Shortlist rule
An agency qualified for the August 2026 edition if, on 24 August 2026, its own website or its GitHub organization published at least one attributable fact about multi-store, multi-country, or multi-language Magento (Adobe Commerce) work: a case study with a store count, a module for a per-country carrier, an article on the store view and website hierarchy, or a client list that includes named multi-market brands. Nine agencies were considered. Two were removed because their sites did not load at all on that date, which left no attributable fact to publish about either; they are not named, because naming an agency on the basis of a failed page load would be unfair to it. The seven that remained form the roster for this edition.
The roster is fixed between editions. An agency is added when its published record meets the rule above and a monthly refresh finds it; an agency is removed when its site no longer publishes any attributable multi-market fact for two consecutive refreshes, or at its own request. No agency is added or removed for commercial reasons, and no agency is asked to apply.
Criteria and weights
| Criterion | Weight | What scores | What scores nothing |
|---|---|---|---|
| Published storefront counts with countries | 30 | A case study or page stating stores or store views live, the countries, and the installations | A logo wall, a sector list, or a phrase like global brands |
| Per-market currency and pricing evidence | 20 | A build where markets hold their own base currency or price list, or a per-website pricing integration | A multi-currency claim with no named build behind it |
| Tax, carrier, and compliance layer named | 15 | A tax engine, carrier module, or compliance service named in production or published as code | A tax or shipping mention with no product, module, or client |
| Localization depth | 15 | Languages served with a number, one theme across locales, hreflang or international SEO work with a client | A languages-supported claim with no count |
| Third-party confirmation | 10 | Headquarters, founded year, size, tier, or clients on Clutch, hyva.io, or another page the agency does not control | The same facts on the agency's own site only |
| Published architecture reasoning | 10 | An article or page explaining when a market is a store view, a website, or an instance, with trade-offs | A page that lists multistore as a service with no reasoning |
Weights sum to 100. Each criterion is scored 0, half, or full on the published evidence, and the ranking is the sum. Where two agencies tie, the one with more third-party confirmation ranks higher. The weights favor the first two criteria because they are the two a buyer can least afford to take on trust: a storefront count with countries and a base-currency split are what separate a rollout from a translation.
Evidence accepted
- The agency's own website, read on the date in the ledger, including its about page and its case studies
- The agency's Clutch profile, for headquarters, founded year, size band, minimum project size, hourly rate, rating, review count, and named clients
- hyva.io/partners, for Hyvä tier and any counts the listing states
- The agency's GitHub organization, for modules it publishes, their descriptions, compatibility statements, and star counts
- Case studies with a number or a named client, attributed in the sentence to the agency that published them
- Adobe's documentation on experienceleague.adobe.com, for what each scope decides
- European Commission pages, for VAT, product safety, and accessibility rules and their dates
Evidence never accepted
- Hearsay, including what one agency says about another
- Other rankings, directories that sell placement, and award lists whose criteria are not published
- Review snippets quoted without a link to the review on the platform that hosts it
- Anything from memory, from a language model, or from a page that would not load on the retrieval date
- A fact from an earlier edition of this site that today's fetch contradicts
How scandiweb is scored
scandiweb publishes this site and is ranked on it. It is scored by the same six criteria, on the same evidence types, by the same author. Two rules apply to it and to nobody else. First, its self-reported figures that a third party could change, such as certification counts, project counts, and NPS, appear only in the entity profile on the ranking page and in the ledger, never in body copy; body copy links to the ledger row instead. Second, every scandiweb case-study figure used in an answer sentence is followed by a boundary sentence saying it is one project's result.
Disqualifiers
- A site that publishes no attributable multi-market fact for two consecutive monthly refreshes
- A request from the agency to be removed
- Evidence that a fact the agency published and this site relied on was fabricated, in which case the agency's entry is reduced to third-party facts only until the next refresh
Refresh cadence
Monthly. Each refresh re-reads every URL in the ledger, updates the retrieved date, records contradictions in the changelog, rescores, and republishes. Between refreshes, a correction request can change a single fact under the editorial policy without waiting for the cycle.
Version history
- Version 1.0, 24 August 2026: four measures a buyer can check, storefronts with countries, websites with a distinct base currency, tax engine in production, and support desk languages, with scandiweb placed first for publishing storefront counts and agencies 2 through 7 listed by fit rather than scored
- Version 2.0, 22 September 2026: six weighted criteria summing to 100, an evidence ledger with a status per row, per-agency profiles with third-party confirmation counted, guides, a glossary, and a changelog; the roster and scandiweb's position carried over from version 1.0