# We measured the wrong Google surface. Here's how we caught it. > A search-card scan missed booking actions on full Google place pages. We pulled the finding, recounted the destination, and changed the measurement rule. Source: https://blog.xonark.com/blog/2026-08-12-recount-correction/ August 12, 2026 · Jack Jia · 5 min read - dental - data - findability - build-log - measurement Validated source update — August 9, 2026. The June measurements below remain visible as a correction log; neither is presented as the current market result. The current [Canadian dental findability scorecard](https://xonark.com/research/canadian-dental-findability/) reports 14,142 of 18,910 checked Google profiles (74.8%) with no booking link in a defined 19,065-listing cohort . It was generated from a coherent read-only snapshot with the subject-anchored parser committed at 97d5b2c8c8d8358422c52b005c36df3373d4e525 . On June 13, we had a sharp headline ready: 98.8% of Canadian dental practices had no one-tap booking link on Google. The first draft had a denominator, a date, and a large national sample—20,258 of 20,506 listings. It was also wrong. We tightened the sample to 19,374 patient-facing practices before scheduling it. The headline survived: 19,138 of 19,374 still appeared to have no booking link. On June 17, the scheduled post briefly entered the blog’s listed and featured source state. Then we pulled it for a recount. The repository shows that public-source interval; it does not tell us the exact moment every edge cache began or stopped serving the page, so we will not turn that into a made-up “hours live” claim. On June 18, the corrected release carried a different result: 75.0% , or 14,534 of 19,374 Google-listed Canadian dental practices. Those figures are a dated June 2026 measurement history. The validated current source is linked above and uses a separately defined cohort, validation contract, and release receipt. ## The search card was not the place page Our first pass read the compact card Google shows in search results. That card exposes only a small set of actions. A booking link usually appears there when Google treats it as the listing’s primary action. Many practices had a Book button on the full Google place page even when the search card did not show it. By treating “not visible on the card” as “not present on the profile,” our scan turned a display limitation into a market statistic. The data was real. The interpretation was not. That distinction matters because a large sample does not rescue a bad measurement. Twenty thousand consistently misread listings produce a very precise wrong answer. ## The check that broke the headline The result looked too extreme. Almost every province was near 100%, yet manual visits to full profiles kept finding booking actions that the card data said were missing. So we changed the test: - Open the full place page for each dental listing. - Record whether a Book action exists there. - Follow the destination far enough to classify it as live self-scheduling, an appointment request, a weak homepage, a dead link, or a blocked check. - Tighten the cohort to exact patient-care dental business types and exclude narrow institutional or service categories. The June recount found 14,534 practices without a Google-profile booking link out of 19,374: 75.0% . That was still a large gap. It was simply not 98.8%. ## What changed—and what did not The June number changed by 23.8 percentage points. The public conclusion became narrower too. We could say that, in that dated recount, three in four Google-listed Canadian dental practices had no booking link on the Google profile. We could not say three in four had no online booking anywhere on their own website. Those are different questions. We also could not treat every link as a completed booking path. Presence and completion are separate measurements. The validated scorecard now carries every current count with its own denominator, coverage, observation window, method, and source. This article remains the record of why those boundaries matter. ## The measurement rule we kept Measure the destination where the person has to finish the job, not the preview surface that points toward it. For a booking claim, that means opening the profile and checking the booking destination. For a contact-path claim, it means loading the practice website and finding the actual form. For an AI-visibility claim, it means checking the answer and its cited destination—not merely whether a brand name appeared somewhere in a result. Every public number should also travel with its denominator, exact question, observation window, method, source, and known blind spots. If one changes, the headline may need to change with it. ## Why keep the earlier result visible? The convenient version of this story would leave only the corrected number and forget the first one existed. That would hide the useful part. Measurement systems fail at the boundary between what a field appears to mean and what it actually captures. The defence is not confidence; it is a recount that can disagree with you. The first version got farther than it should have. The correction gate still worked, but late: the implausible result was challenged, the destination was opened, and the method was allowed to overturn a headline that had entered public source state. For the current aggregate and its full method, read the [Canadian dental findability scorecard](https://xonark.com/research/canadian-dental-findability/). To check one practice rather than a national cohort, [run the free Dental Practice Health Snapshot](https://xonark.com/dental-practice-health-snapshot?utm_source=blog&utm_medium=post&utm_campaign=recount-correction). Validated release receipt Source snapshot SHA-256: d20c1fe2cd6c6245c3077eba6d9be224715c1711908fad5850b61b81fe97e0aa . Parser SHA-256: 1ce38df0c5b129cbb20591c1d5f846f8a4b6e3592d36ca7149bef70486a38123 . Aggregate generator SHA-256: 3e614ef6c4d6c0fe11ac5e5728dd961263ab6d161a2e841cae12e13839a30683 . [← All posts](/)