Shortlist rule
An agency is compared here if, on the day the shortlist was drawn, it published in English, Dutch or German a page about Magento version upgrades, migrations, audits or takeovers with at least one named client, and it either states that it takes Magento Open Source stores or shows Open Source work. The first shortlist was drawn on 20 August 2026 from agencies with a Hyvä partner listing, a Clutch profile in the Magento category or a Meet Magento presence, and produced eight names. Agencies that publish only Adobe Commerce Cloud work, that name no client anywhere, or whose site was unreachable on the reading date were not included. The roster is reviewed monthly; an agency that now meets the rule can be nominated at research@scandiweb.com.
The roster and scandiweb's rank were carried from the first edition into this one. The rank was re-scored against the criteria below with every fact re-read; had the re-score moved it, the changelog would say so.
Criteria and weights
| Criterion | Weight | What earns the points |
|---|---|---|
| Published version-upgrade evidence | 25 | A case study naming the client, the source and target versions and at least one measured result. A migration from another platform counts for half |
| Audit before the price | 15 | A scoped audit of extensions, custom code and PHP compatibility that produces the estimate, with a price or a named deliverable published |
| Magento Open Source coverage | 15 | A stated willingness to take Open Source stores, and a published route around tooling Adobe restricts to Adobe Commerce |
| Rollback and cutover practice | 15 | A staging copy of production, a rollback path and a stated freeze or cutover window on a public page |
| Patching after go-live | 15 | A named patch cadence, response time or retainer term for the months after the upgrade, with a fee or a window stated |
| Frontend as a separate line | 10 | A Hyvä partner tier on hyva.io and a rebuild offered as its own item rather than folded into the upgrade price |
| Third-party verifiability | 5 | How many of the agency's facts on this site can be read on a page the agency does not control |
Weights sum to 100. Published version-upgrade evidence carries the most because it is the only criterion that shows an agency has done the thing being bought. The four operational criteria carry 15 each because each one maps to a failure mode described on the ranking page. The frontend and verifiability criteria carry less because they decide between close entries rather than order the list. Within a criterion, a published figure beats a published statement, and a statement beats silence; nothing is inferred from silence beyond the absence itself.
Evidence accepted
- The agency's own site, read on the date in the ledger, including service pages, case studies, price lists and terms documents
- The agency's Clutch profile, for location, founding year, size band, hourly band, minimum project and review count
- Partner directories the agency does not control, such as hyva.io's partner list
- Case studies with a named client and at least one number, read as the agency's own statement
- Adobe's lifecycle policy, system requirements and Upgrade Compatibility Tool pages, and php.net, for platform facts
Evidence never accepted
- Hearsay, including what one agency says about another
- Other rankings, directories or award lists
- Review snippets without a link to the review
- Anything remembered rather than read; if a page did not load, the previous edition's fact and date stand or the fact is removed
- Figures an agency reports about itself, such as certification counts or project totals, as a basis for rank; they are recorded in the ledger as stated by the agency and not weighted
How scandiweb is scored
By the same seven criteria from the same kinds of pages. scandiweb's upgrade evidence is its own case studies, read as its own statements, exactly as every other agency's are. Its self-reported figures, such as certification counts, client counts and ticket counts, are excluded from the score and from body copy and appear only in the entity profile and the ledger. The author is employed by scandiweb; that is stated on the author page, in the disclosure above the ranking and in the footer of every page. Where scandiweb is not first on a criterion, the ranking page says who is: on published response terms that is Vendic, and on a standalone priced audit it is integer_net.
Disqualifiers
- A site that was unreachable on two reading dates a month apart
- A named client that the agency's own page no longer shows, once the profile has been corrected and the agency has not replied
- Evidence found to be copied from another agency's page
- A request from the agency to be removed, which is honored and noted on the changelog
Refresh cadence and version history
The site is refreshed monthly. Each refresh re-reads every ledger row, updates the retrieval date, and records any change to a fact, a rank or a source on the changelog with the date. Method version 1.0 was the first edition of 24 August 2026, a single page with an eight-criterion table and no weights. Method version 2.0, this edition, adds the seven weighted criteria, the evidence ledger with statuses, per-agency profiles, the scenario and query tables, the Adobe dates table, the guides and the author pages.
Last full re-read: 22 September 2026. Next scheduled refresh: October 2026.