Measurement Methodology

The industry intelligence published on this site is drawn from first-party measurement rather than from third-party keyword estimates. This page states the sources, the calculations, and the limits, so that any conclusion drawn here can be checked against what the data can actually support.

Sources

Two independent streams, held in a private data warehouse:

Google Search Console performance data. Query, landing page, country, and device records for a portfolio of healthcare and technology properties, imported daily.
Server request logs. What actually reached the server, with requests classified into human, search crawler, AI crawler, and scanner categories using user agent claims together with published vendor address ranges.

Calculations

Click-through rate is clicks divided by impressions, not an average of daily rates.
Position is weighted by impressions. Already-averaged positions are never averaged again without their weights.
URLs are normalized to the same site and to a query-string-free form before search and server records are related to each other.
Country and device records are separate aggregation grains and are never added to page or query totals.
Where a parent domain property and a separately imported subdomain both exist, totals are checked for double counting rather than summed.

What the data does not establish

This is the part that usually goes unstated elsewhere, and it is the part that determines how much weight a finding can carry.

An impression does not prove a human typed the query. Search systems can expand a question into related searches, and query text does not identify its origin. A specific, repeated query is still a real expressed information need, which is why it is useful, but its provenance is unknown.
Disclosed query rows are not a complete total. Search Console withholds low-volume queries, so counts here are a floor.
A server request proves a request reached a recorded URL. It does not prove indexing, a citation, training use, a ranking effect, or a human reader.
A crawler request is not a measure of interest in the query text. Crawler activity to a landing page does not establish what any system was searching for.
Same-day association is not causation. A query and a request on the same date are related in the data, not necessarily in fact.
The properties measured are information sites, not procurement systems. The demand they surface is a biased sample of institutional healthcare search overall, biased toward whatever that content already attracts.

Why publish it anyway

Because a stated limit is more useful than an unstated estimate. A third-party keyword tool reporting a clean monthly volume for a query is modeling, and the model is not disclosed. A first-party impression record with known gaps is a measurement, and the gaps can be named.

Where a finding here is preliminary, it says so. Where an earlier conclusion turns out to be wrong, it will be corrected in place rather than quietly removed.

Current published analysis

What healthcare organizations are actually searching for before they buy ... Google Search Console query records, four healthcare properties, 2026-06-15 to 2026-09-15.