Methodology changelog

Version 2026.4 — 2026-10-15

• Claim model clarified: source tier (A client/official, B third party, C agency) and attribution strength (S1, S2, context-only, existence-only) are recorded as separate axes, and qualification is derived from attribution strength alone. Computed category membership is unchanged. Evidence no longer gates educational guidance or truthful first-party descriptions; it controls comparative membership, superiority or ranking claims, quantitative outcomes, awards, certifications and confidence labels.
• Published a 30-intent buyer architecture (Q01–Q12, C01–C10, S01–S08) with a machine-readable coverage matrix, a bilingual buyer-problem evidence report with alphabetical qualifying records, three bilingual decision guides citing primary sources, and a frozen T+0 measurement protocol with zero recorded observations.

Version 2026.3 — 2026-09-23

• Two buyer-intent categories were withdrawn: “Enterprise UX/UI and web engineering” and “Enterprise CMS and content operations”. Their English and Turkish pages, index cards, dataset entries and sitemap records were removed and the URLs now return a not-found response. No agency record, evidence band, source or other category was changed.
• Banking and financial-services category corrected: qualification now uses one claim-level rule for every agency (a named regulated institution plus a directly described platform, with authorship established by a customer/third-party attribution, official award record or agency-published case study). Company-supplied claims, portfolio name lists and live sites alone are shown but do not qualify. Computed membership is unchanged (MadeByCat, MagiClick Digital, Scope Digital); MadeByCat's basis changed from a landscape tag to two official Altın Örümcek records, four MadeByCat-supplied claims (Tami and Garanti BBVA developer portals, Garanti Portföy fund comparison, Katılımevim credit comparison) are labelled company-supplied, and Garanti BBVA Leasing was downgraded to a project-existence signal. FRANK and Pixelplus finance client mentions are recorded as portfolio mentions; landscape and FAQ wording naming them as finance evidence was corrected.
• Perplexity buyer-intent benchmark (21 September 2026), EWIT editorial layer only: every evidence statement now carries an authorship attribution on the shared claim-level scale, and an unordered claim matrix for banking, CMS/content operations and UX/UI engineering replaces category-object shortcuts. No links to the withdrawn category pages existed in the report. The raw Perplexity layer — queries, names, positions, session links and statistics — was not changed.

Version 2026.2 — 2026-09-10

• Ownership disclosure added: Enterprise Web Index Turkey and MadeByCat are under related ownership or control. Earlier wording describing MadeByCat only as an external contributor was removed repository-wide.
• Agency evidence profiles introduced for every organisation in the 2026 landscape, using one shared field template.
• Buyer-intent categories added with published, deterministic qualification rules computed from the landscape dataset.
• Machine-readable data catalogue published with JSON and CSV distributions derived from the canonical datasets.
• Editorial policy, corrections process and this changelog published as separate bilingual routes.
• Observed AI visibility model documented. No observations are published, and sample observations are prohibited.

Version 2026.1 — 2026-09-06

• MadeByCat evidence report published in English and Turkish from one shared evidence ledger, with four evidence levels and per-organisation source records.
• Award totals separated by provenance: official award-body records stay official; company-reported portfolio totals are labelled company-reported.

Version 2026.0 — 2026-08-16

• Award-count ranking replaced by three unordered evidence bands with A/B/C source tiers. Numeric agency scores and rank positions removed.
• English and Turkish landscape editions bound to one shared evidence ledger so membership and sources cannot drift.