Methodology
Indexes
The measures this directory computes on top of the register, and exactly how each one is made. There is one so far, the popularity index. None of them rates an insurer, and none changes how a company is listed, classified or verified.
Popularity index
Insurer pages with enough search data carry a popularity index: an estimate of how often people search Google for that insurer over a rolling twelve months. It is built from the impressions Google Search Console reports to us for the insurer’s page in this directory — each one a search in which someone looked for the company and our page appeared — and it treats them the way a pollster treats a sample.
The current index covers 2,793 insurers, from 4 September 2026 to 26 September 2026, and is recalculated every day.
What it is built from
Google Search Console reports, for each insurer’s page in this directory, how often it appeared in Google results, the searches that produced those appearances, and where the page ranked. Each appearance is a search in which someone looked for the company and our page was shown. Clicks are not used: they measure how appealing our page is, not how sought-after the company is.
The window
The index covers a rolling twelve months, so seasonal swings even out. Until the site has a full year of data, it covers every day since the first, and insurer pages say “since” that date rather than “12 months to” a date.
Three corrections, before any ranking
- Automated searches are removed. Agents and scrapers send structured, repetitive queries that people do not type, and searches of that shape are subtracted day by day. The filter is deliberately cautious: it removes too little rather than too much. In the current window, 15% of searches were set aside as automated.
- Each day is weighted for where the page ranked. A page in the top 10 results is seen by nearly everyone who searches; one further down is seen by only a fraction of them. Each day’s searches are therefore scaled up by how unlikely it was that a searcher saw the page, by at most 10 times, so that a single day far down the results cannot dominate.
- Small samples are drawn towards their market. An insurer with only a few searches can look popular or obscure by chance, so its estimate is pulled towards what is typical for insurers in its market, or in its region where a market has fewer than 8 insurers. The more search data an insurer has, the less it is pulled.
Reading the number
The index is a percentile from 0 to 100 among the 6,768 insurers people searched for in the window: 84 means more search interest than 84% of them. It is measured against searched-for insurers rather than the whole register because about half the register draws no searches at all, which would crowd every published value into the top of the scale.
Likely range and confidence
Every value carries a likely range: the span within which the true value most probably lies, a 90% interval. The confidence says how much search data stands behind the estimate:
- Low: fewer than 50 searches in the window.
- Medium: from 50 to 400 searches.
- High: more than 400 searches.
In the insurer tables, one, two or three bars beside the index show the same confidence.
Rank in the market
The rank places an insurer among every insurer licensed in its market, including those with no index, so it reads against the whole register and not only against the insurers people search for.
When an insurer has no index
An insurer needs at least 10 searches in the window before its index is published. Below that, its page shows no index and the tables show a dash, which sorts last. And if fewer than 50% of the window’s searches carry ranking data, nothing is published at all, rather than an uncorrected figure.
Limits
- It measures people looking a company up by name in Google, mostly retail customers. It cannot see searches in which our page never appears, and an insurer with a strong website of its own may push our page off the first results for its own name.
- How visibility falls with rank is a modelled assumption, not a measurement. It is capped, and it can be retuned.
- The automated-search filter is a floor: some automated demand survives inside searches for a plain company name.
Open data
The values the pages show (index, likely range, rank in the market and confidence) are published daily under CC BY 4.0, without any search counts: /api/v1/popularity.json
What it is not
It is not a rating of the insurer. It says nothing about the company’s financial strength, service, claims record or standing with its supervisor, and it plays no part in whether or how a company is listed here. Licensing, classification and verification come from each market’s register alone, and the popularity index never touches them.