The Growth Cockpit: The Weekly Marketing Report That Replaces the Status Meeting

What should a weekly marketing report include?

Monday, 9:30. Eight people are on a call. One of them shares a screen and reads the dashboard out loud for forty minutes.

Clicks are up. Cost per lead is down a bit. Nobody knows why, so the call ends the way it always ends: we will keep watching.

The dashboard was already watching. That was its whole job.

A weekly marketing report is the one page that tells your team what changed in marketing last week, why, and what to do next. Most teams picture a dashboard or a slide deck. We mean a memo that makes decisions.

We call ours the growth cockpit. It is one of the systems a marketing engineer builds: scripts pull the numbers, an AI drafts the memo, and a person signs it before Monday.

Here is the method in one breath. Every line says what changed, the likely cause with proof, and the call. Check the data before you explain anything, and grade last week's calls before you make new ones.

Get only a handful of leads a month? Run the same memo monthly. Weekly numbers that small are mostly noise.

Put the decision in the message, not behind a link

A dashboard waits for someone to open it. A memo shows up with the answer already inside. That small difference changes what people do.

The cleanest proof comes from government letters, not marketing. In Writing for Busy Readers, Todd Rogers and Jessica Lasky-Fink describe a test on 5,873 Medicare recipients choosing drug plans.

Half got a letter that pointed them to a website with their personal options. Half got a letter that printed those options, the cheapest plan and the savings. Switching to a better plan went from 17 percent to 28 percent.

Same facts. One group had to go and fetch them.

Donella Meadows tells an older story in Thinking in Systems. Identical Dutch houses used far less electricity when the meter sat in the front hall, where people walked past it, instead of the basement. She tells it as a story she heard, not a study, and so do we.

Our reading for marketing is simple. A dashboard is the meter in the basement. The Monday memo moves it to the front hall.

Who reads a dashboard aloud on purpose? Nobody. They do it because nobody wrote down what it meant.

Change-Cause-Call: every line makes a decision

We write every line of the memo with one rule: Change-Cause-Call. Three parts. Always in this order.

The change is what moved, by how much, and against what. The cause is the most likely reason, with a receipt: a link, a quote or a log entry. The call is keep, stop or test, with an owner and a target for next Monday.

Here is one line from a made-up kitchen remodeling company.

Booked estimates hit 31, above the eight-week high of 28. A text reply to every new lead within five minutes went live Tuesday, and leads reached within an hour rose to 81 percent from 52. Keep it; owner, the office manager; target, 30 or more next week.

The idea is older than AI. Gary Lilien and Arvind Rangaswamy coined the term marketing engineering for decisions built from data, models and judgment together, not gut feel alone.

Their textbook contrasts two managers. The first sees sales down 2.3 percent and asks for 10 percent more promotion money, quietly betting that more spend wins the sales back. The other feeds the new week into a model and names the spend that earns the most profit.

Today's marketing engineer goes one step further. We build the system that writes the second manager's answer every Monday.

The cause is where AI gets dangerous. Ask a model why a number moved and it will always find a reason. Like a TV pundit after a loss, it is never short of a theory.

Statisticians named this trap long ago: Twyman's law. The more surprising a number looks, the more likely the measurement broke.

So the rule stays short. No receipt, no cause.

Lead with one number and five or fewer that feed it

Most weekly reports fail before line one. They carry forty numbers, and nobody can watch forty numbers.

Start with the one number the business runs on: booked calls, signed jobs, paid orders. Alistair Croll and Benjamin Yoskovitz, in Lean Analytics, call it the One Metric That Matters: one number above all others for your current stage.

Their Moz example shows the shape. One headline number, new paying customers minus cancels, sat on top of a few numbers that fed it. Even so, an investor pushed the team to track fewer, because nobody can move dozens of numbers at once.

Then add the inputs that feed your one number, five at most. An input is something your team can push this week: leads, contact speed, show rate. Revenue arrives late and explains nothing on its own.

That split is the core of the weekly business review described in Working Backwards by Colin Bryar and Bill Carr. An output tells you growth slowed. An input like new customers tells you why, and much sooner.

Give every number one owner, a normal range and a target you can actually hit. A target nobody can reach turns red every week. A number that is always red works like a car alarm on a busy street: everyone stops hearing it.

Set targets from your best recent weeks, not your average. Meadows warns that goals tied to past performance slowly slide down until bad feels normal.

Forty numbers is not rigor. It is a hell of a place to hide.

Separate noise from news before explaining anything

Most weekly movement is noise. Numbers wobble for reasons nobody can fix, and explaining a wobble wastes everyone's morning.

Our rule is to show every number three ways: this week, last week, and the average of the four weeks before.

Then mark the normal range: the lowest and highest week of the last eight. Inside the range, the line says noise and moves on. Outside it, or three weeks in a row in one direction, it is news and earns a cause.

Amazon's review has a phrase for the first case: nothing to see here. A control chart does the same job with more math; our eight-week range is the simple version.

Check the data before you read it. A broken tag or a half-loaded day looks exactly like a great week. Even the Search Console API warns that its newest days may still change.

Tests get the same discipline. Report a test only when it reaches the size you set before it started.

Why so strict? Evan Miller showed that if you check a test after every visitor and stop at the first good result, a change that does nothing looks like a winner 26.1 percent of the time. You were told to expect 5.

Running tests get a name and a stop date. That is all.

The ten-line Monday memo, in reading order

The memo has ten lines and reads on a phone in about two minutes. The order matters.

Line 1 is the verdict: one sentence the boss could act on without reading further. That is the army's bottom line up front rule, applied to marketing.

Line 2 is data health, because everything below depends on it. Lines 3 to 6 are the evidence: the numbers, finished tests, what customers said and what your own team changed.

Finished tests come from your creative testing engine. What customers said comes from your customer truth system, with quotes and links.

Line 6 is the one teams skip. If nobody logged that the page headline changed Tuesday, nobody can explain Wednesday.

Lines 7 and 8 are the decisions. Line 8 holds anything that spends money, reaches customers or changes tracking, written as one yes-or-no question.

A narrow question gets a fast answer. "Approve moving $400 a week to search, yes or no?" gets a reply. "Thoughts on budget?" gets a meeting.

Line 9 is what moved outside: competitor launches, price changes, search positions. Line 10 grades last week's calls.

Bold one thing only: the verdict. In tests from Writing for Busy Readers, one highlighted instruction was followed by 84 percent of readers. Highlight five sentences and that drops to 65 percent.

If it needs scrolling, it is a newsletter.

Grade last week's calls, or the memo never learns

Line 10 turns a report into a system. Every call from line 7 goes into a dated decision log with its owner and target.

Next Monday the memo reads that log and grades each call: yes, no, or too early to tell. After a quarter you have twelve weeks of graded decisions instead of twelve weeks of slides.

Why bother? Unchecked promises die quietly.

Gino Wickman, author of Traction, saw it in his own client meetings. Of ten weekly promises, only a few got done. Once each one had to be reported the next week, nine of ten got done.

The approver's edits teach the memo too. Every cause they cut and every number they fix goes into the agent's rules file, so the same mistake does not come back next Monday.

Keep those rules concrete. "Ignore the last two days of search data" works. "Be careful with data" does nothing.

The rules file and the log live in your growth repo, next to the job specs. A chat window forgets. A folder does not.

Let AI draft the memo and a person sign it

A marketing engineer can build the first version in about a day, if the data already sits in tools with exports. The order we use is below.

Give the agent read-only access. A memo writer has no reason to touch a budget, so the rule from our marketing engineering stack applies: read wide, write narrow.

Write the agent a job spec, not a one-line prompt. The format is in How to Write a Job Spec for an AI Agent.

Backtest before you schedule. Run it on last week, a week you already know, and count every wrong claim. Fix the spec and rerun until the count is zero.

Then keep a person in the pilot's seat. It is a cockpit, after all. Andy Grove argued in High Output Management that writing a report matters more than reading it, because writing forces you to be precise.

Our version for the AI era: the machine gathers and drafts. The owner of each number still rewrites its cause and its call. Skip that, and you outsource the thinking along with the typing.

The field's founding textbook drew a similar line: automate a decision only when a wrong one is cheap. A headline swap is cheap. A budget shift is not, which is why it waits in line 8.

Does a weekly memo replace the Monday meeting?

Not entirely. It replaces the reading aloud, and it gives the meeting a better job.

Look at the weekly meeting in Traction. It runs 90 minutes, and only 20 of them are reporting: the numbers, the quarterly goals, the headlines and the to-do list. The other 60 go to solving problems.

The rule inside it is the one we keep: report, never discuss. Anything off track drops to the problem list for later in the meeting.

The memo does that reporting before anyone joins. So the call starts with line 8 and ends when every off-track number has an owner.

How to waste a Monday in four easy steps. Step one: share your screen. Step two: read every damn number aloud, including the ones that did not move.

Step three: explain each wobble with a confident theory nobody checks. Step four: end with "we will keep an eye on it" and book the same call for next week.

That is not a meeting. It is a very expensive screensaver.

Low volume changes the cadence, not the method. With only a few leads a week, run the memo every two or four weeks and compare equal periods.

Run your first Monday memo this week

The prompt below drafts the memo for you. Paste it into your preferred AI with last week's numbers, your change log and your test log. It asks for anything missing and never guesses a number.

This week: pick the one number and write its last eight weeks by hand. Then list what your team changed last week, with dates. Then run the prompt and have each number's owner rewrite its cause and call.

If the ad numbers in your memo look strange, read Read 'Results', Not 'Leads' before you trust them. If you would rather have the cockpit built, it is part of our marketing engineering work.

Back to that Monday call. The dashboard never needed a narrator. It needed one page that said what changed, why, and what to do.

A number without a cause and a call is wallpaper.