The weekly numbers review: 30 minutes, same 10 numbers
The habit that separates operators from busy founders is a fixed weekly review of the same ten numbers, every week, no exceptions. Same numbers, same slot, one action per red number, so trends surface before they become surprises.
Fastest path: one prompt, end to end
🤖 AI prompt — paste into ChatGPT / Claude
You are my weekly business analyst. Turn my raw exports into a 10-line readout.
I will paste: this week's numbers and last week's, plus a 4-week list where I have it.
The 10 numbers (use these, in order): revenue, orders, AOV, blended CAC, contribution margin %, repeat-purchase rate, email revenue share, inventory weeks of cover, support ticket backlog, cash weeks of runway.
Do this:
1. For each, show this week vs last week (% change) and vs the 4-week average.
2. Flag red (worse and meaningfully off-trend), yellow (drifting), green (fine).
3. For every red number, propose ONE concrete action for the coming week.
4. Keep it to 10 lines plus the actions; no fluff.
Do not invent any figure; if I omit one, list it as missing and move on. If you cannot open a linked export, ask me to paste the values.
Output: the 10-line readout + a short action list, red numbers first.
Or do it in 4 steps
- Fix the ten numbers and the slot. Revenue, orders, AOV, blended CAC, contribution margin %, repeat rate, email revenue share, inventory cover, support backlog, cash weeks. Same ten, a recurring 30-minute block (e.g. Monday 9am). The discipline is in never changing the list.
- Pull the exports and let AI assemble the readout. Export from your store, ad platform, email tool, and helpdesk; paste them in with the prompt above. AI's job is turning messy CSVs into ten clean lines; it must not fabricate a single figure.
- Compare two ways: last week and the 4-week average. Week-over-week catches shocks; the 4-week average catches slow drift a single week hides. A number is "red" only when it's worse AND off its own trend.
- One action per red number, written down. Don't debate all ten. Pick the reds, commit one action each, and check them off next week. That closes the loop and makes the review compound.
Worked example (labeled): Monday review: revenue +4% (green), but blended CAC $31 -> $39 against a $33 four-week average (RED), and contribution margin 41% -> 36% (RED).
Two reds, two actions: pause the worst-ROAS ad set, and re-check the free-shipping threshold that's eroding margin. Everything else is noted and left alone. Thirty minutes, done.
Same ten numbers, same slot, every week; the ritual matters more than any single metric. Save the readout as your kpi-dashboard artifact, and pair it with daily-ops-dashboard under operations, that daily scan is operational while this is the weekly business review.