How it works
Five parts, in the order you meet them.
A company answers questions and gets a result. The result names what to fix. Ninety days later it is taken again. Everything else exists to make those three things trustworthy and repeatable.
Each part depends on the one before it. Click any part to jump to it.
Part 1 of 5
The method
Everything that produces the number, written down.
Published in fullThe model, the questions, both answer scales, the arithmetic, the bands, the version record and the conformance test vectors.
You do not have to trust us about how the number is made, because you can read exactly how it is made. Run it yourself in the browser, or implement it in your own systems and check your implementation against the published vectors. Everything above this part rests on it being inspectable by anyone.
Part 2 of 5
The instrument
Eight minutes, and the company keeps the result forever.
Free to every company, alwaysThirty seven questions. A score from 0 to 100, capability reported separately from momentum, the stage where the problem shows up, the root cause underneath it, and the first actions.
No upgrade, no paywalled result, no charge at any point, whether or not the company ever speaks to anyone. That is not generosity. Commercial capability cannot be scraped from filings or web traffic, it can only be asked about, and a company only answers honestly if the answer is useful to them and costs them nothing.
Part 3 of 5
Diagnosis and action
A number nobody acts on is a report. This is the part that is not.
Where the work happensThe written diagnosis, the three actions that follow from it, the category of support it points to when a person is needed, and the reporting across a cohort.
Most of what a diagnosis calls for is work the company does itself. Where it needs a person, the diagnosis names the category of help, never the supplier: the institution decides who provides it, from provision it already funds. Ninety days later the measure is taken again, and the movement is the thing everyone was actually after.
Part 4 of 5
Checking a result
A result anyone can confirm, without seeing anything private.
Open to anyone shown oneEvery result carries a short reference. Enter it and you learn four things: whether it is real, which method version produced it, when it was taken and whether it is still current, and its evidence level.
A report can be screenshotted and a number can be edited. A reference cannot. It returns no score, no company name and no answers, so a company can prove a result is genuine without disclosing anything it did not choose to share. Same idea as checking a certificate number.
Part 5 of 5
How it improves
The measure gets sharper because companies used it.
Over timeA record of which kind of support tends to move which part of the engine, for which kind of company. It builds only as companies measure, act, and measure again.
This is why a published, unchanging instrument is worth more than a bespoke one. Every company measured on the same method makes the next diagnosis a little better founded. Nothing about an individual company is ever published, and no peer benchmark is released until there are enough comparable observations to make one honest.
Who does what
| If you are | What you actually do |
|---|---|
| A company | Answer 37 questions in about eight minutes. Read your result. Do the first action. Take it again in ninety days. You are never charged and the result stays yours |
| A programme, bank, university or fund | Invite a cohort. Watch coverage. Read the cohort report, and the movement report ninety days later |
| Someone shown a result | Check the reference. Confirm it is real, dated and produced by the version it claims, without seeing anything private |
| Someone running the method themselves | Read the specification, implement it, and check your implementation against the published test vectors |