Marketing Systems Audit: 20 Questions That Show Where Your Growth System Breaks

What a marketing systems audit should tell you

Someone asks where last month's leads came from. The marketing lead says the CRM tracks that. Easy.

So you open five leads. Three of them say Unknown.

The tracking was set up. Nobody checked it.

That gap is what a marketing systems audit checklist is for. Ours has twenty yes-or-no questions, and a yes counts only if someone opens the proof in five minutes.

Most marketing audits grade your ads, your site and your copy. This one grades the machine behind them. Do leads, feedback and decisions move on their own, or only when one person remembers?

A low score is not an insult. It is a short list, in the order you should fix it.

Marketing engineering means deciding with data, then building the tools

The term is older than the job title. Gary Lilien and Arvind Rangaswamy used it in the late 1990s, and their textbook Principles of Marketing Engineering and Analytics, written with Arnaud De Bruyn, gives the definition.

In plain terms, marketing engineering means turning data, know-how and judgment into tools that help marketers decide better. The authors set it against two other modes. Conceptual marketing decides from the gut, and automated marketing hands the whole decision to software.

Their rule for choosing still works. Automate a decision when a wrong call is cheap, like a product recommendation or a keyword bid. Keep a person in the loop when a wrong call is expensive, like how the company is positioned.

We add one thing: the building. The book already says the models have to be built and put to work inside support systems.

Today those systems are pipelines, agents, tests and reports. Someone has to build them, and that person is the marketing engineer described in what a marketing engineer is.

So what does an audit check? Whether those systems exist. And whether anyone can prove it.

A yes counts only if you can open the proof in five minutes

The Five-Minute Proof is our scoring rule. A yes counts only when someone opens the evidence on screen within five minutes: a file, a log, a report or a dated note.

Self-scored checklists are not new. The Organizational Checkup in Gino Wickman's Traction has owners rank 20 statements from 1 to 5, turns the total into a percentage, and repeats every 90 days.

We like the format. We do not trust the self-rating.

Wickman writes that most organizations run below 50 percent, on a checkup they score themselves. Picture the score when every yes needs a receipt.

Software researchers hit the same wall. The Accelerate team of Nicole Forsgren, Jez Humble and Gene Kim never asked teams if they practiced continuous integration, because nearly everyone says yes. They asked whether automated tests start each time new code goes in.

Ask about the behavior, not the label. "Do we track lead source?" gets a yes. "Open five leads from last week" gets the truth.

"We have that somewhere" scores zero. So does "Sam handles it," unless Sam's method is written where the next person can find it.

Engineers call that risk the bus factor: how many people could leave before nobody knows how a thing works. Any yes that lives in one head has a bus factor of one.

Check the boring basics first, because that is where the misses are

Rare problems are fun to find. Basic misses cost more.

Atul Gawande tells this story in The Checklist Manifesto. A Johns Hopkins doctor asked ICU nurses to watch, for one month, how often doctors followed five well-known steps for placing a central line. In more than a third of patients, at least one step was skipped.

Later, Gawande's team ran a World Health Organization pilot that tracked six basic surgical safety steps in eight hospitals. The best one missed a step about once in every 16 patients. On average, the hospitals missed at least one step in two of every three patients.

Most of those hospitals had no idea of their own rates until someone counted. Our reading: marketing teams are no different.

So the audit opens with four foundations: lead source on every lead, data checked for blanks, a report someone can rebuild, and a spend plan tied to revenue. Skip them and no other score can be checked.

The data check matters more than it sounds. The Lean Analytics authors pass on a data scientist's first warning: never assume your data is clean, and ask whether a tracking bug is blanking out 30 percent of your numbers.

Question 4 comes from the marketing engineering textbook. Its authors call a plan "dumb" when ad spend shows up only as a cost line, with nothing tying it to sales.

The 20 questions of the marketing systems audit

Here is the full list, in five groups. Copy it into a sheet with four columns: score, link to the proof, owner and date.

Each line scores 0 or 1. A half-yes is a no.

The groups are not random. They climb from the basics to the deepest causes, and the next section explains why that order decides what you fix.

Each zero also has a build. Question 8 is server-side conversion tracking, questions 9 and 10 are the customer truth system, and question 7 is the growth cockpit.

Questions 16 and 19 live in the creative testing engine. For a sense of pace, the authors of Trustworthy Online Controlled Experiments say a team just starting runs about one test a month, and each stage after that runs four to five times more.

Rank what you find by how deep the fix goes

A list of twelve zeros is not a plan. You need an order.

Donella Meadows gave us one. In Thinking in Systems she ranks twelve places to intervene in a system, from weakest to strongest.

Numbers sit dead last: budgets, rates, settings. She guessed that 90, then 95, then 99 percent of our attention goes there, for very little effect.

Delays, information flows, rules and goals sit higher up the list. Each one changes more of the system than the one below it.

Her favorite example is a Dutch housing estate. Homes with the electricity meter in the front hall used 30 percent less power than the same homes with the meter in the basement.

Nothing changed except who could see the number. She calls missing information flows one of the most common reasons a system fails.

Our reading for marketing: most audits stop at the bottom rung. Bids, budgets, button colors.

Our groups map onto her list. Speed is her delays, feedback is her information flows, rules are her rules, and goals are her goals.

The rule: fix any zero in the foundations first. Then fix the zero highest on the ladder that your team can change this quarter. Higher levers do more, but they are harder to move.

Find the cheap step that is holding back an expensive one

Andy Grove ran Intel, and in High Output Management he treats any kind of work as a production line. Every line has a limiting step, and the whole flow should be built around the right one.

His sharpest example is the justice system seen as a factory. By his rough math, each conviction cost well over a million dollars, while a jail cell cost about $80,000 to build.

So when cells ran short, the cheap step limited the expensive one. Grove calls that letting the wrong step limit the process.

Marketing does this every day. You pay for a lead, then lose it to a reply that takes a day. Same mistake, smaller numbers.

Picture a company that pays $150 for each lead. A reply within five minutes costs a few minutes of someone's time. If slow replies are what cap sales, the cheapest step in the chain is limiting the most expensive one.

Grove adds a second rule: catch a defect at the lowest-value stage. A junk lead is cheapest to stop at the form, costs more on a sales call, and costs the most after a contract is signed.

If question 5 scored zero, start with speed to lead.

Watch for the one person everything waits on

Some zeros hide inside a yes. The work gets done, but always by the same person. One person is not a system.

Matthew Skelton and Manuel Pais list the warning signs in Team Topologies. Requests keep going to the same people, even when they are busy or away. Other teams wait on one team, and people complain that nothing is written down.

They wrote that about software teams. Our reading: marketing teams show the same signs, usually around the CRM, the website forms and the ad accounts.

That is why question 13 asks for proof from an absence. When the owner was last away, did the work still ship?

If not, the fix is a written process plus a second person who has run it once. How to write that process so a new hire or an AI agent can follow it is in the agent job spec.

What a 7 out of 20 means, and what to fix first

Add the points. The bands below are our rule of thumb, not a benchmark.

Here is an example company at 7. It tracks lead source, its spend plan shows cost per customer, and it notices a broken form within a day. It also uses customer words in ads, names owners, writes processes down and keeps a test log.

Everything else is a zero. Nobody checks the data, the report cannot be rebuilt, and replies take hours. Closed deals never reach the ad platforms, and outbound is judged on emails sent.

Foundations come first, so this month it cleans the data and scripts the report. Then it changes what it counts: outbound moves from emails sent to replies that became meetings.

That second fix costs almost nothing. It is also the highest lever on the sheet.

Count sends and you get more sends. That is Goodhart's law: once a measure becomes a target, it stops measuring well.

How to ace a marketing audit without changing a thing, in four easy steps:

1. Let the head of marketing score it alone, from memory.

2. Count "we have that somewhere" as a yes.

3. Fix the lowest score by buying a dashboard.

4. Never run it again, so the number cannot drop.

That is not an audit. It is a compliment you paid for.

Re-score in 90 days, and set each new target from your best recent week, not your average one. The change in the total is your first honest number on whether the work is working.

Is this audit just an agency lead magnet?

Partly, and we would rather say it first. Our paid work starts with an audit, and we explain why audits make good offers in customized audits as lead magnets.

But the 20 questions work without us. There is no form on this page, and your own team can answer every line in an afternoon.

Score 16 or more and you do not need outside help for this. Spend the money on your next test.

Want a smaller paid first step? The shape we prefer is in quick win offers.

How do you run a marketing systems audit in 90 minutes?

Get three people in a room for 90 minutes: whoever owns marketing, whoever owns sales, and whoever lives in the CRM.

Go line by line. The owner opens the proof on screen or the line scores zero. Proof or zero, five minutes max.

Then paste the prompt below into your preferred AI with your answers. It ranks the zeros by how deep the fix goes and flags any cheap step that is holding back an expensive one.

It never invents a score. Every yes needs proof you name, and every cost needs a number you give it.

What to do this week

1. Book the 90 minutes with the three people. Ask them to bring proof, not opinions.

2. Open five leads from last week before the meeting. You will know your first answer before anyone sits down.

3. Pick one fix, name its owner, and put the re-score date in the calendar.

Remember the three leads marked Unknown? That was not bad luck. That was your score, showing up early.

When the fix is plumbing, like lead routing or CRM fields, that work sits under CRM and RevOps. Once a system runs, grade it with how to evaluate a marketing engineer's work.

An audit from memory is a mirror. An audit with proof is an X-ray.