An index that ranks named companies is only worth publishing if someone who disagrees with their placement can check the work. This page is written for that reader.
In one line: the universe is the MDM Top Distributors lists, deduplicated; revenue is each company’s most recent reported figure, sourced per company; archetypes are editorial; the Digital Readiness Index is measured — four pillars, fourteen signals, 100 points, scored from live pages.
The index covers 224 distributors — every company appearing on the MDM Top Distributors sector lists, deduplicated across the twenty verticals so a company charting in ten of them appears once. We use MDM’s lists to define the universe because they are the industry’s reference for who counts as a major North American distributor. Everything we do with that universe — the revenue refresh, the archetype classification, and the Digital Readiness Index — is our own.
Each company’s figure is the most recent annual revenue it has publicly reported, researched individually rather than carried over. Sources are preferred in this order: SEC filings, company announcements and annual reports, MDM, then established trade press. The fiscal year and the source URL are recorded for every figure, and the table shows what the figure actually covers — total company, North America only, or a distribution segment — because many of these firms are global and a global number presented as a North American one would be misleading.
Companies that disclose nothing are not assigned a number. Roughly a third of this industry is private and says nothing about revenue. Estimating them from headcount or branch count would produce a more complete-looking table and a less true one, so they are listed below the ranking instead, in their own tier. Where the only figure available is a third-party estimate, it is marked as an estimate in the table.
Not every figure is a 2025 figure, and the dated ones say so. A handful of private companies last published a number several years ago and have said nothing since. Rather than drop them or quietly imply the figure is current, they are ranked on what they have actually disclosed and labelled dated in the table with the fiscal year attached. Reading a ranking without reading the fiscal year next to it will mislead you, which is why the year is never hidden behind a tooltip.
Consolidation is tracked as of the build date. Where two companies in the source lists have since merged, the surviving entity carries its own reported figure and the acquired entity is listed but excluded from the ranking — marked now part of — so the combined business is not counted twice. Deals that are announced but not yet closed leave both companies independent, because on the day we publish they still are.
The six archetypes describe where a distributor’s advantage comes from, not what it sells. Two companies in the same vertical at the same scale routinely run opposite playbooks — a large electrical distributor may compete on national purchasing leverage or on the density of its local branches, and those are different businesses wearing similar clothes.
Assignment is editorial. It is made from published operating facts — branch counts, vending device counts, service and fabrication revenue, acquisition cadence, whether the catalog is self-serve — and each company’s classification carries a one-line rationale citing the specific fact behind it. A company can carry a secondary archetype where two genuinely apply; the primary is the one that would hurt most to lose.
This is the subjective half of the index and we would rather say so than dress it up. If you run one of these companies and think we read your model wrong, tell us — that argument is more interesting than the classification.
The Digital Readiness Framework is Anglera’s own. We built it, we score it, and we publish it — which means the honest thing to say up front is that no third party has reviewed or endorsed this index. It is not an industry standard and we are not presenting it as one.
What it is built on is a set of published specifications that already define what machine-readable commerce data looks like. Every signal in the framework tests against one of them rather than against our opinion:
The rest comes from operating experience. Anglera’s business is completing and correcting product catalogs for distributors and manufacturers, and the framework’s weighting reflects what we repeatedly see decide whether a catalog performs: attribute depth above almost everything, identifiers close behind, and consistency mattering more than any single page’s best-case quality. The proprietary part is the scoring and audit system itself — the extraction pipeline, the rubric, and the deterministic scorer that turns a live URL into a reproducible number.
We would genuinely like this to be argued with. If you work in B2B commerce and think a weighting is wrong — that imagery is overvalued, that gated catalogs deserve different treatment, that a pillar is missing — that argument is more useful to us than agreement, and the next edition is where it would land.
The measured half. For every distributor ranked in the top five of any vertical, we sample live product pages from their own public catalog, extract each one with the same extraction service Anglera runs in production, and separately probe the site for structured data, sitemap, and crawler access. There is no model judgement anywhere in the scoring: the same site scores the same number twice in a row.
The framework is deliberately wider than product data alone. A distributor can have excellent attributes on a page that no crawler is allowed to fetch, or immaculate structured data on a catalog that requires a login — and in both cases the practical result for a buyer using an AI assistant is the same as having no data at all. So the index scores the whole path: what is on the page, whether it answers a real question, whether the commercial terms are visible, and whether a machine can reach any of it.
Five product pages, from five different top-level categories in the site’s own navigation, spread across the breadth of what the company sells. Within each category listing, the page is taken from the middle of the results — roughly the fifth to fifteenth item — never the first item and never anything flagged as featured, best-selling, recommended, or promoted.
That rule exists for one reason: featured products are the pages a company has groomed. The middle of a category listing is what the catalog actually looks like, which is also what a buyer or an agent hits when they search for something specific.
A few of the largest distributors serve their product pages to human browsers but refuse them to every standard client, including this measurement’s production extractor. Leaving the biggest names in a vertical blank would misrepresent them, so for those companies (4 in this edition) the page-level signals are read in a full browser session — from the same five sampled pages, scored by the same rubric. Two things stay honest about it: the scorecard discloses the browser measurement, and the Machine & Agent Readiness pillar is still scored from what a standard client can retrieve, because for a crawler or a shopping agent the block is the reality.
Can a machine tell what this product is, and match it to the same product elsewhere?
Scored per product page, averaged across the sample
Does the page answer what a buyer actually asks before they commit?
Content scored per page; consistency scored across the sample
Can a buyer find out what it costs and whether it ships, without asking a human?
Price and stock per page; gating assessed once per site
Can a crawler, a marketplace, or an AI shopping agent actually consume any of it?
Probed once per site
The page-level signals are averaged across the sample; consistency, gating, and all four agent- readiness signals are properties of the site and are scored once. Alongside the index we publish two numbers the average hides: median attributes, the typical structured attribute count on one of the company’s pages, and consistency spread, the gap between the richest and thinnest page in the sample. The spread is usually the more uncomfortable number, because it measures how differently a company treats its own products.
Four different situations, reported separately, because conflating them would misrepresent most of them.
None of the three is scored as a zero. A zero would say the pages are bad; these say we could not read them, which is a different claim.
The index records what each storefront runs on — commerce engine, site-search vendor, CMS, and the CDN or bot wall in front of it. A platform named next to a company is a claim about that company’s technology stack, so it is held to a two-sided rule, and the two sides are gathered independently.
The first side is the storefront itself. A script fetches the homepage, one of the sampled product pages, and a site-search results page — found from the site’s own schema.org SearchAction or header search form rather than guessed — and matches the raw HTML, the asset URLs the page loads, the response headers and the Set-Cookie names against a signature dictionary. Every hit saves the string that matched and the URL it matched on. Where a site-search page loads its client from a bundle, the first-party bundles are read too, because that is where a search vendor’s endpoint usually lives. Signatures are weighted: one unambiguous match, or two weaker matches on two different surfaces, is the floor for calling a field detected.
Twenty-one of these storefronts never showed a standard HTTP client a page at all. For twelve of them the same checks were run in a supervised Chrome session, reading the resources the page actually loaded — which is how Grainger’s SAP Commerce back end and Ferguson’s Demandware storefront became visible at all, neither being mentioned in anything a crawler is served.
The second side is the public record. One research agent per vendor family searched vendor customer lists, case studies, press releases, systems-integrator write-ups and job postings for any of the 203 companies, returning a public URL and a verbatim quote for each claim; a second agent then tried to refute every claim by opening the URL and checking the quote was really on it, and six of forty-one claims died there — dead links and mismatched entities.
Only where the two sides agree is a platform published as confirmed. Where only the storefront shows it, the value publishes as detected and says so on its face. Where only a vendor says it and the storefront could not be read, the claim stays in the dataset as reported and is never stated as fact on a company row. CMS and CDN are exempt from the two-sided rule because they are directly observable in the response and nobody publishes a customer list of who sits behind Akamai.
Two things this cannot see. A signature dictionary only knows the platforms it has signatures for, so a storefront that matches nothing is recorded as “no known vendor” rather than as an absence of technology — for the largest companies in this index that usually means a build of their own. And platforms migrate, so every value carries the date it was observed and is re-checked on the same quarterly cadence as the scores.
A five-page sample characterises a catalog; it does not audit one. A distributor with excellent data in its core categories and thin data in a long tail will score somewhere in between, which is the intent, but the specific five pages matter and a re-run will move a score by a few points.
The index measures the public shelf. Several of these companies pass far richer data to customers through EDI, punchout, or a syndication feed than they publish on the open web. That data is real and it does not show up here — though it is worth asking why the public catalog is the thin version, given that is what search engines and shopping agents actually read.
Finally, this is a vendor’s index. Anglera sells product-data enrichment, which means we have an obvious interest in the finding that product data is thin. That is exactly why the scoring is deterministic, the rubric is published in full, and the sampling rule forbids us from choosing which pages to look at.
Send a correction with a public source and it gets updated, with the revision recorded in the public corrections log. Revenue, fiscal year, ownership, headquarters, and archetype are all correctable by argument.
Digital Readiness Index scores are not adjusted by hand. If your catalog or your robots.txt has improved, the score is recomputed by re-running the measurement against your live site — which is the only version of the number that means anything.
Index built 2026-08-06; digital shelf measurement run 2026-08-06.