What changed on 25 July 2026
The directory was rebuilt from scratch. The previous build queried Wikidata for instances/subclasses of what its own source notes described as "advertising agency" (wd:Q1341478) — but wd:Q1341478 is actually Wikidata's item for “energy company”, not advertising agency (re-verified live against Wikidata on 2026-07-25). That mismatched class query is why the previous 456-entry directory was almost entirely investment funds, SICAVs, pension schemes, sports conferences/leagues, and a university communications office — not marketing agencies. This build re-queries Wikidata against the correct, spot-verified classes below.
Sourcing
The primary and only source is the Wikidata SPARQL endpoint at https://query.wikidata.org/sparql. We query for entities whose own wdt:P31 (instance of) resolves, via wdt:P279* (subclass of, transitive), to one of six verified genuine-agency classes:
wd:Q216931— advertising agency (includes the media agency and full-service-agency subclasses)wd:Q12043474— public relations agencywd:Q9592701— marketing agency (includes the digital agency subclass)wd:Q967140— media agencywd:Q1369723— digital agencywd:Q131427387— marketing solutions company
Query pattern used (run once per class, results merged and de-duplicated by item):
SELECT ?item ?itemLabel ?itemDescription ?countryLabel ?hqLabel ?founded ?website WHERE {
# run once per class, results merged + deduped client-side:
# Q216931 advertising agency Q12043474 public relations agency
# Q9592701 marketing agency Q967140 media agency
# Q1369723 digital agency Q131427387 marketing solutions company
?item wdt:P31 ?d . ?d wdt:P279* wd:Q216931 . # (repeated per QID above)
?item rdfs:label ?itemLabel . FILTER(LANG(?itemLabel)="en")
OPTIONAL { ?item schema:description ?itemDescription . FILTER(LANG(?itemDescription)="en") }
OPTIONAL { ?item wdt:P17 ?c . ?c rdfs:label ?countryLabel . FILTER(LANG(?countryLabel)="en") }
OPTIONAL { ?item wdt:P159 ?hq . ?hq rdfs:label ?hqLabel . FILTER(LANG(?hqLabel)="en") }
OPTIONAL { ?item wdt:P571 ?founded . }
OPTIONAL { ?item wdt:P856 ?website . }
}
For each match we pull the English label, English description (where present), country (P17), headquarters city (P159), founding date (P571), and official website (P856) where present.
Spot-verification
Before this data replaced the previous directory, a random sample of results was checked against their Wikidata item pages and, where a P856 official-website value existed, against the live site (title tag / HTTP status). The sample confirmed real, operating advertising, PR, marketing, media, and digital agencies with no funds, sports bodies, or generic institutions. One item did surface a genuine edge case — a university's internal communications office, tagged on Wikidata as a public-relations-agency instance — and was excluded by hand: it serves one parent institution rather than external clients, which is not what this directory is for.
What is excluded
Entities without any English label on Wikidata are excluded, so every listing has a real, human-readable name. Entities whose only P31 class match is one of the six above but which on inspection describe an internal institutional office rather than a client-serving business are excluded by hand (see above); this directory does not attempt to auto-detect that pattern at scale, since it was a single occurrence in the current dataset.
Region assignment
Region is assigned from the P17 country value into one of six super-region buckets (Asia-Pacific, Europe, Global, Latin America, Middle East & Africa, North America) used across the site. Entities with no stated country sit in Global — that is a statement of missing data, not a claim that the firm is literally global. A handful of transcontinental/ambiguous countries (e.g. Turkey, Azerbaijan) are bucketed by editorial judgment call, disclosed here rather than hidden: Turkey and Azerbaijan are grouped into Middle East & Africa and Europe respectively, matching common marketing-industry regional reporting conventions rather than a Wikidata property.
Classification (formerly "specialization")
Each agency's classification tag(s) on the site are exactly the Wikidata class(es) it matched during the query above — not an invented specialization. An agency classified on Wikidata as both an advertising agency and a digital agency, for instance, carries both tags.
Cross-links on service pages
Three service pages (Public Relations & Earned Media, Public Affairs & ESG, Marketing Automation & MarTech) show a "related agencies" block. Agencies shown there are drawn from the classification(s) most plausibly aligned with that discipline (e.g. Public Relations Agency → Public Relations & Earned Media) — a disclosed editorial mapping, not a claim of verified capability depth in that specific discipline. Industry vertical pages (/verticals/*) do not carry an agency cross-link block: Wikidata does not record an industry-vertical field for these entities, and inventing one would not be honest.
Refresh cadence
The directory is regenerated by re-running the query above. There is no automated schedule; the site is regenerated manually when a correction or refresh is warranted. Every regeneration updates the build metadata below and the footer timestamp sitewide.
Limitations to keep in mind
- Wikidata coverage skews toward agencies with an English-language Wikipedia/Wikidata presence; many strong independents, especially non-English-market shops, are missing.
- Classification tags are the agency's Wikidata instance-of class, not a current capability audit.
- Region assignment is based on stated country of headquarters and does not reflect global delivery footprints.
- Founding year and official website are only shown where Wikidata records them; blank means not recorded, not zero.
Read this methodology as a working description of the current build, not as a permanent specification.
Corrections
Spot an error in an agency entry, or know of a genuine agency this restricted query should have caught but didn't? Use the contact page — corrections are the fastest way this directory gets better.