Analytics says «conversion». But did a customer arrive?

The report says eighteen conversions last month. A fine number.

Now ask yourself something harder: in how many of those eighteen did you speak to a person, or get paid? If the answer does not come straight away, the problem is not your memory.

We have written elsewhere that this comparison, between what the tool counts and what actually reached you, is worth making. Here we go one step further, to the useful part: exactly where measurement breaks. Because it does not break where people assume. It breaks at four specific points, and all four look perfectly normal until someone looks at them closely.

First, though, let us agree on what the words mean, because that is where the misunderstanding starts.

What «conversion» actually means

Measurement tools do not know what success means for your business. They have no way of knowing.

They know only what you told them, or what whoever set them up told them. «When a visitor reaches this page, write a conversion.» «When they click this button, write a conversion.» That is all. The tool faithfully counts an event inside your site, and you read the result as though it counted customers.

Between that event and an actual customer there is a distance. Let us see at which points along it the customer gets lost.

One: a tap on the phone number is not a call

This is the most common and the most treacherous, because it affects precisely the businesses that live off the phone.

The visitor sees your number on their phone and taps it. At that exact moment the conversion is recorded. Whatever happens next happens outside your site, and the tool sees none of it.

And plenty happens next. They cancel before pressing call. The line is engaged. It rings and nobody picks up, because it is Saturday afternoon. Or they simply drag a finger across the screen and hit it by accident, which on a phone happens far more than you would imagine.

None of that is subtracted from the number. And the first three cases, the cancelled call, the engaged line and the unanswered ring, are lost customers who already had the phone in their hand for you.

The useful move here is not to stop counting the tap. It is to name it properly: these are not «calls», they are «call attempts». And if the attempts are many while the phone is not ringing, you do not have a marketing problem. You have a problem with who answers, and we have been here before.

Two: clicking «Send» is not submitting

Here the difference is one word, and that one word costs money.

If measurement is built on the button click, then it also records every time the form never went anywhere. They missed a required field. They typed the email wrong and the field objected. The site froze at the moment the button was pressed. Or the message went out perfectly well and landed in your spam folder. In all of those the visitor clicked, the conversion was written, and nothing reached your hands.

The check is simple and takes five minutes. Submit your own form properly and see whether the message arrives. Then submit it deliberately wrong, leaving a required field empty, and watch whether a conversion is written again. If it is, you are not measuring submissions. You are measuring button clicks.

Three: the thank-you page counts every time anyone sees it

A lot of measurement is set up the easiest way: a conversion is the visitor reaching the page that says «thank you, we have your message».

It works beautifully, until you remember how people behave. Someone refreshes the page to be sure it went through. Someone goes back and then forward. Someone has it in their history and opens it again days later, looking for their order number. Someone forwards it to their partner so they can see it too.

Each of those is counted as a fresh visit to that page, and therefore as a fresh conversion. A single order can be counted three or four times that way. You watch the number climb and assume more customers came, when it was the same one opening the page four times.

Four: your own tests stay inside the numbers

This is not a fault. It is dust that gathers over time.

Every time you test whether the form works, it counts. Every test order you pushed through to see whether the receipt comes out right, it counts. Every visit from the person building your site, it counts. The day you sat with your accountant showing them how your shop’s checkout works: every click you made was counted.

In a large shop all of that disappears into the noise. On a site with twelve conversions a month, ten tests of your own are nearly the whole total. And because they happen on particular days, they also build false peaks that somebody later tries to explain with advertising.

The check sheet, three columns and nothing else

All four, together with two more cases worth looking at, are checked with one table you can build in an afternoon. It has three columns:

The last two rows do not belong to the four cases we went through. They are there because in an online shop they hide exactly the same kind of error: an order that completed without being recorded anywhere, and an abandoned order counted as though it had gone through.

You perform each of those actions yourself, in order, and fill in the two right-hand columns. The second one you read inside the measurement tool. The third you read in your inbox, on your phone and in your orders.

Wherever those two columns do not agree, you have found the problem. And you have found it precisely, which is the whole point: you are no longer saying «something is off with the numbers», you are saying «the form records a conversion even when the message does not go out». The second sentence leads to a fix. The first one leads nowhere.

And once it is fixed, the numbers will fall

I have to say this, because it is the moment many people back out.

After a proper check, conversions almost always shrink. It is unpleasant to watch eighteen become nine. But the nine is real, and on real numbers you can decide whether the advertising is worth it, whether the new page works, whether anything changed when you changed the copy.

On the eighteen you can decide nothing, because you do not know which part of it is real.

If you would rather first see what the tools you already have are saying, we have explained it for non-technical readers and it costs nothing. If you want measurement running on our own server, leaving no third-party cookies on the visitor’s computer, there is MS-Analytics, and I say that with the caveat it deserves: it is ours.

Whichever road you take, the question stays the same, and it is better than any report: of the things the tool calls successes, how many knocked on your door?

If your conversion numbers do not match the calls and messages that actually arrive, we check it for you. We look at what your measurement records today, fix whatever counts wrong and hand you the same checklist filled in, so you know which conversions correspond to real customers. Send us through the contact form your site’s address and the measurement tools you use. Correct measurement is the first step of every campaign in our Google & Social Ads service.