Memory
Performance Analyst
Your week's numbers read back to you in sentences: the three to five things that actually moved, why each matters, and one action apiece, with every figure traced to your input.
Reads your operating system before it starts. Without one, it calibrates from a few questions and tells you plainly what the house would have added.
The task
The month's figures have landed in a spreadsheet and you have a pricing call to make on Thursday; you need to know what actually moved before you decide.
By hand
You open the export and sort by percentage change, because that feels like analysis. The top of the list is a 40% swing on a line worth almost nothing, and you spend twenty minutes on it before realising it's noise. Meanwhile the number that actually decides Thursday's call moved quietly by a few percent, three columns over, and you haven't looked at it yet.
An hour in, you have a feeling rather than a finding, and somewhere along the way you've started saying 'on track to' about a trend the data doesn't actually show. Nobody checks, because there is nobody to check.
With the specialist
Three answers
The figures, the prior period if there is one, and the decision this readout should inform. Where your metrics ledger is installed, everything is read against your own targets and definitions, never a fresh one.
- QThe figures themselves, pasted or as a file
- QThe prior period to compare against, if any
- QThe one decision this readout should inform
The ship gate
Every number is traced to your input before you see it, derived figures are labelled 'about', and anything resembling a forecast is cut unless you supplied the trend yourself. Findings are ranked by what matters to your decision, not by the biggest percentage.
- Every number maps to a supplied input, or is a labelled 'about' figure computed from two of them; anything unsourced is deleted
- No forecasts: no 'on track to' or 'at this rate' unless you supplied the trend yourself
- Findings ranked by what matters to your decision, not by the size of the percentage swing
- The readout admits at least one thing the data cannot answer, rather than filling the gap with a guess
The deliverable
A one-sentence bottom line, a what-moved table of the three to five findings that matter, one concrete this-week action per finding, and an honest line on what the data cannot tell you.
Sample output
What the deliverable looks like, on a neutral example.
A monthly readout for a small online shop
Bottom line: takings held steady while one cost line did most of the moving, so the story this month is spend, not sales.
Do this next: one concrete action per finding, each startable this week. Then the honest line: the question these numbers raise but cannot answer.
By hand
typically 2 to 3 hours in a spreadsheet, often skipped in a busy week
With the specialist
your figures pasted in, minutes later a readout, then your review
Illustrative comparison from the authors' own use; estimates, not measurements.
Performance Analyst: £19, yours to keep.
Buy once, updated free forever. No subscription. Leave your email and you’ll hear the moment it launches.
Runs on your own Claude subscription. Prices are indicative while we finish arguing about them.
