A method is a recipe for making a choice well. Each recipe here says what it's for, what you need, the steps, an example, and where it can go wrong.
The method library collects the reusable tools on this site, such as templates, scoring scales and rules. Each one follows the same layout so you can pick it up quickly. You can reuse and adapt them; the worked examples show how, using made-up numbers.
A method here is a repeatable procedure with defined inputs and outputs, applicable across organisations. Pages follow a fixed structure (purpose, inputs, steps, worked example, limits, sources). Examples are illustrative; verify any product facts against the cited sources and their access dates.
Why a library
I learn by explaining things. When I explain something well enough, it often turns into a method: a set of steps someone else could follow without me. Over time, the site has collected quite a few of these. The library is where they're gathered, so they can be found and reused. The full list is in Every method on this site, on one page.
What counts as a method
I count something as a method if it passes three tests.
- It's repeatable. Two people following it on the same decision should get a similar answer.
- It has clear inputs and outputs. You know what you need before you start and what you'll have at the end.
- It works outside one organisation. If it only makes sense with one company's systems, it's a case study.
That rules out a lot of good advice, which is fine. Advice lives in the other articles.
The layout
Every method page follows the same pattern, though the headings vary to suit the topic.
| Part | What it covers |
|---|---|
| Purpose | The question the method answers, in a sentence or two |
| Inputs | What you need: data, people, time |
| Steps | Numbered where order matters |
| Worked example | A realistic case with illustrative numbers, labelled as such |
| Limits | Where the method breaks down, or what I'm still unsure about |
| Sources | Primary sources for any product or factual claim, with the date I checked them |
Each page also has three introductions at different levels: one for a curious child, one for a fifteen-year-old, and one for a professional. That's the Feynman habit behind the site. If I can't explain a method simply, I don't understand it well enough yet.
Worked examples are illustrative
The numbers in worked examples are made up to show the method working, and they're labelled that way. They aren't benchmarks. Where an example uses a real price or limit, such as a Microsoft licence price or a Copilot Credit rate, that figure comes from the cited source on the date shown, and it may have changed since.
Adapting a method
Most methods have settings you should change. The money bands in the scoring guide suit a mid-sized organisation; a small charity would lower them. The review cadence in Review dates that happen is a starting point.
When you adapt one, I'd write down what you changed and why, using the decision record format. Changing a method's settings is itself a decision.
A worked example
This example is illustrative. A council's IT team wants to use the threshold method for an agent that suggests password reset approvals.
| Step | What they do |
|---|---|
| Purpose | Decide the confidence above which the agent resets without a person |
| Inputs | Cost of a wrong reset (security risk), cost of a wrong refusal (an hour of lost work) |
| Adapt | They add a rule that some accounts always go to a person, whatever the score |
| Record | They capture the change and the reason in a decision record |
| Review | They set a review date for three months |
The method didn't need to change. One setting did, and it's written down.
What I'm still checking
One planned method, the Decision Diagnostic, is listed in this collection as coming. I haven't written it up in full yet, and I'd rather wait than publish a half-version.
Sources
This page describes how this site's method library is organised. It uses no external facts or figures.