Skip to content
JAC Growth Marketing
All playbooks

Automate Your Weekly Marketing Report

A step-by-step playbook to turn scattered metrics into a one-screen weekly marketing report that writes itself, with the prompts to copy.

Marketing Engineering Playbook beginner 6 steps 9 min read

The short version

  • A weekly report should answer one question: is the machine working, and what do we change? One screen, five numbers, a short narrative, and at most two actions. Everything else is decoration.
  • The build is four parts: pick the five numbers that matter, centralize where they come from, run a summarize-and-flag prompt, then force a 'so what' on every metric.
  • AI writes the narrative, you own the verdict. The model turns numbers into readable prose and flags anomalies. Deciding what to do stays human.
  • Automate the assembly before the collection. A report that writes itself from pasted numbers ships this week. Full data plumbing can come later, if ever.

Every founder-marketer has a version of this story: metrics live in four tools, the report is a screenshot ritual nobody reads, and the honest answer to “is marketing working” is a shrug with charts. The problem is not missing data. It is that nobody turned the data into a decision.

A weekly report should do one job: tell you whether the machine is working and what to change. That takes five numbers, a short narrative, and at most two actions. AI makes this cheap to produce every week. The playbook below builds it in six steps, prompts included.

Reporting is a judgment problem, not a data problem

Marketers do not lack numbers, they lack verdicts. Content teams have named measurement among their hardest problems for years in the Content Marketing Institute’s annual research, and HubSpot’s State of Marketing research keeps finding teams under pressure to prove ROI with data spread across more tools than anyone can watch. Adding another dashboard does not fix that. A system that ends in a verdict does.

That is why this playbook spends as much effort on the “so what” pass as on collection. The model is genuinely good at the summarize-and-flag work: turning a log of numbers into readable prose and pointing at drift. What it must not own is the verdict. That split, machine handles volume while the human keeps judgment, is the core pattern of the marketing engineer role, and this build is deliberately the gentlest introduction to it.

Why this is the right second system

If you build only one system, build the research agent. Build this one second, because reporting is where systems skills become visible to everyone around you. A crisp weekly report is proof that the machine runs, and demand for exactly this skill is climbing: analysis of automation-heavy go-to-market roles shows reporting and data workflows near the top of what companies now hire for.

It is also the system that keeps the others honest. Your content pipeline and research agent both make claims about what is working. The weekly report is where those claims meet numbers, which is why it maps to days 61 through 75 of the 90-day career path.

The build, step by step

Six steps below. One and two are decisions, three through five are the machine, six is the habit. Ship the manual version this week: pick the numbers, paste them, run the prompt. Automating the collection is an optimization for later, and some operators never need it.

Implementation

Build it
step by step.

  1. 01

    Pick the five numbers that matter

    Most reports fail by including everything. Choose the handful of metrics that actually indicate whether marketing is working for your stage, and write down why each one earns its slot. If a number cannot change a decision, it does not make the report.

    Prompt 
    Help me choose the five metrics for my weekly marketing report.

My business: [ONE-LINE DESCRIPTION]
My current goal: [E.G. MORE INBOUND PIPELINE, LAUNCH TRACTION, CONTENT COMPOUNDING]
Channels I actively run: [LIST]
Metrics I currently look at: [LIST EVERYTHING, EVEN OCCASIONALLY]

From these, propose the 5 that belong in a weekly report. For each:
- Why it earns the slot, in one line
- The decision it can change
- A healthy range or direction for my stage, if you can infer one

Then list the metrics I should deliberately IGNORE weekly and check monthly
instead, with one line on why.

    Tip. Vanity metrics survive by hiding in crowds. A five-number cap forces every metric to justify itself against the others.

  2. 02

    Centralize the pulls

    Write down where each number lives, who or what fetches it, and when. One collection note or sheet, updated the same morning each week. Automate exports where the tool makes it easy, and paste by hand where it does not. Do not let plumbing stall the habit.

    Prompt 
    Here are my five weekly metrics and where each one lives:
[METRIC 1] - [TOOL / DASHBOARD]
[METRIC 2] - [TOOL / DASHBOARD]
[METRIC 3] - [TOOL / DASHBOARD]
[METRIC 4] - [TOOL / DASHBOARD]
[METRIC 5] - [TOOL / DASHBOARD]

Design my weekly collection routine:
1. The fastest manual path to each number, as a checklist in order
2. Which of these tools can export or schedule the number automatically,
   and how, in plain language
3. A simple template for logging the five numbers plus date, so history
   accumulates in one place
4. A realistic estimate of total collection time per week

Optimize for the routine surviving a busy week, not for elegance.

    Tip. Log the numbers in the same place every week even if collection is manual. Trend history is where reports get their power, and it only exists if the habit does.

  3. 03

    Write the summarize-and-flag prompt

    This is the engine. Paste this week's numbers and last week's log, and the model returns a short narrative with anomalies flagged: what moved, what is drifting, what looks off. Numbers become prose you can read in one minute.

    Prompt 
    You are writing my weekly marketing report. Be concise and neutral. Do not
celebrate or catastrophize.

This week's numbers:
[PASTE THE FIVE METRICS]

History log:
[PASTE PRIOR WEEKS OR "FIRST WEEK"]

Context worth knowing this week (launches, holidays, campaigns, outages):
[PASTE OR "NONE"]

Produce:
1. STATUS: one sentence on whether the week was normal, up, or down overall
2. THE NUMBERS: each metric with this week's value, the trend arrow, and
   a one-line note
3. FLAGS: anything that moved more than usual, with your best guess why,
   labeled as a guess
4. QUIET TRUTH: one thing the numbers suggest that I might not want to hear

Under 250 words. If it was an ordinary week, say so plainly.

    Tip. The 'quiet truth' line is the part you will come to rely on. Models are good at pointing at the drift you have been rationalizing.

  4. 04

    Force the 'so what' pass

    A report describes. A useful report decides. Every flagged item gets one of three verdicts: act now, watch another week, or accept and move on. This pass is yours, but the model can draft the options.

    Prompt 
    Here is this week's report:
[PASTE FROM STEP 3]

For each flagged item, lay out my options:
- ACT: the single most sensible action this week, sized under 2 hours
- WATCH: what specifically to look for next week that would settle it
- ACCEPT: the case for doing nothing, stated honestly

Recommend one verdict per flag and say why in one line. Then list what,
if anything, should change in next week's plan. Two actions maximum
across the whole report. If nothing warrants action, recommend none.

    Tip. The two-action cap is the discipline. A report that generates five todos a week becomes a report you dread, then a report you skip.

  5. 05

    Lock the one-screen format

    Design the final template once: status line, five numbers with trends, flags with verdicts, actions, done. The same shape every week is what makes it readable in a minute and comparable across months.

    Prompt 
    Turn this into my permanent weekly report template.

Here is a finished example week:
[PASTE THE OUTPUT OF STEPS 3 AND 4]

Produce a reusable template with:
- A fixed section order: status, numbers, flags, verdicts, actions
- Placeholder markers where each week's content goes
- A hard length budget per section so the whole report stays on one screen
- A two-line footer: what changed in the system itself this week, and
  anything to revisit at month end

Format it so I can paste it into my notes or a message to my team as-is.

    Tip. Resist redesigning the template when a week looks bad. Changing the format to soften a trend is the report lying to its only reader.

  6. 06

    Schedule the loop and review monthly

    Put collection and reporting in the same calendar slot every week, ideally before you plan the week. Once a month, run a review pass on the report itself: are these still the right five numbers, and did the actions from past reports actually happen?

    Prompt 
    Here are my last four weekly reports:
[PASTE]

Run the monthly review:
1. Are these still the right five metrics for my current goal? Propose
   swaps only if a metric changed no decisions all month.
2. List every action the reports recommended. Which happened, and what
   resulted?
3. What is the one-paragraph story of the month, written for me in
   January looking back?
4. Anything the weekly format is hiding that a monthly view reveals?

Be specific. The goal is a system that gets sharper, not a ceremony
that repeats.

    Tip. The action audit in question 2 is the highest-value minute of the month. Reports earn trust by being connected to things that actually happened.

Free bonus

The Weekly Report Kit

The six prompts in this playbook build the loop. This kit is the finished versions: the template filled in, and the prompts chained so one paste runs the whole report.

Common questions

Quick
answers.

How do I automate a marketing report with AI?

Pick five metrics that can change decisions, log them in one place on a weekly schedule, then run a summarize-and-flag prompt that turns the numbers into a short narrative with anomalies highlighted. Add a verdict pass for flagged items and keep the whole report to one screen. The assembly automates first; deeper data plumbing can follow later.

Do I need a dashboard tool for this?

No. Dashboards show numbers; this system produces judgment about them. A simple log of five metrics plus the prompts in this playbook outperforms an unread dashboard, because the output is a narrative with flags and verdicts, not another screen to interpret. If you already have a dashboard, it becomes one of your collection sources.

Which metrics should a small B2B company report weekly?

It depends on the goal, but the shape is consistent: one demand signal like qualified inbound, one attention signal like organic or AI-search visibility, one engagement signal like replies or booked calls, one pipeline signal, and one delivery signal like content shipped. Five total, each tied to a decision it could change.

How long does the weekly loop take once it is set up?

Typically 15 to 25 minutes: a few minutes collecting the numbers, a minute for the model to draft the report, and the rest reading, setting verdicts, and deciding the week's actions. The one-time setup, choosing metrics and locking the template, is an hour or two.

Where to next

Build it, or
have it built.

Want to build it yourself

Take the shortcut.

Get one practical build like this in your inbox each week. The prompts, the workflow, the wiring.

Free weekly playbook