The pixel is there, the conversions aren’t: the ads run blind

Most sites have a pixel. The agency installed it, or a plugin did. The problem is not that the pixel exists but what it counts. If it only counts “page viewed”, the ad platform learns one thing: who clicks. It never learns who filled the form, who called, who bought. And what it can’t learn, it can’t optimise. The ads run blind — they find the clicker, not the buyer.

Where the leak is. In two places. The first is invisible: with no conversion signal, the platform steers budget toward cheap clicks, and the cheap click is usually the person who doesn’t buy. The second is visible but misread: the figure labelled “conversions” in the report is actually page views, or everyone who reached the thank-you page — including the staff member who tested the form. Reports like that show a wonderfully low cost per conversion, because what’s being counted isn’t a conversion.

Three levels. At level one there is no pixel, or nobody knows whether there is. At level three a pixel exists and one or two events are defined — form submission, say — but phone taps, WhatsApp taps and purchases are missing; the setup was done once and never looked at again. At level five every main conversion is tagged, the tags are verified monthly, and server-side tracking (a setup unaffected by browser blocking) is in place.

A field example. One company’s report showed four hundred “conversions” a month at an enviable cost. The site’s own form log showed sixty-nine. The gap came from the thank-you page being reachable from the menu: every visitor passed through it once, and the pixel counted each pass as a conversion. For three months the platform had been optimising toward “people who click the menu”. When the tag was fixed, conversions divided by six, cost per conversion multiplied by six, and a real number was discussed for the first time.

The five-minute check. Open the “events” or “conversions” screen of the ad platform — Events Manager in Meta, Conversions in Google. Look at what was counted in the last seven days. If the list only shows “PageView”, the ads are blind. If “Lead” or “Purchase” is there, ask the second question: how many forms were submitted on the site last week, and how many “Lead” events did the platform count? If the two differ by more than twenty percent, the setup is wrong. No technical knowledge needed; putting two numbers side by side is enough.

First action. Write on paper what should count as a conversion on your site: form submission, phone-number tap, WhatsApp tap, purchase, booking. Usually four or five items. Hand the list to the agency or the web person with one sentence: “One event for each of these, and a count for each of them a week from now.” The setup is a day’s work; the decision has to be yours, because you — not the agency — know what a conversion is.

Growth 360 measures this point as “retargeting and tracking”. One of the Full Assessment’s twelve evidence questions sits exactly here: “How many conversions did the platform count last month, and how many did you?” The gap between the two is one of the first stones in the Evidence Index.