A Weekly Marketing Report in 15 Minutes: Three Numbers, One Insight, One Decision
Most small businesses either report nothing or produce a monthly deck nobody reads. The first means you drift without noticing what works. The second takes half a day, arrives too late to change anything and mixes forty numbers with no conclusion. A better habit is a one-page weekly report you can finish in fifteen minutes: three numbers that matter, one thing you learned and one decision for next week. It keeps you honest, it gives anyone you answer to a clear picture and it builds, week by week, a record of what actually moves your business. This tutorial shows you how to choose the three numbers, how to collect them with little effort, how to use AI to draft the summary without letting it flatter you, how to separate what you observed from what you guess and how to automate the boring parts safely.
Difficulty: ★★☆☆☆ (no technical background needed)
Tools needed: A spreadsheet or notes app, any chat AI and access to your marketing numbers (ads, email, website, sales)
Last updated: November 2026
Overview
Reporting is not about proving you have been busy. It is about learning, quickly and cheaply, what to do next. A good weekly report does three jobs.
1. It shows where you stand. A few numbers, compared with last week and with your target.
2. It explains one thing. Not every change, just the most useful insight.
3. It forces a decision. What will you start, stop or change this week, and why?
The common failure is the opposite of this: a long report full of numbers with no insight and no decision. People call it "data-driven" when it is really data-heavy and decision-light. The fix is constraint. Three numbers, one insight, one decision, one page.
The constraint works because most small businesses do not need more data. They need a habit of looking at the same few numbers every week and acting on what they see. A report that fits on one page and takes fifteen minutes is a report you will actually write in week twelve, when you are busy and tired.
AI helps in specific, limited ways. It can turn a table of numbers into a clear summary, point out changes you missed, draft the "what to do next" options and tidy the format. It cannot know your business, so it will happily invent explanations that sound plausible. It also tends to flatter: left alone, it will describe a flat week as "steady progress". So we will use it for drafting and checking, with firm rules, and keep the judgement for ourselves.
The tutorial has seven steps. First, choose your three numbers. Second, set up a simple data sheet. Third, the fifteen-minute routine. Fourth, turn numbers into an insight. Fifth, use AI to draft and check. Sixth, the one-page template, and how to report honestly. Seventh, automate the collection, and review monthly. The earlier tutorial on reading ad results gives you the vocabulary, and the quest in this module checks the judgement about claims. This tutorial ties your numbers together.
You will finish with a one-page report template, a data sheet, a fifteen-minute routine and a prompt for an honest AI summary.
Who This Is Useful For
What You Will Learn
What You Need
1. A few marketing numbers you can get each week: from ads, email, your website or shop, and your own sales records.
2. A spreadsheet or notes app.
3. Your goals and targets, if you have them, such as the ad test targets and break-even numbers from the earlier tutorial.
4. A chat AI.
5. Thirty minutes to set up, then fifteen minutes a week.
Information as of 14 November 2026. Dashboards, analytics tools and automation features change often, so this tutorial describes the steps in general terms and names no specific product. If you give an AI a summary of your numbers, include aggregate figures only, and no customer names or personal data. Check each tool's data settings before connecting it to anything.
Step 1 — Choose three numbers that matter
The hardest part of a short report is choosing what goes in it. Most dashboards offer dozens of numbers. You need three.
One main number: the result you want. This is the outcome your marketing exists to produce, measured as close to real money or real customers as you can.
Choose the number nearest to the business result, not the easiest to count. Followers, likes and impressions are easy to count and weakly connected to the outcome.
Two supporting numbers: the levers behind it. These explain why the main number moved. Choose from the funnel:
| Stage | Example numbers |
|---|---|
| Reach | Visits to the site, email list size, search impressions |
| Interest | Click-through rate, email replies, time on a key page |
| Action | Conversion rate, sign-ups, add-to-basket |
| Cost | Cost per customer or sign-up, ad spend |
| After the sale | Repeat orders, refunds, reviews |
Pick the two that best explain your main number. For a tea shop focused on subscribers:
1. Main: new paying subscribers this week.
2. Supporting: visits to the sample-box page, and cost per sign-up from ads.
Test each number. Before you include it, ask:
Avoid vanity metrics. A vanity metric looks impressive and does not guide action: total followers, total impressions, page views with no context. They are not forbidden, but they do not belong among your three. You can keep a small "also watching" line.
Set a target or a comparison for each. A number alone says nothing. For each, record last week's value and your target or break-even, so you can see at a glance whether it is good.
Review the choice every quarter. As your marketing changes, so do the numbers that matter. Swap one when it stops teaching you anything.
Step 2 — Set up a simple data sheet
The report should be a view of a record, not a fresh piece of work every week. A data sheet keeps the numbers in one place so you can see trends.
One row per week. A simple spreadsheet:
Week starting | Main: new subscribers | Support 1: visits to sample-box page | Support 2: cost per sign-up (ads) | Ad spend | Emails sent | Replies | Notes (what changed)Add a column for each of your three numbers, plus the few inputs you need to explain them: spend, emails sent, anything you changed. The "Notes" column is the most valuable one: it records what you did, so that later you can connect changes to results.
Add targets in a header or a second row. For each number, your target or break-even.
Record where each number comes from. A short list: "new subscribers: email tool, subscribers tab. Visits: website analytics. Cost per sign-up: ad dashboard spend divided by tracked sign-ups." Write it once, so the number means the same thing every week and anyone can reproduce it.
Use the same period every week. Monday to Sunday, for example. Mixing periods makes comparisons meaningless.
Check for gaps. Tracking can miss things: people switching devices, blockers, delays. Compare the platform's numbers with your own sales or orders. If they differ, note by how much, and use your own records for the main number where you can.
Add a simple chart. A line chart of each of the three numbers over the last eight to twelve weeks shows the trend better than any sentence. Most spreadsheet tools make this easy. If you need help setting it up, the dashboard prompts tutorial in this track shows how to ask an AI to build charts from a sheet.
Keep customer data out. The sheet holds counts and totals, not names, emails or personal details.
Step 3 — The fifteen-minute routine
Do it at the same time every week. A Monday morning or a Friday afternoon both work. Set a timer.
Minutes 1 to 4: Update the sheet.
Minutes 5 to 7: Look.
Minutes 8 to 10: Find the one insight.
Minutes 11 to 13: Decide.
Minutes 14 to 15: Write and send.
Protect the time. If the routine regularly runs past fifteen minutes, you have too many numbers or too complicated a sheet. Cut something.
Skip rarely, and never skip silently. A missed week leaves a gap in your trend. If you must skip, add a note.
Step 4 — From numbers to an insight
An insight is not a number or a description of a number. It is a sentence that explains a change and points to an action. Learning to write one is the core skill of reporting.
Observation versus insight.
| Level | Example |
|---|---|
| Observation (what happened) | "Sign-ups were 7 this week, up from 4." |
| Interpretation (why, with a label) | "Likely because we moved the button above the fold on Tuesday. Not yet certain." |
| Insight (what it means for action) | "Page layout seems to matter more than ad copy right now, so I will fix the other pages before touching ads." |
A report that stops at observation is a log. A report that reaches insight guides action.
Ask the same five questions each week.
1. What changed most? Look for the largest movement among your three numbers.
2. What did we do? Check your notes: a change in ads, content, price, email or the site?
3. What else could explain it? Seasonality, a holiday, a competitor, a one-off mention, a tracking change, a small sample.
4. How sure are we? Rate it: seen several weeks in a row, seen once, or a guess.
5. So what? What does this suggest we do next?
Beware of the common traps.
Separate fact from guess. In the report, mark each statement: Fact (counted), Likely (supported by the pattern and your notes) or Guess (a hypothesis to test). This keeps you honest and tells readers how much weight to give each line.
Write it in one or two sentences. "Sign-ups rose from 4 to 7 after we moved the booking button above the fold (likely, one week of data). Next week I will check whether it holds, and I will move the button on the second page too."
Step 5 — Use AI to draft, and to challenge
Now bring in AI, for the parts it does well.
Give it numbers, not people. Paste your three numbers, last week's values, your targets and your notes. No customer names, emails or personal details.
A drafting prompt.
Role: a careful marketing analyst who values honesty over optimism.
Task: Below is this week's marketing data for a small business, with last week's numbers, my targets and my notes on what I changed. Write a one-page weekly report with: (1) the three numbers versus last week and versus target, in a small table; (2) one insight, in one or two sentences, that explains the biggest change and points to an action; (3) one recommended decision for next week; (4) a "confidence" label for each statement: Fact, Likely or Guess; (5) a list of anything that makes these numbers unreliable (small samples, tracking gaps, other changes).
Rules: Use ONLY the data and notes I give. Do not invent causes. If the sample is small, say so. Do not call a flat or down week progress. Do not recommend increasing spend unless the data clearly supports it. If something is missing, say "MISSING: ...".
Data and notes: [paste]Then make AI attack its own draft. A second prompt is very useful.
Role: a sceptical colleague.
Task: Here is a draft weekly report. Tell me: (1) which statements are not supported by the data, (2) what other explanations could account for the change, (3) where the sample is too small to trust, and (4) what you would want to check before acting. Do not rewrite the report.
[paste the draft]Check the arithmetic. AI can miscalculate percentages and sums. Recompute at least the main number and one ratio yourself.
Edit for voice and brevity. Cut anything vague. A report is for decisions, not for impressing anyone. Remove empty phrases ("showing positive momentum", "strong performance") and replace them with the number.
Do not let AI choose the decision. It can list options. You choose, because you know what you can afford in time, money and energy.
Keep AI out of the raw data. Do not connect an AI tool to customer records. Paste only the aggregated figures you have chosen to share, and check your tool's data settings.
Use it for the monthly view. Once a month, paste the last four weeks of your sheet and ask: "What is the trend? What has this month taught us? What should we try next month?" Again, check against your own judgement.
Step 6 — The one-page template, and reporting honestly
Use the same template every week. A familiar format is quick to write and quick to read.
WEEKLY MARKETING REPORT
Week of: ___
1. THE THREE NUMBERS
| Measure | This week | Last week | Target | Status |
| Main: ___ | | | | |
| Support 1: ___ | | | | |
| Support 2: ___ | | | | |
Notes on reliability (small samples, tracking gaps): ___
2. ONE INSIGHT
___ (one or two sentences)
Confidence: Fact / Likely / Guess
3. WHAT WE DID THIS WEEK
___ (the one to three changes that matter)
4. ONE DECISION FOR NEXT WEEK
Action: ___ Owner: ___ By: ___
How we will know it worked: ___
5. RISKS AND UNKNOWNS
___
Also watching (optional, one line): ___Reporting honestly. A report is only useful if it is true. These habits keep it that way.
A short note on claims. The next tutorial covers honest marketing in depth: claims, disclosure, copyright and AI labels. A weekly report is an internal tool, but it is also where you check that what you tell customers is backed by what you see.
Step 7 — Automate the collection, and review monthly
Once the routine works by hand for a few weeks, you can reduce the effort. Automate the collecting, never the thinking.
What is worth automating.
What to keep manual.
Automate with care.
Review monthly. Once a month, spend thirty minutes with the last four reports.
1. What is the trend in your three numbers?
2. What did our decisions achieve? For each of the past four decisions, did it work, was it inconclusive or did it fail?
3. What did we learn? Write three sentences you could tell a colleague.
4. Are these still the right three numbers? Swap one if it has stopped being useful.
5. What is the plan for next month? One or two priorities.
Review quarterly. Once a quarter, check your targets, your break-even numbers and your data sources. Prices, costs and tools change.
Keep the archive. Over a year, your weekly reports become a record of what you tried and what happened. That record is worth more than any advice, because it is about your business.
Worked Example: a week at the tea shop
Mount and Leaf, the small online tea shop, wants 40 new subscribers in six weeks. The owner has chosen her three numbers: new paying subscribers (main), visits to the sample-box page and cost per sign-up from ads. Her targets: 7 new subscribers a week, and a cost per sign-up below HK$100.
Minutes 1 to 4. She pulls the numbers. This week: 9 new subscribers (last week 5), 380 visits to the sample-box page (last week 340), and cost per sign-up of HK$88 (last week HK$130). Her notes: "Moved the order button above the fold on Tuesday. Sent a newsletter on Thursday. Added 'free tea samples' as a negative keyword."
Minutes 5 to 7. She compares them with last week and target: subscribers above target, visits up modestly, cost per sign-up below target. The trend chart shows the last three weeks rising. She notes the ad sample is small: 14 ad sign-ups.
Minutes 8 to 10. The biggest change is subscribers and cost per sign-up. Possible reasons from her notes: the button move, the negative keyword and the newsletter. She cannot separate them, and a newsletter gives a short-lived bump. She labels her insight "Likely, one week of data".
She uses AI with her numbers and notes. Its draft says "strong growth driven by the button change". She asks it to attack the draft. The second answer points out that three changes happened in one week, the newsletter alone could explain the rise and 14 sign-ups is a small sample. She rewrites the insight in her own words.
Minutes 11 to 13. Her decision: do not change anything else for a week, so she can see whether the improvement holds. She will check the sign-up rate on the page, separate from the newsletter bump, using the numbers for Friday to Sunday.
The report. One page: the three numbers against last week and target, her insight ("Sign-ups and cost per sign-up improved after three changes at once, so I cannot say which mattered. Likely, small sample. Next week I change nothing and watch."), one decision, and a risk note about the newsletter bump. She sends it to her part-time helper and files it. The whole thing took fourteen minutes.
A month later. Her monthly review shows the number of subscribers rising steadily, but the cost per sign-up creeping back toward HK$110 in week four. She writes: "Ads are slowing. Test a new hook." That is her next decision, and it came from a record, not from a feeling.
Common Mistakes to Avoid
Three habits make reports useless.
Mistake #1: Too many numbers. Forty metrics hide the one that matters, and the report takes half a day, so it stops. Fix: choose three numbers that would change a decision, keep a small "also watching" line and review the choice quarterly.
Mistake #2: Reporting numbers without an insight or a decision. A list of figures is a log, and nobody acts on a log. Fix: always write one insight with a confidence label, and one decision with an owner and a date.
Mistake #3: Letting AI or optimism tell a neat story. AI flatters flat weeks, invents causes and credits whatever you changed last. Fix: give it only aggregate numbers and notes, ask it to attack its own draft, check the arithmetic, label facts, likely explanations and guesses, and report the bad weeks.
Going Further
Build a dashboard for the trend. A simple chart page of your three numbers over twelve weeks. The dashboard prompts tutorial in this track shows how to ask an AI to build one from a spreadsheet.
Share a team version. If several people work on marketing, give each person one number to own, and meet for ten minutes a week to go through the report.
Add a "what we learned" log. A running list of one-line lessons from each decision. After six months, it is a handbook for your business.
Use a Bot for the routine, with limits. A recurring task that gathers numbers into your sheet and drafts a summary can save time, if it is read-only, the output is always checked by you and nothing is sent without your approval. The Grok Bot marketing guide in this track covers this kind of use.
What to learn next. The last tutorial in this track is on honest marketing: claims, disclosure, copyright and AI labels, and a checklist before anything goes live. It connects to the report because honest numbers and honest claims go together.
Key Takeaways
In this tutorial you learned:
Before you move on, a challenge. Choose your three numbers and build the data sheet with eight weeks of past data if you have it. Run the fifteen-minute routine this week, draft the summary with AI using the two prompts, and write one insight with a confidence label and one decision with a date. Next week, check whether the decision worked.