You are a month into a campaign. The dashboard shows conversions, the count is going up, and you still cannot say whether any of them became work you were paid for. Nothing here is broken. The report is telling you the truth about the wrong thing.

The form was never the hard part

Forms are the part everybody complains about. They have not been the expensive part in my experience.

When I have run ads against a site somebody else controlled and asked for a thank-you page — a page that loads only after a submission succeeds, which is a cheap and honest conversion signal — it took about a week. Sometimes two. It went on somebody's list and it got done.

Where it is not available, the fallback is to measure the button press. That number counts a click on Send whether the message left the browser or died on a blank required field, so the real cost per enquiry sits above the reported one by an amount I cannot state — which belongs in the report rather than left for somebody to discover.

But a week is a week. If the form were the whole problem, this would be a scheduling complaint.

The phone call is measurable, and that is not the same as easy

Plenty of the businesses worth advertising for are phoned, not emailed. Somebody searches, reads two paragraphs, and calls the number in the header.

Calls placed straight from the ad are tracked by the platform without touching the site. The ones that come after the click, from the number on the page, are the harder case. Measuring those needs a number that changes per visitor, a script with permission to rewrite it in the page, and somewhere for the result to land. Every one of those is a change to a website, not a setting in an account.

I have built it on a site I did not manage, and what it took was not skill. It was weeks of somebody else's attention. Until it is in place the calls happen, the business is pleased that the phone is ringing, and none of it reaches the campaign that caused it — so the platform is scored on a share of the outcome nobody has counted, and told to go and find more of it.

Whether the lead was any good comes back a month late

The second missing number is the one that decides whether any of this was worth doing.

Enquiries are not equal. Some are the work you want. Some are people who wanted something you do not sell. Some are spam, and some are real but wrong in a way only whoever answered the phone could tell you. That information exists, immediately, in the head of the person taking the calls.

It has reached me in month two more often than in week one.

By then the account has had four more weeks of budget pointed at whatever the fast signal said was working. If half of those enquiries were the wrong kind, the campaign has spent a month getting better at producing the wrong kind, and it has done that competently. The optimization was never broken. It was aimed using the only number that arrived in time.

None of that is inevitable. Google publishes a mechanism for sending lead quality back into bidding — offline conversion imports (opens in a new window), which exist to upload what happened to a lead after it left the website. It requires whoever answers the phone to write down, every week, which enquiries were worth having. No vendor can do that step, which is why it is the one that does not happen.

Nobody in this chain is being difficult

The story that circulates online has a villain in it. Vendors report clicks because clicks flatter them; agencies hide behind vanity metrics. Some of it is true, some of the time.

It has not been the ordinary case for me.

The ordinary case is that "fire the conversion event on the submit callback rather than the click handler" is a sentence with no meaning outside a narrow trade, and it lands in somebody's inbox next to forty other things. So there is a teaching stage before there is a change. I explain what a conversion event is, why a click is not the same as a send, and what the thank-you page is for. Then it gets built, usually quickly.

The lead-quality gap is the same shape and nobody has even named it as a task. Nobody sells a website with a note explaining that the person answering the phone is holding the most valuable marketing data in the business, and that it goes stale.

A campaign optimizes at the speed of its slowest useful signal

The fastest signal is the button press, and it is the least informative. Next is the confirmed form submission, a week away and useful. The call is measurable too, on a longer timetable and somebody else's goodwill. Lead quality is the one that decides whether the money worked, and it arrives last, filtered through somebody's memory of a month they have stopped thinking about.

The two the title means are the last two, and they are the two an ad platform never waits for.

What I will not do about it

I will not report a conversion count as though it measures customers when it measures button presses. If a proxy is all that can be had, the report says which proxy and what it inflates.

Until the calls are tracked, I will say before the first invoice that the account cannot see the phone, and which direction that pushes the cost per enquiry — not which number it would have been.

And I will not run a second month on numbers nobody has checked against reality. If nobody can tell me within about a week whether last week's enquiries were any good, that gets fixed before the second month is funded.

Two things worth checking this week. If you have access to the ad account, open the conversion action and look at what actually fires it; if the word is "click", you now know something about your cost per enquiry that your reporting does not say. The second one needs nobody: ask whoever answers your phone how many calls came in last week and how many were worth having, and notice how long it takes them to find out.

What to do about it depends on what the account is currently attached to, which is worth settling before you spend another month on numbers you cannot judge.