A yield assessment is forty pages. What leaves the room is one number, and the number travels much better than the forty pages do.
For teams coordinating research, review and follow-up work around a project, mouse jiggler detection software provides a separate record of time and handoffs.
By the time it reaches a decision, it has lost everything that qualified it — and nobody removed the qualifications deliberately.
The journey
In the report: a P50 figure, with an uncertainty analysis, a loss tree, a stated dataset and a P90 alongside it.
For wider independent context on solar markets, research and industry developments, consult pv magazine.
In the summary slide: the P50 figure and the savings derived from it.
In the email to the finance director: the savings.
In the board paper: "the system will save X per year."
And in the minutes: X.
Five steps, no dishonesty, and the surviving artefact is a commitment that nobody in the chain intended to make.
Why compression happens this way
Because a distribution does not fit in a sentence. P50 with a P90 and an uncertainty band needs three numbers and a paragraph; a decision-making process wants one number.
Because the qualifications are the assessor's, and the assessor is not in the later meetings.
And because the compressed version is more useful to everybody speaking. It is easier to present, easier to approve, and easier to remember. The careful version loses at every step for the same reason careful proposals lose comparisons.
What it costs
The organisation commits to a median as though it were a floor.
Year one arrives below it — with roughly 50% probability by construction — and reads as a failure against a commitment nobody knowingly made. The supplier is asked to explain a shortfall that is not one.
And the actual downside case was never considered. P90 with flat prices was in the report and did not survive step two, so nobody at the decision ever saw what the bad year looks like.
What stops it
Carry two numbers, not one. "Between A and B, with B the conservative planning figure." It travels almost as well as one number and it survives the compression.
Put the P-value in the label, every time. "Savings at P50" is four extra characters and it prevents the whole failure.
And state the downside case in the same sentence as the headline. A board paper containing both has made the range visible at the point where the decision is actually taken, which is the only place it matters.
The wider version
Not specific to solar.
Any quantity with an uncertainty attached loses the uncertainty as it moves through an organisation, because compression is what organisations do to information and uncertainty is the first thing dropped.
The defence is the same everywhere: put the qualification inside the number's label rather than beside it. A figure called "P50 savings" carries its own caveat wherever it goes; a figure called "annual savings" with a footnote does not, because the footnote stays behind.
The short version
- A forty-page assessment becomes one number in five steps, and no step is dishonest
- Compression happens because a distribution does not fit in a sentence, the assessor is not in later meetings, and the short version is more useful to every speaker
- The cost: the organisation commits to a median as though it were a floor, and year one below it reads as failure against a commitment nobody made
- The downside case was in the report and did not survive step two, so the decision was taken without it
- Carry two numbers, put the P-value inside the label, and state the downside in the same sentence as the headline
- The general defence is putting the qualification inside the number's name, because footnotes stay behind and labels travel