The three numbers in your marketing report that are usually wrong

There is a structural problem in marketing reporting that has nothing to do with anybody’s integrity, and pretending otherwise makes it harder to fix.

The report is produced by the people being measured. The agency runs the campaigns, the agency holds the platform access, the agency builds the deck. Nobody has to lie for that arrangement to bend quietly toward the flattering interpretation. It only takes a definition that nobody wrote down.

Three definitions do most of the damage. All three are defensible if you squint. All three move the number in the same direction.

One: your lead count counts touches, not people

Ask a marketing platform how many leads you got and it will tell you how many conversion events fired. That is not the same question.

Two real behaviours break the equivalence, and both are large enough to see in production data:

People contact you more than once. Somebody calls, does not get through, calls again an hour later. That is one person and two conversion events.

People switch channel mid-decision. Somebody fills in your contact form, gets impatient, and phones about the same job. Every attribution system on the market treats that as two conversions, frequently credited to two different channels.

Deduplicate properly against the actual contact records and roughly 30% of raw contacts disappear. Not because the tracking was broken, but because the tracking was answering a different question from the one you asked.

The correct definition is narrow and worth writing down: a new lead is a distinct person whose first contact with the business falls inside the reporting window. Not a session. Not a form fill. A person.

If your reporting has never made that distinction, roughly a third of your lead volume is an artefact.

Two: your cost per lead is divided by the wrong denominator

This one is the most expensive, and it is the easiest to do by accident.

The natural calculation is total ad spend divided by total leads. It is one line in a spreadsheet and it looks like an efficiency metric. It is not. It is a blend of your paid performance and your organic performance, expressed as though it described your paid performance alone.

The correct denominator is only the leads that ad spend actually produced. Everything else, your organic traffic, your referrals, your repeat customers, your reputation, is not something your media budget bought.

The size of the error depends on your organic share, and on a real account it understated true cost per paid lead by 61%.

Sit with that number for a second. The account was not underperforming by a rounding error. The reported figure was less than half of the truth, and it was wrong in the direction that made the media buying look efficient. Nobody falsified anything. The spreadsheet just divided by everything.

If you take one thing from this note, make it this: ask whoever produces your reporting what the denominator is. If the answer is all leads, your cost per lead is fiction and every optimisation decision made from it is unanchored.

Three: the wrong channel is getting the credit

Most reporting credits the last touch before conversion. It is the default in most platforms because it is the easiest thing to observe.

It is also close to the least useful. The last touch before a conversion is very often branded search: somebody who already knew your name typed it in and clicked. Crediting that click with the acquisition is like crediting the cashier with the customer’s decision to shop there.

Meanwhile the thing that actually created the demand, the campaign that got you into consideration months earlier, gets nothing. And because it gets nothing in the report, it gets less budget next year.

This is the mechanism by which measurement quietly destroys the marketing that works. Not through malice. Through a default setting.

Credit the first touch. What brought this person in, not what they clicked on the way to a decision they had already made.

That is not a perfect model either. No attribution model is. But it errs in the direction of understanding what created demand, and last-touch errs in the direction of harvesting demand somebody else created.

Why all three fail the same way

Look at the direction of the errors.

• Touch counting makes your lead volume look about a third larger than it is.
• Spend over all leads makes your paid channel look up to twice as efficient as it is.
• Last-touch credit makes branded search look like your best performer.

Three independent methodological choices, and all three flatter the report. That is not coincidence, and it is not fraud. It is what happens when the person choosing the definitions is also the person the definitions describe. Each choice is individually defensible. Nobody ever has to make a decision they would be embarrassed to explain. The bias lives in the aggregate.

The fix is not to find more honest people. It is to publish the definitions and apply them the same way in a bad month as in a good one.

What to actually do

You do not need to buy anything to start on this. Ask four questions of whoever produces your marketing reporting, and ask for the answers in writing:

1. What is the definition of a lead in this report? Specifically, is it deduplicated to distinct people, and across which channels?
2. What is the denominator in cost per lead? All leads, or only leads attributable to paid?
3. Which attribution model is this, and what is the lookback window? If the answer is a shrug, the answer is last touch.
4. Are inbound vendor and sales solicitations excluded, and how many were excluded this period?

Those four questions are not adversarial. A good agency will welcome them, because a report that survives them is worth considerably more than one that has never been asked.

If the answers do not come back, or come back vague, you have learned something too.

The underlying principle

A number nobody can check is an opinion with a decimal point.

The way out is not more dashboards. It is fewer numbers, each with a published definition, applied identically whether the month went well or badly. That is a lower-drama form of reporting and a much harder one to argue with, which is exactly the point.

We publish every metric definition we use, for the same reason. If a definition is doing quiet work in somebody’s favour, it should be possible to catch it.