Four location pages that finally told the truth about who works where

Multi-location sites drift. Each location page gets written by whoever is closest to it, at a different time, and nobody sees all four side by side until something goes wrong.

Here, three things had gone wrong at once. One location was displaying four physicians who had never practised there — leftover demo content from an earlier draft, presented to patients as fact. Another had lost its map section entirely, so a patient could not find the building. A third was offering three services where the other locations offered four, implying a treatment was unavailable there when it was not. And several physician profiles existed twice, with the older, thinner version still reachable.

These needed separating. Making the pages look alike is a templating problem. Making them say true things is a fact-checking problem, and solving the first while ignoring the second produces four tidy pages that are still wrong.

The most serious defect was not on a location page at all.

The homepage carried a physician filter that mapped doctors to locations. The mapping was substantially wrong, which meant a patient could search for a specialist, be shown a clinic, and travel to a building where that doctor does not practise. For a cancer practice that is not a broken link — it is a wasted journey at the worst possible moment in somebody's life.

Correcting it required the roster to exist as a single verified source rather than as a copy on each page. That is now how the location pages and the filter both get their data.

All four location pages now carry the same sections in the same order, and each one is accurate about the doctors and services at that specific site. The orphaned duplicates return 404 rather than quietly existing. The sitewide link defects are gone.

No ranking or traffic figure is claimed here. This work was a correctness and structure pass, and the honest measure of it is that the pages no longer contradict each other or the practice's own records.

These pages were marked as builder-edited, and the builder's stored data was a complete, separate copy of the page. But the live site was rendering from the standard post content instead — confirmed by the fact that editing the builder data alone changed nothing visible.

Two copies of every page existing in the database, with only one of them live, is a trap. Whoever edits next will open the builder, make a change, see no effect, and conclude the site is cached. It is worth fixing properly before that happens; the alternative is a site whose content nobody can confidently edit.

The site is also still served from a temporary platform domain rather than its public one. Pointing the real domain at it is outside this work.