Lesson 8.1 · 50 minThe weekly report

What you will have

The student has a report template and a reusable prompt saved in docs/reports/, and a first weekly report written from their real data that ends with at most three concrete actions.

Video

The video for this lesson is not recorded yet.

Prompts used in this lesson

Part 5 — Build the template and the prompt

Purpose: I want a weekly routine that turns my marketing data into one page that
says what happened and what to do next. Set up the reusable parts.

Context: read docs/ads-plan.md (targets and the stop point), docs/measurement-plan.md
(the funnel) and docs/crm-spec.md (pipeline stages). Each week I will save exports
for one seven-day period in docs/reports/data/: Google Ads CSV, Meta Ads CSV, a GA4
funnel screenshot, my Clarity notes as text, a CRM lead CSV with contact columns
removed, and a CRM dashboard screenshot.

Please create:
1. docs/reports/TEMPLATE.md with these sections: Period and sources; Numbers (a
   table by campaign: spend, clicks, leads, cost per lead, qualified leads, cost
   per qualified lead, and the target from the ads plan); Funnel; What visitors
   did (from Clarity notes); Data problems (where sources disagree); Findings;
   Decisions; Not doing this week.
2. docs/reports/PROMPT.md: the instruction I will give you each week to fill the
   template from the files in docs/reports/data/.
3. Add docs/reports/data/ to .gitignore, because exports can contain personal data.

Rules the weekly prompt must contain:
- Give findings, not a summary. A finding joins at least two sources or compares
  a number with a target.
- Every rate is shown with its counts, for example "5 of 38".
- Label each finding solid, tentative or too early to say, with the reason. If
  moving one or two leads would change the conclusion, it is not solid.
- List every place where two sources disagree, with both numbers.
- End with at most three actions. Each action has: what to change, why, and what
  number we will look at next week to see whether it worked.
- Never invent a number. If a figure is missing from the files, say which file
  and which column you expected.

Done when: both files exist, .gitignore has the new line, and you have shown me
the template. Do not write a report yet.

Part 6 — The first report and the decisions

Follow docs/reports/PROMPT.md for the period 2026-09-28 to 2026-10-04. The files
are in docs/reports/data/. Save the report as docs/reports/2026-10-04.md. Before
you finish, re-add the lead counts yourself from the CRM file and confirm they
match the numbers in your table. Then show me the Findings and Decisions sections.

Do along

Work on your own project and pause the video where a step says so.

1. Pause after Part 2. Choose your period: the last seven full days. Write the dates down.
2. Export the five sources described in Part 2 for that period. Watch five Clarity recordings and write your notes as plain sentences.
3. In the CRM export, delete the name, email, phone and message columns. Save all files in docs/reports/data/.
4. Pause after Part 5. Run the setup prompt from Part 5. Adapt the list of sources if you run only one ad platform, and the stage names if your pipeline differs from new, contacted, qualified, won, lost. Run git status and confirm that nothing inside docs/reports/data/ is listed.
5. Pause after Part 6. Run the weekly prompt from Part 6 with your own dates and file name.
6. Verify two numbers by hand against the exports. Apply the one-person question to each finding.
7. Edit the Decisions section yourself until it holds one, two or three actions that you will really carry out this week.
8. Commit: git add -A, then git commit -m "Lesson 8.1: weekly report template, prompt and first report", then git push.

Check your work

1. Open docs/reports/. Expected: TEMPLATE.md, PROMPT.md and one dated report.
2. Open the report and scroll to the end. Expected: a Decisions section with no more than three actions, each naming the number you will look at next week.
3. Search the report for a percent sign. Expected: every rate has its counts beside it.
4. Run git status. Expected: no file from docs/reports/data/ appears.

Common problems

• The report repeats the numbers and draws no conclusions → the weekly prompt was given without the prompt file, or the rules were shortened → tell Claude: "Re-read docs/reports/PROMPT.md and rewrite Findings; each finding must join two sources or compare with a target."
• Numbers in the table do not match the export → mixed date ranges, or a currency or decimal format read wrongly → tell Claude which figure is wrong and ask it to show the rows it added up; re-export with matching dates if needed.
• GA4 and the CRM disagree on lead count → own test submissions, consent-rejected visitors, or duplicates → rerun the comparison from docs/tracking-qa.md (lesson 6.9) for the same dates.

Homework

About 30 minutes, on your own, after the lesson. Lesson 8.2 does not depend on it.

1. Carry out one decision. Pick the action in your Decisions section that needs no building (a budget move between ads, a paused ad) and do it, without raising your total ad budget. If every action needs building, write beside each the date you will do it. Deliverable: one line under the action with the date and what you changed. Done when: that line is in the report and the change is visible in the ad platform.
2. Watch the visitors who did convert. In the lesson you watched five people who gave up. Now watch up to five recordings that end on /thank-you, and add to "What visitors did" what these visitors did differently, with counts ("4 of 5 opened the FAQ"). Done when: the section holds at least three such observations, each with its count.
3. Book the routine. Create a recurring weekly slot in your calendar for the report, with the five exports and docs/reports/PROMPT.md named in the description. Done when: the event repeats weekly and the first one falls within the next eight days.

The report changed, so commit without a tag: git add -A, git commit -m "Homework 8.1: report notes", git push.