is whoever spoke most forcefully, or whichever finding happens to map onto a ticket somebody already knows how to write.
Why prioritisation quietly defaults to the loudest voice
Research produces a flat list. A finding about a form that loses people at the last step, a finding about a letter nobody understands, a finding about a handover between two teams that fails roughly once a week: on a wall they all look the same size. Nothing in the artefact says which one matters more, and the team has no shared language for comparing them, so the comparison happens implicitly and privately inside each person's head.
Two forces then take over. Vividness beats volume, because one distressing story told well in a workshop outweighs a dull problem that most of the caseload meets every day. Tractability beats importance, because the finding that already resembles a piece of work gets picked up, while the one that would require a difficult conversation with another department stays on the wall until the sticky notes come down.
The problem is not that teams fail to prioritise. It is that they prioritise without agreeing what they are comparing.
The five questions this framework asks
The framework puts the same five questions to every finding, in the same order. Four of them are scored. The fifth does not change the score at all; it changes what you do about it. This is a Curious Society way of working rather than an established named model from the service design literature, and that is worth stating plainly. There is no citation behind it. It exists because the conversation has to happen anyway, and it goes better when everyone is answering the same questions in the same sequence.
Severity asks how badly the problem affects the person who meets it, judged from their side rather than the organisation's. A confusing label and a wrongly refused claim are not the same kind of problem. Spread asks how many people meet it and how often, which is the dimension teams skip most, because answering it honestly needs service data rather than a workshop opinion. A striking problem affecting a handful of people and an unremarkable one affecting most of the caseload look identical on a wall of sticky notes.
Grip asks whether the organisation can act on this at all with the levers it genuinely holds: policy, budget, staffing, the contract, the system of record, the wording it controls. Reach asks whether fixing it in one place fixes it everywhere or only locally, which is the difference between a root cause and one instance of it. Cost of being wrong asks what happens if the team acts on this finding and has misread it, and in particular how easily that action could be undone.
How to run it on a Monday morning
Start by rewriting each finding as a problem somebody experiences, not as a solution. "The eligibility letter does not say what happens next" is a finding. "Redesign the eligibility letter" is not. Cap the list at around twenty-five, because beyond that the session stops being a conversation. Bring whoever holds the service data into the room, since the spread scores are guesswork without them, and score together rather than circulating a spreadsheet. Score severity, spread, grip and reach from 1 to 5 against anchors the group agrees before it starts:
severity: 1 is friction the person barely registers, 5 is the person losing access to something they are entitled to
spread: 1 is a genuine edge case, 5 is most people, most of the time
grip: 1 needs another organisation or a change in the rules, 5 sits entirely inside this team's control
reach: 1 fixes one screen or one office, 5 fixes the cause everywhere it appears
Add the four scores. Above sixteen is a candidate for the next block of work, below eight is a note for later, and the middle band is where the argument belongs. The total is not the interesting output. The disagreement is: when two people score severity at 2 and at 5, they are almost always picturing different people, and the two minutes spent establishing which is worth more than the number. Write one sentence of reasoning beside every score, because that sentence is what makes the decision defensible in three months when nobody remembers the room.
Cost of being wrong then sets the size of the first move rather than whether to make one. Where the action is cheap and reversible, do it and watch what happens. Where it is expensive, hard to undo, or touches something people depend on for money, health or housing, the first move is a test with a small group rather than a commitment.
What to do when the worst problem is not yours to fix
The uncomfortable case is a finding that scores 5 for severity and 1 for grip. Someone cannot get help from the local service until a decision arrives from a separate organisation, and the wait for that decision is the single worst thing on the whole map. Services delivered across several organisations produce this constantly. The two usual responses are both poor: dropping the finding because it is not actionable, which deletes the most important thing in the discovery from the record, or taking it on regardless and spending a quarter negotiating with a body that has no particular reason to move.
The framework handles it by splitting the output into two lists instead of one. The build list holds work the team has grip on. The second list, worth naming out loud as an escalation and evidence list, holds high-severity findings the team cannot act on directly, each recorded with three things: the evidence, the named actor who does hold the lever, and the smallest thing the team can do on its own side. That last part is not a consolation prize. Telling someone honestly how long a decision will take, or not asking them for the same information twice, is real work and it sits within grip.
This is where thinking about services as ecosystems earns its keep. Vargo and Lusch describe value as co-created through interaction among actors rather than delivered by one organisation to a passive recipient, and their later work replaces producer and consumer language with actor-to-actor networks. Loban and colleagues, in a 2021 cross-case analysis of multi-stakeholder primary health care partnerships in two Canadian provinces published in Health Science Reports, examine how clinicians, health authorities, community organisations and patients coordinate across exactly these boundaries. Neither grants a team authority it does not have. Both suggest the coordination problem is itself the design problem.
How to use this without turning it into a scoreboard
The scores are not the decision. They are a record of an argument the team had, with the reasoning attached, and a group that starts treating the total as an authority has reintroduced the original problem in more respectable clothing. Anyone in the room should be able to override a ranking by saying why, provided the why is written next to the number.
Re-score quarterly, and re-score grip in particular. Severity and spread move slowly. Grip can change overnight when a contract comes up for renewal, a new director arrives or a system is replaced, and the finding that was unactionable last year is occasionally the cheapest thing on the list this year. That is the main reason the escalation list is kept rather than thrown away.
Prioritisation is not a method for finding the right answer. It is a method for making the choice visible enough to be argued with.
Companion resource: 🎯The Design Prioritisation Canvas