FP&A Assessment:Operating Model Review
Eight findings, each confirmed by two blind peer reviews, on a finished seven-year three-statement model. Everything a script can detect is published in advance and scores nothing.
Nobody is asked whether a finding is clever
Machines decide what was already known. People decide only whether a finding is real. That split is deliberate: asked to rank original work, people report valuing originality and choose the familiar.
What a finding has to clear
- 01
Specified
Decided by machineEvery field is present, long enough, and states a condition somebody else could go and check. A finding that hedges instead of naming what breaks is refused before a person reads it, which is what keeps the rest of the process affordable.
- 02
Not already listed
Decided by a person, against the published listCompared against the risks the exam publishes in advance. Restating one scores nothing, and the list is public precisely so that recall is worth nothing and nobody can be surprised by it.
- 03
Real
Decided by two blind checkersTwo people who sat the same exam rule on whether the thing described actually happens. They never see who wrote it, and they are never asked whether it is clever, interesting, or better than their own.
What the numbers mean
Reported side by side and never averaged into one score, because they answer different questions and a composite hides which one moved. Every figure a row carries is published beside it, so the order can be checked rather than taken on trust.
- Confirmed
- How many of the findings somebody submitted came back confirmed. Before an attempt has been checked, the same column reads how many cleared the first, mechanical check instead. That is the only thing decided about them so far.
- Reviews
- How many of other people's findings that author went on to rule on. It is here because the checks that settled their own findings ran on somebody else's time, and an attempt that returns less than half of what it owes is left unranked.
- Time
- How long the attempt took, where that was observed. There is no clock on the exam and no credit for being quick.
- Precision
- Confirmed over judged: whether somebody can tell their own good findings from their bad ones. It breaks ties rather than deciding the order, because it stops at 100 per cent and a cohort of competent people all pile against that ceiling.
- Rarity
- The column the order is made on. Each confirmed finding is weighted by how few other participants found the same thing, and the weights are summed. It has no ceiling, so it is printed as a figure rather than a percentage.
What is never published
The findings themselves. What somebody writes goes to two other participants with their name removed before either of them reads it, and it stays there. There is no public archive of anybody's findings, and there is no plan for one. A name appears on the board only where that person has public visibility switched on; everybody else holds their position anonymously, and their row stays, because removing it would print rank three as rank two and nobody reading the board could see that happen.
