The things we measure and refuse to score
Every business in the Local Digital Visibility Index is measured on six scored pillars. It is also measured on about twenty things that are deliberately not scored at all: a crawl of its site, real-user speed data, its backlink profile, its Companies House record, the detail of its Google profile.
Those go on the scorecard as context and never touch the number. This article is about what happened when we finally tested whether they should.
Every figure is computed from the published datasets when this page is built, across 1371 businesses in 34 indices.
The test
Section titled “The test”For each signal, one question: does it move with whether a business can actually be found? Findability is already measured as the Visibility pillar — whether a firm appears anywhere in the first hundred results for its own sector’s keywords.
A signal that matters should rise and fall with that. The scored pillars are included on the same axis, so the comparison is like for like.
| Signal | Scored? | Businesses | Moves with findability |
|---|---|---|---|
| AI search presence a scored pillar | 15% of the score | 1371 | 0.64 |
| Local presence a scored pillar | 20% of the score | 1366 | 0.40 |
| Content & trust a scored pillar | 10% of the score | 1222 | 0.32 |
| Backlink authority DataForSEO's domain rank | no | 1153 | 0.30 |
| Technical a scored pillar | 20% of the score | 1371 | 0.22 |
| Pages on the site how much there is to find, from our own crawl | no | 1371 | 0.21 |
| Lighthouse performance the lab speed score | no | 1055 | -0.20 |
| Speed & CWV a scored pillar | 20% of the score | 1246 | -0.09 |
| Lighthouse accessibility the lab accessibility score | no | 1063 | -0.05 |
| Referring domains how many distinct sites link to it | no | 1153 | 0.03 |
| Broken backlinks share of inbound links that no longer resolve | no | 1153 | 0.03 |
| Google profile photos how many images the profile carries | no | 1089 | 0.02 |
| Company age years since incorporation, Companies House | no | 300 | 0.01 |
Those are correlation coefficients. Zero means no relationship. For observational data like this, above about 0.3 is worth noticing and below 0.1 is noise.
The uncomfortable part
Section titled “The uncomfortable part”Backlink authority and Pages on the site carry no weight at all and sit at the top of that table — above Technical, above Local presence, and far above Speed, which is a fifth of every published score.
Read plainly: two things we collect and refuse to score predict findability better than most of what we do score.
The obvious move is to start scoring them. We are not going to, this quarter, and the reasons are worth setting out because they are the same reasons the index is worth reading at all.
Why they stay unscored
Section titled “Why they stay unscored”A published score must not move because a third party changed their mind. Backlink authority is a vendor’s proprietary metric. If they revise how it is calculated — which vendors do, without notice — every business in this index would move in the rankings without touching their website. A score nobody can reproduce from first principles is a score you have to take on trust, and the whole point of this index is that you do not have to.
Coverage is uneven, and uneven coverage silently punishes. Real-user speed data exists for a quarter of the businesses we measure, because the quiet ones do not generate enough traffic to report. Scoring it would mean the smallest firms are marked on evidence the larger ones supply and they cannot. That is not a measurement, it is a tax on being small.
Correlation here is not causation, and the direction is genuinely unclear. Bigger sites with more pages rank better. Do they rank because they have more pages, or do they have more pages because they are bigger businesses that also invest in everything else? This data cannot separate those, and a scoring weight is a causal claim whether or not you intend it as one.
We have been wrong about a weight before, this quarter. The review velocity component turned out to be measuring which trade a business was in rather than the business. Adding two new weighted signals on one quarter of correlational evidence would be repeating that mistake with more confidence.
What we are doing instead
Section titled “What we are doing instead”The signals stay on the scorecard as context, clearly marked as unscored, where a business can see them and act on them without them silently moving its rank. And in September, when a second quarter lands, the same analysis runs again.
A relationship that holds across two independent quarters is worth ten that appeared once. If pages and backlink authority still lead the table then, that is an argument for reweighting — announced in advance, with old and new scores published side by side for one quarter, exactly as the changelog promises.
What this means if you run one of these businesses
Section titled “What this means if you run one of these businesses”The practical reading is narrower than the table looks.
More pages is not the instruction. Publishing forty thin pages will not move you. The signal here is that businesses with something substantial to find are the ones that get found — which is the same thing the Content pillar says, arriving by a different route.
Backlink authority is not something to buy. It is a summary of who already links to you, and the market in buying that is exactly the thing search engines spend most effort discounting. It is a symptom of being worth linking to, not a lever.
Speed remains the thing most often sold and least often justified. It correlates −0.13 here, it is 20% of the score, and we have published that inconsistency rather than quietly fix it. Separately, the lab number most audits quote is a median 7× worse than what real visitors get.
The data is open
Section titled “The data is open”Every signal above is in the published dataset as unscored context on all 1371 businesses, and the pipeline is open source. If you think the analysis is wrong, it is checkable — and we would rather be corrected than cited.