For each choice your team makes, give three scores: how often it happens, how much it matters, and how easy it is to undo. Use the same ruler for everyone.
The decision inventory asks how often each decision happens, what's at stake, and whether it can be undone. Without shared scales, people score these very differently. This guide gives each a 1 to 5 scale with plain descriptions, and shows how to read the result.
Anchored 1 to 5 scales for frequency (per role per month), stakes (worst plausible single-error cost across money, people and compliance) and reversibility (time and cost to undo). Use frequency for opportunity, stakes and reversibility for control level; score in pairs and reconcile differences of two or more.
Why scales matter
The decision inventory template has columns for how often a decision happens, what's at stake and whether it can be undone. When two people fill those in without guidance, one writes "high" where another writes "medium" for the same decision. The inventory then reflects who filled it in more than the decisions themselves.
Anchored scales fix most of that. Each score has a description that's specific enough for two people to agree.
Frequency
Count per role, per month. If a decision is shared by a team, divide by the number of people who make it.
| Score | Frequency per person per month |
|---|---|
| 1 | Less than once |
| 2 | 1 to 4 times |
| 3 | 5 to 20 times |
| 4 | 21 to 100 times |
| 5 | More than 100 times |
Frequency tells you about opportunity. A small improvement to a decision made 200 times a month adds up; a big improvement to one made twice a year may not be worth the effort.
Stakes
Score the worst plausible result of one wrong decision. "Plausible" means something that could happen in a normal year, so leave out disaster scenarios. Take the highest of the three columns.
| Score | Money | People | Compliance |
|---|---|---|---|
| 1 | Under £100 | Minor inconvenience | None |
| 2 | £100 to £1,000 | Annoyance, a complaint | Internal policy breach |
| 3 | £1,000 to £10,000 | Real harm to one person, fixable | Reportable internally |
| 4 | £10,000 to £100,000 | Serious harm to one person or harm to several | Regulator interest likely |
| 5 | Over £100,000 | Serious, lasting harm | Regulatory action or legal claim |
The money bands are illustrative. A small charity and a large bank would set them differently, and that's fine as long as everyone in one organisation uses the same bands.
Reversibility
Score how hard it is to undo a wrong decision once it's made.
| Score | Reversibility |
|---|---|
| 1 | Undone in minutes, at no cost |
| 2 | Undone within a day, small effort |
| 3 | Undone within weeks, real effort or cost |
| 4 | Partly undoable; some harm remains |
| 5 | Can't be undone |
There's more on why this deserves its own score in Why reversibility deserves its own column.
Reading the scores together
I read frequency on its own, and stakes and reversibility as a pair.
| Frequency | Stakes and reversibility | What it suggests |
|---|---|---|
| High | Both low | Good candidate for a decision model that acts, with sampling |
| High | Either high | Decision support: the model suggests, a person decides |
| Low | Both low | Probably leave as it is |
| Low | Either high | Better information and a clear record; keep people in charge |
A worked example
These scores are illustrative. A housing repairs team scores three decisions.
| Decision | Frequency | Stakes | Reversibility | Suggests |
|---|---|---|---|---|
| Priority of a repair request | 5 (about 150 a month each) | 3 (a missed emergency harms a tenant) | 3 | Decision support |
| Which contractor gets a routine job | 4 | 2 | 2 | Model that acts, sampled |
| Whether to approve a major works variation | 1 | 4 | 5 | Better information, people decide |
The first row is the interesting one. It happens often enough to be worth improving, but a wrong call can hurt someone. So the model should suggest a priority and show why, and the call handler decides.
How to score
Score in pairs. Two people who know the work score separately, then compare. Where they differ by two or more, talk it through and write down why. The notes from those conversations are often more useful than the scores.
What I'm still checking
The stakes bands mix money, people and compliance, and taking the highest can overstate decisions that are mainly about money. I've kept it simple for now. A fuller version might score the three separately, but I'm wary of adding columns people won't fill in.
Sources
This guide describes my own scoring scales. It uses no external facts or figures; all money bands and examples are illustrative.