Your marketing is working if the outcome each piece of work promised in advance actually arrived. Judge it one project at a time. Before the work ships, somebody writes down which single number should change and by when.
Afterwards, you look at that number together. Work that carried no advance promise cannot be graded later.
The monthly marketing review is a strange hour. Somebody walks you through twelve slides. Traffic is up, impressions are up, the email list is bigger than it was.
Three new pages went live and a video did better than the last one. Everyone seems pleased. You say thank you, the call ends, and you have no idea whether any of it was worth what it cost.
Why the review leaves you with nothing
The gap is vocabulary rather than intelligence. You can read a balance sheet and tell inside ten minutes whether an engineering estimate is honest.
Marketing is the one function where you have no instinct for what good looks like. So you cannot tell which of the twelve slides deserved a question, and the questions stay in your head.
The few that escape come out wrong. "Is any of this actually doing anything?" lands as an accusation, you soften it to keep the relationship intact, and the review goes back to being a slideshow.
Left alone, this runs eighteen months without much trouble. Every month the charts are green and the bank balance sits where it sat the previous spring. Everybody has been telling the truth about numbers that were never wired to money.
How can I tell if my marketing is working?
You do not need to know how marketing is done in order to judge whether it worked. Those are two separate skills, and the second one is far cheaper to acquire than founders assume.
You cannot perform surgery, and you can still ask a surgeon what recovery ought to look like at week two, then notice when week two arrives and looks nothing like that.
Auditing a claim takes less expertise than making one. That gap is the founder's leverage.
Founders skip that step with marketing alone, because the work arrives described as a list of things that were made. Twelve slides of output, zero stated predictions, nothing that can be marked right or wrong afterwards.
The correction is small. Somebody writes one sentence about what a piece of work should change, and later you both look. Most of the time that sentence is hard to write, and the difficulty is itself the finding.
Question one: what was this supposed to do?
Start with something already finished. Pick one item off the slides, ideally the one you understand least.
Ask what number would have to move for this to have been worth doing.
A usable answer contains three things: a single number, a direction, and a date. "We expected trial starts from organic search to go from about 40 a month to 60, within two months of the new pages being indexed."
You may have no idea whether 60 was a sensible target, and you do not need to know. You now hold a claim with a deadline on it.
An evasive answer takes you on a tour. It describes the work in loving detail and reaches for a figure that was already climbing before the project started.
The reliable tell is how long the answer takes. Somebody who was aiming at something gets it out in a breath. Somebody who was aiming at nothing needs a warm-up.
Then there is the third case, more common than either: work that never had a target and never pretended to. Marketing teams inherit a queue of jobs that arrive as requests rather than as bets.
Refresh the website. Post more. Do something for the conference.
Every one of those is a fair request. None had a number attached, so none can be graded, and at year end nobody can say what the year bought.
How to introduce this without it landing as distrust
The habit is easy to install badly. Announce that every project now needs a projected result, and you have handed a performance review to people who thought they were doing fine.
So say up front that the first pass is a baseline, and put yourself on the same hook. Half the untargeted projects on those slides exist because you asked for them in a hallway on a Tuesday.
That converts the exercise into a review of how work gets started, and it gets you an honest inventory instead of a defended one.
Question two: could that number have gone down?
This is the shortest of the three and it does the most work. Take whatever number the team offered and ask whether it was structurally capable of falling.
Some cannot: total followers, emails sent, posts published, leads generated all time. Each is a running tally of work performed, and it moves in one direction for as long as anybody keeps working.
A dashboard that only ever goes up is measuring effort.
Other numbers can go the wrong way. That is what makes them useful. Trial to paid conversion drops when the onboarding gets worse, and cost per paying customer climbs when the audience is worn out.
Each of those can deliver bad news, and only a number that can deliver bad news makes good news mean anything.
So the test fits in a breath. Could this have gone down? If the honest answer is no, ask what the activity was supposed to cause, and go find whichever number would have registered it.
Check one thing first. Sometimes the reporting cannot see past a signup, so a signup is the furthest thing anyone on the team can honestly claim. That is a wiring problem with how conversions are tracked, and engineers fix it.
Question three: is this counted the way it was last quarter?
The third question catches the most embarrassing failure of the set, because it is the one where everybody was honest and the number still misled the room.
A metric improves when the thing being counted quietly changes shape. The chart takes one step up in the month the wording changed, then carries on at its old slope.
Unless somebody says it out loud, that step reads as a result.
Put illustrative numbers on it. A team reports qualified leads at around 40 a month for a year. In April the criteria are widened to count anyone who downloaded a guide.
May comes in at 95. The slide shows a 138% improvement and the team is congratulated. Nothing changed in April except one line in a spreadsheet.
The question carries no accusation in it, because the change is nearly always legitimate and simply went unannounced. Ask whether the number is defined the way it was the last time you looked at it.
Ask it every time a chart takes an unusually clean step. Then ask for definition changes to be marked on the chart from now on. That costs a minute and removes the problem permanently.
What should I ask my marketing team?
Run the three in order, on one item at a time, and stop when one fails. Without a stated target there is nothing to sort into activity or outcome.
There is one more failure worth naming, and it is yours. If you run these questions monthly and nothing changes between reviews, the answers have nowhere to land.
Decisions about spend need a shorter loop than a month. That is what a standing weekly slot provides: the same handful of numbers, faced on a schedule, by whoever can act on them.
These three questions judge individual pieces of work. They sit underneath the short set of numbers that judges the whole machine, and underneath a full teardown of the growth system. Start here: it costs an hour and needs access to nothing.
What changes once you can ask
The first month is uncomfortable. You will ask about six items and get clean answers on one. That ratio is the finding: most of what has been shipping was never designed to be checked.
By the third month the conversation has changed shape on its own. The team starts bringing the target with them, unprompted, because writing it down in advance is far easier than reconstructing it under questioning.
That single behavior change is where most of the return sits. Work designed with a check waiting for it gets designed differently.
And you get the thing you actually wanted out of that hour: the ability to sit in it and know, honestly, whether to fund more of what you just saw.
By the time a founder calls me, they have usually sat in that meeting for a year and stopped believing it somewhere around month eight. Running these three questions across the last two quarters of work is where we start. Let's talk.