Weekly Metrics Summary
Turn a metrics CSV into a plain language weekly readout with deltas.
A prompt that turns a raw metrics export into a plain language weekly readout, with week over week deltas called out and the noteworthy changes flagged, so a stakeholder can read it in a minute.
By Tri Vo
A grid of numbers is not a report. This prompt takes a metrics export and writes the recap: what moved, by how much, and what is worth a second look. It calls out the week over week deltas and flags the changes that matter instead of listing every row. You paste the numbers, you get a readout you can send with light edits.
The prompt
You are a marketing analyst writing a weekly readout for a busy stakeholder.
I will paste a metrics table (this week and last week, or a CSV). Produce:
1. Headline: one sentence on the overall story of the week (up, down, flat,
and the main driver).
2. What moved: a short list of the metrics that changed most, each with the
absolute value and the week over week change as both a number and a percent.
Mark each as good, bad, or watch.
3. Why it likely moved: 1 to 2 plausible drivers per notable change, framed as
hypotheses, not certainties. Do not invent causes the data cannot support.
4. Callouts: anything that needs a decision or a closer look this week.
5. Next step: one concrete action for next week.
Keep it under 200 words total. Plain language, no jargon, round numbers where
it aids reading. If a metric is missing or looks like a tracking error, say so
rather than guessing.
Here is the data:
[paste your metrics table or CSV]How to use
- Export your weekly metrics as a table or CSV (this week and last week side by side).
- Paste the prompt into Claude Code and drop your data in at the bottom.
- Check the “Why it likely moved” hypotheses against what you know before you send.
- Send the readout, or paste it into the Newsletter command if it goes to a wider list.
Questions and answers
Is Weekly Metrics Summary free to use in Claude Code?
Yes. Weekly Metrics Summary is a free entry in the marketing library. Copy the prompt and use it in Claude Code at no cost.
What sections does the readout include?
A one-sentence headline, a What Moved list with each change as a number and a percent marked good, bad, or watch, likely drivers framed as hypotheses, callouts that need a decision, and one next step, all under 200 words.
How should I prepare the data?
Export this week and last week side by side as a table or CSV and paste it at the bottom of the prompt. If a metric is missing or looks like a tracking error, the prompt flags it instead of guessing.
Can I trust the Why It Moved explanations?
Treat them as hypotheses, not facts. The prompt is told not to invent causes the data cannot support, but check each driver against what you know before sending.