How the method works
The question we are asked most often is how the number is calculated. It deserves a page of its own, and this is it.
What follows is the shape of the method. The calibration itself belongs to your organisation and is agreed with you before a trial begins.
Four principles
-
Code and organisation, read together
Read apart, each one shows half the picture: the code says what is tangled, the organisation says who has to untangle it. The cost lives in the pair.
-
Every number traces back
For any figure we show, you can walk back to the observation that produced it and to the system it was read from. No figure appears without that path.
-
The score and the estimate stay apart
What we observe is a diagnostic score and it stands on its own. The translation into money is an estimate, calibrated on your organisation, and it is labelled as one. It is an opportunity cost, not cash leaving a bank account.
-
The error is declared
Every estimate carries its uncertainty in plain view. You should see the error before you see the number.
What we read
We start from the systems that are already instrumented and reliable, and we say what each one contributes.
- Version control which parts of the system change, how often, and who actually changes them
- Project management what was planned, what it turned into, what waited and for how long
- Continuous integration and delivery how long the path from a change to production takes
- Agent gateway logs how much of the output is written by people, by assistants, by agents
What we do not ask for
No tracking of individual people, and no measurement of individual productivity
No dependency on human resources data to begin: the organisational picture starts from ownership as it is practised in version control
No access to the content of your product, your customer data, or your source code beyond what the analysis needs
What we report back
For every connector, what was read and what was not. A gap in the data is part of the result, not something to discover later.
Formulas and coefficients are not on this page. The model is calibrated on the first customers, and publishing it today would mean publishing numbers that are going to change.