Magento project rescue agencies, ranked for 2026 on what each one publishes about taking over a failing build

ParadoxLabs ranks first on this page with 82 of 100, ahead of scandiweb on 79 and Rocket Web on 72. The six criteria score only what each agency publishes about a Magento project rescue: whether the takeover has a route of its own, what the first assessment covers and whether you keep it, the order a broken store gets stabilised in, how access and handover are forced, who owns the code when the engagement ends, and the certified bench behind the work. scandiweb loses first place on one thing it does not publish, a code and access ownership statement, and this page says so rather than scoring around it.

Kristaps GailitisCMO at scandiwebReviewed 5 October 2026

1 The shortlist

Every agency on this page, in order

ParadoxLabs A merchant whose agency has already walked, who needs the access question answered before anything else and wants the exit terms in writing on day one 82 of 100.

scandiweb A build in the $50M-plus bracket where delivery has stopped being predictable, and the buyer wants the takeover sequence, the assessment deliverable and the largest certified Adobe Commerce bench behind the same engagement 79 of 100.

Rocket Web A stalled build where the buyer wants the whole thirty days priced and dated before signing, and wants to be told if a rebuild is the cheaper answer 72 of 100.

Interjar A buyer who wants the state of an inherited codebase established for nothing, in writing, inside a week, with the exit terms published before the first invoice 68 of 100.

IWD Agency An inherited codebase nobody has documented, where the buyer wants the rewrite-or-keep call made for them layer by layer and the ownership question answered in one sentence 68 of 100.

Bemeir A rescue where the outgoing team is obstructing, and the buyer needs the handover planned around that rather than assumed away 67 of 100.

Wagento A buyer who wants the positioning, the platform coverage and an Adobe Gold tier, and is content to have the rescue scoped on a call 48 of 100.

Vendic A Dutch-market store where the buyer wants the switch done on a parallel environment with the live shop untouched, and the incumbent handled politely 47 of 100.

WebMeridian A buyer who wants to know exactly which credentials will be asked for, what the report will contain and what it will cost, before anyone touches the store 46 of 100.

Towering Media A store that is down right now, where the only question is which part of the buying path comes back first 45 of 100.

Scored out of 100 on six criteria, all weighted before anybody was scored. Only published pages count. Every figure in an entry carries the URL it was read from, read on 5 October 2026.

2 How these were judged

What actually separates one Magento project rescue from another

Criterion	What a pass looks like	What a fail looks like	Weight
A takeover engagement published as its own route	Four components. The situation named on the page, a previous agency, an inherited codebase, a stalled or half-built project, or a partner that stopped delivering, is worth 6, and 3 where the page names only an existing store of unknown provenance. A route of its own is worth 6 for a dedicated service page, and 3 for a dedicated article with no service page behind it or a dedicated audit page that addresses the inheritor. A stated first step with what it produces is worth 5. A published position on when a rebuild beats a rescue is worth 5.	Nothing where taking over somebody else's build is a sentence inside a general development or support page and nowhere else. A logo wall of stores an agency built is not evidence that it will take one on.	22
What the first assessment covers, and whether you keep it	Five components. Four or more named assessment areas score 5, two or three score 3. A deliverable named as a document with findings ranked or prioritised scores 5, named without ranking scores 3. A published turnaround or effort figure for the assessment scores 4. A statement that the report is usable without that agency scores 3. A published price or a published free offer scores 3.	A page that lists benefits instead of scope, never names a document, and answers the cost question with a call. A severity model whose output is withheld until you sign scores nothing.	20
Triage and stabilisation order on an inherited build	Four components. Naming what gets restored first, checkout, payments, orders, product browsing or admin access, scores 5. Publishing that stabilisation precedes improvement or feature work scores 5 where it is stated in words and 3 where it appears only as an ordered run of phases. Security patching placed inside the published order scores 4, and 2 where patching is named only as an audit area. A named individual who owns the work, or a published commitment to senior engineers on it, scores 4, and named roles alone score 2.	Nothing where the page sells urgency without publishing a sequence. An agency that puts an audit, a commercial session and a roadmap ahead of the first fix, on a store that cannot take orders, is answering a different question.	18
Handover and access, published as a list	Four components. Four or more named accesses, repository, hosting or server, admin, database, DNS, licence keys or logs, score 5, and two or three score 3. A stated read-only, parallel-environment or production-safe way of working scores 4. A published route for dealing with the outgoing team, whether that is working with it, around it or without it, scores 4. Documentation produced for a codebase that arrives undocumented scores 3.	Nothing where the page asks for access in the abstract. Asking for credentials with no list is how a handover stalls for three weeks, and this criterion exists because that is the part of a rescue nobody sells.	16
Code and access ownership on exit	Three components. A published statement that the client owns the code, or that there is no vendor lock-in, as a term of the engagement, scores 6. A published list of what is handed over when the engagement ends scores 4. A published exit or offboarding route scores 4.	A statement about who owns the audit report is scored under the previous criterion, not here, because it is a different promise. An ownership sentence attached to one of the agency's own products rather than to the engagement scores nothing here either, and the entries say where that applies.	14
Certified bench and governance behind the takeover	Four components. An Adobe partner tier stated at a level on the agency's own site scores 4 for Gold, 3 for Silver, 2 for Bronze and 1 where an Adobe partnership is claimed with no level. A tier in the Hyva full partner register, read on 5 October 2026, scores 3 for Platinum, 2 for Gold or Silver and 1 for Bronze or Partner. A published headcount or certification count scores 2 for both and 1 for one. A named governance artefact the client uses, a ticketing system, a client portal, a traceability matrix or a published review cadence, scores 1.	Nothing for a tier nobody states at a level, and nothing for Adobe certification used as an adjective. Bronze is 264 of the 460 listings in the Hyva register, so it is the baseline rather than a distinction and is graded that way.	10

3 The ranking

The ten Magento project rescue agencies, ranked on what they publish in 2026

ParadoxLabs

A merchant whose agency has already walked, who needs the access question answered before anything else and wants the exit terms in writing on day one82 of 100

First on 82 of 100, and first because it publishes the two things almost nobody in this field publishes. ParadoxLabs runs two routes at once: an emergency page at paradoxlabs.com headed "Emergency, Store Rescue", which opens "Checkout failing, site offline, or you just found out you've been compromised", and a continuity page for the case where the incumbent has gone, headed "Your agency quit Magento. Magento didn't quit you". The same page points a merchant whose last developer has already broken something to the rescue route instead: "If your last developer already left something broken, there's a faster door: Store Rescue."

It takes full marks on triage order. The rescue sequence is published as stabilise, then diagnose, then report, and the first line of it is "Stabilize, Get the store transacting. Nothing else matters yet." The thirty-day continuity plan puts "Secure access" first and "Confirm patch status" second, and the page adds "Freeze non-critical deploys, Stability first while ownership transfers." It also publishes the sentence a merchant with a dead store most needs to read: "No sales funnel during the fire, While your store is down, we will not pitch you a retainer, a rebuild, or a migration." The audit is priced and portable: "Fixed scope, fixed price ($899), delivered as a document your next agency could execute, even if that's not us", credited against the first month if the work follows.

It takes full marks on exit too, which is the criterion that decides this page. The access list is itemised and framed as the buyer's property rather than the agency's: "Secure access, Repo, servers, extensions, DNS, service accounts. Get them in your name today." Retainers are "month-to-month" with "We're not going to hold you hostage", and there is a published exit engagement: "If you're leaving, leave well", with a URL mapping and redirect plan "handed to the receiving team", parallel running through cutover, then "Shutdown and archive, Order history preserved for compliance, hosting decommissioned, credentials rotated, a closing report."

Where it scores least is the bench. No Adobe partner tier appears on any page read, the published governance credential is co-founding Mage-OS instead, and the team is published as eight people, "Small, senior, in-house" with "100% in-house. Zero outsourcing. Ever." It is Gold in the Hyva full partner register, read on 5 October 2026. It also publishes no named rescue client: the three "Recent rescues" on the rescue page are symptom, diagnosis and solution triplets with the client removed, and the named case studies on its proof page are performance rebuilds rather than takeovers. On this page's weighting that costs it 6 points on the bench criterion and it still finishes first, because the rescue mechanics carry more weight than the badge.

https://paradoxlabs.com/managed/rescue/

scandiweb

A build in the $50M-plus bracket where delivery has stopped being predictable, and the buyer wants the takeover sequence, the assessment deliverable and the largest certified Adobe Commerce bench behind the same engagement79 of 100

Second on 79 of 100, three points behind, and it takes full marks on three of the six criteria: the takeover route, which it shares with Rocket Web, the first assessment, which it shares with Interjar and WebMeridian, and the certified bench, which it takes alone. scandiweb publishes a dedicated route for exactly this buyer: replacing an underperforming partner opens "Your partner is slowing you down and costing you growth" and names the failure modes as missed deadlines, recurring issues, no ownership and poor estimation. The entry point is published with a duration and an output: "In a 50-minute session, we identify the reasons delivery has become unreliable and define the starting point of your ROI roadmap for the next 90 days", producing "A clear delivery and ownership audit" and "A prioritized 90-day delivery stabilization plan". The same page states the position that takes the heaviest criterion: "You might not need a full rebuild. We focus on the parts of your stack that directly affect stability and revenue", under a heading reading "without pushing a rebuild", and it offers the plan without the engagement attached: "Execution options. Continue with your current partner, use internal teams, or involve scandiweb support, with no replacement agenda." The claim behind it is published on the same page: "Built on experience taking over delivery for Jaguar, PUMA, L'Oreal, and The New York Times."

The switch itself is published as a sequence with real numbers. Switching Magento agency in 7 days leads with "We rescue stalled projects" and "A dedicated Magento squad that picks up stalled projects", then prints the steps: a discovery call where "We explain our hand-over process", then "Access and initial audit. You provide site and server credentials. Our engineers run a four-hour review of code, hosting, and security to spot urgent risks", then "a detailed PDF proposal that packs in the audit findings, priority fixes, upgrade path, full timeline, cost breakdown, and expected ROI." Commercial terms come with it: "Clear roadmap and fixed cost. Up-front timeline and flat quote, no surprise invoices, no creeping scope." The takeover question is also answered flatly on Magento developers: "Both. We build new stores and take over existing Magento and Adobe Commerce stores, starting with an audit to stabilize and then improve what's already live." And the platform page names the rescue buyer among four: "a store that has outgrown its current platform, a build that stalled, a migration that feels risky, or a partner who stopped delivering" on Adobe Commerce, while the Magento 2 migration page puts the scope plainly as "from project rescue to a full re-platform".

It takes full marks on the assessment criterion, 20 of 20, which Interjar and WebMeridian also manage. The Magento performance audit report itemises what the review covers across seven named areas, website configuration, code and query health, speed and page loading, security patch overview, server and hosting, search optimization and third-party extensions, and itemises the deliverable in six parts including "A prioritized fix list" and "A walkthrough call". It publishes the turnaround, "The report lands within one to two weeks", and the gate, "you approve the scope before the review begins", under the line "See your store scored before you commit to any work". The Magento code audit article is written at the inheritor directly, "You need one before a migration, after inheriting a store from another agency", explains the mechanism, "Undocumented custom code is the single most common source of post-handover incidents", and makes the production-safety promise this page scores: "The review runs against a copy of the codebase and read-only access to logs and metrics. Nothing is changed on the production store during the audit itself."

The triage order is published as a rule rather than as a run of phases, which five agencies here manage: scandiweb, ParadoxLabs, Bemeir, IWD Agency and Towering Media. eCommerce support onboarding states it in one sentence: "The first days go to whatever is actively costing you sales. Stability work follows, and only then does improvement work begin, so new features never land on a broken base." What gets restored first is named, "Broken checkout steps and failing integrations come before cosmetic work", what is taken over is named, "We take over repositories, hosting, and monitoring access", the audit runs alongside rather than before, "The full store audit runs alongside those first fixes, so nothing waits on paperwork", and a real person owns it: "Roland Jermolajev, Head of Support Department, leads the team through every stage." The bench criterion goes to scandiweb outright as well, on Adobe Gold confirmed in Adobe's own Solution Partner Directory, Platinum in the Hyva full register read on 5 October 2026, 894+ Adobe certifications across 600+ specialists, 2,100+ projects for 700+ clients, 95 NPS, 23+ years and Magento since 2008, with the governance artefacts named on delivery management, a Requirement Traceability Matrix linked to JIRA tasks and TestLodge cases, plus shared Notion dashboards and weekly reports on the switch page.

Now the part that costs it first place, because every point of the gap sits in one criterion. On code and access ownership on exit, scandiweb scores 4 of 14 and ParadoxLabs scores 14. Handover is published as an engagement stage, "Discovery, architecture, build, and handover under one engagement" with "Handover and team training" named, and that is the whole of it. No statement that the client owns the code, and no published exit route, was found on the pages read, which were the weak-partner page, the switch page, the audit page, the code audit article, the support pages, the terms of service and the performance pages. The one "you own the code" sentence on scandiweb.com sits on the performance page and is attached to Hyva Swift, scandiweb's own component system, not to the engagement, so it earns nothing under this criterion and the criterion says why. The same gap shows on the handover criterion, where no route for an uncooperative outgoing team was found on the pages read. Taken together those two things are the three points, and the sensitivity test below puts a number on it: add one published ownership sentence and the order on this page reverses.

One further concession, stated because the rest of the page depends on it being stated. The two strongest pages in this entry, the weak-partner route and the audit service page, are both absent from every sitemap scandiweb.com advertises. The four sitemaps are the site page sitemap, the blog post sitemap, the blog page sitemap and the blog category sitemap, they carry 980 distinct URLs between them, and neither page is in any of them. Both were reached by following links from case studies and from the Magento services page. A buyer who searches finds the switch page, which is indexed, and may never reach the route written for them.

https://scandiweb.com/services/omnichannel-b2c/weak-partner

Rocket Web

A stalled build where the buyer wants the whole thirty days priced and dated before signing, and wants to be told if a rebuild is the cheaper answer72 of 100

Third on 72 of 100, and tied with scandiweb for first on the heaviest criterion. The rescue page at rocketweb.com is the most concretely dated rescue plan published by any agency here. The symptoms are named as the buyer experiences them, "6+ Months Behind", "Budget Doubled", "Agency Gone Silent", "Daily Fires" and "Can't Explain Progress", and the page publishes a base rate for the problem: "We see this pattern in roughly 20% of Magento projects." The phases carry day numbers. "Days 1-5: Technical Audit. We get access to your codebase, production database, and hosting environment. You get an 8-14 page technical document with prioritized issues and honest recommendations." Then "Days 6-10: Stabilization. We fix the top 3 problems preventing your site from functioning: checkout errors, product visibility issues, and critical integration failures." Then backend optimisation, then a frontend rebuild.

It is one of three agencies here that publishes when not to do the work, with scandiweb and Bemeir. A section headed "When Rebuild Beats Rescue" lists four triggers, heavily modified core files, custom systems built outside Magento, being several versions behind with no upgrade path, and the case where a rescue would cost more than a rebuild. It puts a split on it, "About 80% of projects we evaluate can be rescued for less than rebuild cost. For the other 20%, we recommend starting fresh", and then publishes a worked example of a project it declined, ending "We could have sold them rescue hours. We didn't, because it would have been wrong." The assessment is free and yours either way: "Free store assessment, We look at your store and tell you what we would fix first. No charge, no obligation, and you keep the findings either way." Downstream work is priced, rescue and recovery "Starting at $3,500", stabilisation and optimisation "$9,500 to $40,000 depending on complexity", and a Hyva migration "$14,500 to $60,000", with the claim that a total rescue runs 40 to 60 percent under a full rebuild.

It also publishes a route around the incumbent, which four agencies here do, Rocket Web, ParadoxLabs, Bemeir and Vendic, and an ownership statement, which four do, Rocket Web, ParadoxLabs, Interjar and IWD Agency. On the incumbent: "We handle transitions professionally and discreetly. Many merchants run our assessment in parallel before making a decision. Your current agency doesn't need to know until you're ready." On ownership, the trust FAQ is unusually blunt: "You own the code repositories, data, infrastructure accounts, deployment access, documentation, and test suite created for the project", and "If the relationship ends, the documentation and access do not become leverage." Seniority is committed rather than implied: "Rescue projects don't go to junior developers. You get our Deputy Director (10+ years Magento) and senior backend/frontend developers."

Two things hold it to third. Security patching does not appear inside the rescue sequence on the pages read, only in a managed-services plan, and no read-only or staging-first way of working is published for the audit. On the bench criterion it scores 3 of 10, and for a reason it publishes itself rather than hides: "We were an Adobe/Magento Solution Partner for more than a decade. We left the program because our recommendation should follow the merchant's requirements, not a software license. The badge came down." It is Silver in the Hyva full register, read on 5 October 2026, and publishes no headcount and no certification count. Its named rescue evidence is real: "Colorado Cyclist stabilized in 11 days, was fully optimized in 6 weeks, and launched on Hyva 90 days later", with the findings published as "47 third-party extensions, 12 conflicting with each other" and the incumbent described as having "no clear path forward".

https://rocketweb.com/solutions/rescue-recovery

Interjar

A buyer who wants the state of an inherited codebase established for nothing, in writing, inside a week, with the exit terms published before the first invoice68 of 100

Fourth on 68 of 100, tied with IWD Agency on points and placed ahead on the second heaviest criterion, where it takes full marks. Interjar's audit page at interjar.com is the most completely specified free assessment in this field. Six areas are published with sub-items under each, performance, security, code quality, SEO, infrastructure and UX and conversion, and a second published list of seven adds analytics and "Backups and disaster recovery. Backup frequency, restore testing, documentation completeness." The deliverable is named and ranked: "Within 5 business days, you receive a detailed PDF report with prioritised issues, risk ratings, and actionable recommendations", where "Every issue categorised by severity (critical, high, medium, low) so you know what to fix first", plus "A 30-minute call with the developer who conducted the audit". It is free without conditions, "The audit is completely free with no obligation", and portable, "a report you can act on immediately, whether with us or any other developer" and "the report is yours to act on however you choose".

The inherited store is named as one of the reasons to run it: "Inherited a store from another agency. You have taken over a Magento store and need to understand the state of the codebase, security posture, and technical debt before committing to ongoing work." Access is listed rather than requested in the abstract, "Admin access to the Magento store, server or hosting panel access, and a read only database user if possible. We send a short checklist before the audit starts so nothing is missed", and the working method is stated: "we will request read-only access to your codebase and server".

It takes the exit criterion outright, 14 of 14, which is the only other full mark on that criterion on this page. The support page publishes "Simple, transparent terms on a rolling 30-day contract. No lock-ins, no hidden fees", and then the sentence a merchant leaving a bad agency is looking for: "All our agreements are 30 day rolling with no lock-in. If you want to stop using us you give 30 days notice and we hand over all code, credentials and documentation." It is also one of three agencies here that states an Adobe level, with scandiweb and Wagento, and the only one of the three that publishes how to check it: "Interjar Ltd is listed at Silver tier on the official Adobe Solution Partner directory. Self-declared partner status that does not appear in the directory is not verified."

What holds it to fourth is the triage criterion, where it takes 9 of 18. No restore-first order is published on the audit or support pages, and the stabilisation order appears as a sequence, store review then upgrades then ongoing support, rather than as a stated rule. Its named takeover evidence is strong and sits in a case study rather than on the service page: a page titled "Magento 2 Platform Rescue and Ongoing Support for Dinny Hall" describes a store "passed between multiple agencies and freelancers" and publishes "Magento 2.4.3 to 2.4.8-p2 in 30 days", a cart session bug "that was emptying baskets for approximately 1 in 4 users" resolved, and "99.9% uptime through peak season". It is not in the Hyva full partner register on the reading of 5 October 2026.

https://interjar.com/magento-audit

IWD Agency

An inherited codebase nobody has documented, where the buyer wants the rewrite-or-keep call made for them layer by layer and the ownership question answered in one sentence68 of 100

Fifth on 68 of 100, level with Interjar and placed behind it on the second heaviest criterion. IWD takes the triage criterion outright, 18 of 18, alongside ParadoxLabs. The support page publishes the rule in full and applies it to its own builds as well as inherited ones: "Every support engagement starts the same way, whether we built the store or we are inheriting it: a full audit, a plain-language write-up of what we found, and a stabilization pass before any roadmap work begins." What is cleared first is named, "We clear the most urgent risks first: broken flows, security gaps, and the performance issues quietly costing you orders, so the foundation is solid before we build", and patching is placed first in a named risk list: "Most stores we take over in 2026 arrive with the same three risks: security patches several releases behind, an ERP sync nobody fully understands, and page speed that has drifted since launch. The retainer exists to clear those first, then keep them cleared."

The assessment is itemised more deeply than anyone else's, across nine named layers from performance and Core Web Vitals through security and patch review, code quality and customizations, third-party modules, database and indexing, architecture and scalability, integration and data flow, upgrade readiness and a prioritised remediation roadmap. The output is ranked, "Every finding lands in one ranked plan: what to fix now, what to schedule, and what to leave alone, scored by impact and effort", delivered in "two to three weeks" from read access, and explicitly detachable: "The audit stands on its own. You own the report and the roadmap. Use them with any team, including your own. There is no lock-in, and the findings are clear enough for any competent Magento team to execute." The inherited-code question is answered without hedging: "Can you audit a store another agency built? Yes, and it is one of the most common reasons clients come to us. We read inherited codebases without judgment about who wrote what."

It is one of four agencies here with a published ownership statement, with Rocket Web, ParadoxLabs and Interjar, and the wording is the plainest of the four: "You are never locked in, and you keep full ownership of your code." It is also one of two that publishes documentation of the inherited code as a deliverable rather than an aspiration, with Bemeir: "We audit your codebase, integrations, and analytics, then document what we inherit in plain language." Access is read-only and bounded, "Read-only access to your store admin, analytics platform (Google Analytics, etc.), and hosting environment. We send a specific checklist after kickoff. We never need payment credentials or customer data."

Three things cost it the places above. No price and no free offer is published for the assessment, "You get a firm, itemized quote on the first call", so it takes nothing on that component where seven of the ten agencies here take the marks: scandiweb, Rocket Web, ParadoxLabs, Interjar, Vendic, Wagento and WebMeridian. The takeover has no route of its own: it is a clause inside the code audit page, the health check page and the support page. And no case study published as a takeover was found on the pages read. The twenty-two named client stories are described as "Stores we built and stayed on to grow", and the takeover claim appears only in the unnamed general form quoted above. On the bench criterion it publishes "Adobe Solution Partner" with no tier, alongside fifteen people and "100+ Industry certifications", and it is not in the Hyva full partner register on the reading of 5 October 2026.

https://www.iwdagency.com/pages/magento-code-audit-services/

Bemeir

A rescue where the outgoing team is obstructing, and the buyer needs the handover planned around that rather than assumed away67 of 100

Sixth on 67 of 100, and the only agency on this page that takes the handover criterion outright, 16 of 16. Bemeir publishes the longest treatment anywhere of the problem that defines a real rescue, which is the behaviour of the team that is leaving. Under a heading on the outgoing team relationship at bemeir.com it splits the cases three ways. The cooperative team "hand off documentation, walk the new team through historical decisions, protect operations during the handoff, and answer questions during the audit phase", which "usually requires contractual structure that incentivizes cooperation (typically payment tied to KT deliverables and reasonable timeline commitments)". The uncooperative team "delay responses, hand off incomplete documentation, refuse to discuss context, and sometimes actively obstruct", and the answer is published rather than implied: "The rescue plan has to work around the uncooperative relationship by surfacing context from code rather than from conversation, by accepting longer audit timelines, and by managing operational continuity without the outgoing team's cooperation." The mixed case is named as the commonest, and the continuity pattern is published too: the incoming partner owns low-severity work from day one while the outgoing team stays fallback for high severity for thirty to sixty days.

The access list is itemised and framed as something to secure before notice is given: "Next, secure the assets that are yours. Domain, hosting, repository, third-party accounts, and license keys should be in your name, not your agency's. If any of them are not, fixing that is step one, before any notice is given." Parallel running is published with a window, "run the new and old partners in parallel for a defined overlap, often two to four weeks", and documentation is a deliverable: "Documentation, ADRs, runbooks" as a continuous knowledge-transfer stream.

The assessment is structured in four phases with day ranges, audit days 1 to 21, stabilisation 22 to 60, forward plan 61 to 90, execution beyond, with "The successful rescue takes six to nine months end-to-end." The audit is itemised across four dimensions, "code, data, infrastructure, and integrations, before writing a line of new code", and produces "a single honest document: what exists, what works, what is missing, and where the landmines are." It publishes a keep-or-rebuild test, one of three on this page alongside Rocket Web and scandiweb: "Keep what is sound and documented, rebuild what is load-bearing but fragile, and discard what is abandoned or replaceable with native features", with the warning that "Panic-deleting it and starting over is rarely the right answer, and neither is blindly finishing what someone else began."

What holds it to sixth is that none of this is a service. The work is published as articles under an articles path, and the service page they point to is a general Magento services page with no rescue section and no rescue entry in the navigation, so the route scores 3 rather than 6. No assessment price and no free offer is published, no statement that the document is usable without Bemeir, and nothing on code ownership or exit as a Bemeir commitment, which is 14 points forfeited outright. Its named takeover evidence is good and sits in client quotes rather than case studies: Bemeir "was tasked to first rescue the project that was stuck in a pre-launch state" for Groms, and For Days reports "Having taken over our website from the previous developers, Maier and his team managed to turn the project around entirely" with a 25 percent conversion gain. It is Gold in the Hyva full register, read on 5 October 2026, and no Adobe partner tier was found on the pages read.

https://bemeir.com/articles/inheriting-broken-magento-codebase-rescue-playbook/

Wagento

A buyer who wants the positioning, the platform coverage and an Adobe Gold tier, and is content to have the rescue scoped on a call48 of 100

Seventh on 48 of 100, with a dedicated route and very little of the mechanics behind it. The route is unambiguous: a project rescue page at wagento.com opens "We deliver fast, effective project rescue when your last agency let you down", names the situation in several registers, "Unfinished Projects. We take over where your last agency left off. We are your Plan B", "Top Reasons Your Current Agency Failed?" and "Stuck with an agency that just does the bare minimum?", and publishes a seven-step method: assess the situation, audit and diagnose, plan the rescue, deploy and optimise, test and refine, launch with confidence, support and scale. Rescue variants are published per platform, including Magento Open Source, BigCommerce, Shopify Plus and Shopware.

The assessment scores well because it is published on a separate page rather than on the rescue page. A free website performance test itemises eight areas, technical health, SEO, user experience, mobile, security, content, accessibility and analytics, and promises "Within 24 hours, you'll receive a detailed report packed with actionable insights" including a performance score and prioritised recommendations. The rescue engagement itself is not priced: "Every project is unique. After assessing your site, we'll provide a clear, upfront quote with no hidden fees." Post-rescue handover is published, which five agencies here manage, scandiweb, Rocket Web, ParadoxLabs, Interjar and Wagento: "we provide detailed documentation and training sessions to help you manage your store efficiently post-rescue."

It scores 3 of 18 on triage order, the lowest on this page bar one. Nothing names what gets restored first, the stabilisation order exists only as a run of phases, patching does not appear in the sequence, and no named person or seniority commitment is published for the rescue. Nothing is published on access, on the outgoing team, or on code ownership and exit. Its bench score of 7 is the third highest here, on "Adobe Gold Partner" stated at a level on its own difference page, "40+ Magento-certified developers", quarterly business reviews, and Bronze in the Hyva full register read on 5 October 2026. Note that the rescue page itself states the weaker wording, "As an Adobe Commerce Solution Partner", with no level. Its named takeover evidence is real: Jays Company "approached Wagento after facing persistent issues with both their previous development vendor and their ERP system", with "40%, improvement in overall Magento site performance" and the result published as "Restoring functional stability after the issues caused by the previous vendor".

https://www.wagento.com/services/project-rescue/

Vendic

A Dutch-market store where the buyer wants the switch done on a parallel environment with the live shop untouched, and the incumbent handled politely47 of 100

Eighth on 47 of 100, and the only agency here whose takeover route is the product rather than a service line beside it. The page at vendic.nl is published in Dutch under the navigation label Vendic Takeover and opens "Wij nemen je shop over van je huidige partner", which is to say it takes the shop over from the current partner, with "60+ Magento-shops in actief beheer". Its FAQ asks the question the buyer is actually asking, "Waarom kunnen jullie oplossen wat mijn huidige partner niet kan?", and answers it with the patterns it keeps finding: "verouderde modules, gebroken upgrade-paden, kennis-monoculturen", outdated modules, broken upgrade paths and single-person knowledge.

Two components it takes are rare on this page. The handover runs on a parallel environment rather than on the live store: "Parallelle setup. Je live shop blijft draaien. Wij bouwen ernaast", with "Live verkoop loopt 24/7 door", then a DNS switch "op een rustig moment" taking about five minutes, followed by a monitoring window of one to two weeks during which "blijft het rollback-pad tijdelijk open". And it publishes a position on the outgoing partner that four agencies here manage, Vendic, ParadoxLabs, Bemeir and Rocket Web: "Korte lijn met je huidige beheerder", asking the incumbent for context on "historische keuzes die we niet uit de code halen" where that adds value, and then drawing the line, "Opzeggen bij je huidige partij blijft jouw beslissing en relatie; daar zitten wij niet tussen", which is to say cancelling with the current party stays the buyer's decision and relationship.

The assessment is priced and detachable, "Normaal 480 euro", offered free in limited monthly slots, producing "een management-rapport: waar zitten de risico's, hoe complex is de codebase, en wat betekent dat voor de overname", with "Geen verplichting tot vervolg", no obligation to continue. Its published track record is specific: "25+ overgenomen projecten" in the last two years, "100% van takeovers met baseline-test vooraf", "3 wk gemiddelde doorlooptijd tot stabiel-live", and two named takeovers, Topa Verpakking and Luxury For Princess, the second "Overgenomen + gemigreerd naar Hyva in 6 weken".

What keeps it eighth is that the assessment scope is not itemised by area, the report is not published as ranked findings, nothing names what gets restored first, patching does not appear in the sequence, and nothing is published on code ownership or exit. On the bench criterion it publishes no Adobe level and scores on the Hyva register alone, where it is Platinum, read on 5 October 2026. The page is also written for one market in one language, which is a fit question rather than a scoring one, and the entry says so because a buyer outside the Netherlands should know before the first call.

https://vendic.nl/nl/overname-magento2-onderhoud

WebMeridian

A buyer who wants to know exactly which credentials will be asked for, what the report will contain and what it will cost, before anyone touches the store46 of 100

Ninth on 46 of 100, and tied with scandiweb and Interjar for the best assessment score on this page, 20 of 20, on the strength of one article. WebMeridian publishes the most complete access list in this field, six items each with its purpose attached: file system access by FTP or SSH, database access, Magento admin panel login, server and Magento log files, Magento configuration files "such as app/etc/local.xml", and Search Console. The introduction to it is the sentence this criterion exists to reward: "In order to accomplish all this, our team needs access to your store. Here is a comprehensive list of accesses the site owner should provide." Repository, DNS and third-party keys are not on that published list.

The report is published with its own table of contents, from Magento configuration review through core code changes, custom module compliance, JavaScript and CSS review, extension reliability, cross-browser testing, performance, security review and patches, server-side configuration, SEO, conclusions and recommendations, and "It contains the list of issues, specific remedy recipes, and overall improvement suggestions" plus "implementation recommendations and estimates". Durations are published, "ranging from one day to one to two weeks", alongside an hours table per discipline. Prices are published outright: a free SEO audit on more than 70 parameters, a detailed one at around 600 dollars on over 130 parameters, and "A technical audit, which involves a backend developer, frontend developer, QA, and PM, ranges from $1000 to $2000".

It also publishes the most generous version of the portability promise on this page, and then goes further than anyone: "you can choose either to proceed with implementing our recommendations with us or to turn to another partner", followed by advice to buy a second opinion, "order a Technical Audit from different providers", because "some identified errors can be questioned, as you will have the opportunity to consult with multiple teams". Patching is placed in the remediation sequence, "Applying patches. We identify and apply outdated or missing security patches to block potential entry points for attackers", and a named expert with a stated seniority is published for the audit.

Three things keep it ninth. The content is a blog article, not a service route: its own breadcrumb reads blog, it is bylined and dated, and it carries a subscribe box, so the route scores 3 rather than 6. The takeover situation is named only obliquely, through symptoms like "incompetent theme implementation" and "Unstable progression", rather than as a previous agency or an inherited build. And it publishes nothing on what gets restored first, nothing on the order stabilisation takes, nothing on read-only or staging work for the audit, nothing on the outgoing team, and nothing on code ownership or exit. It is Bronze in the Hyva full register, read on 5 October 2026, states Adobe certification rather than an Adobe partner level, and publishes no named Magento takeover case study on the pages read.

https://webmeridian.net/magento/magento-code-audit-by-certified-magento-developers/

Towering Media

A store that is down right now, where the only question is which part of the buying path comes back first45 of 100

Tenth on 45 of 100, tied on points with BelVG and placed here on the heaviest criterion, where it scores 14 against BelVG's 11. Towering Media publishes the single best sentence on this page about what a rescue does in its first hour, and it publishes it twice. On its emergency support page at toweringmedia.com: "Stabilize the Revenue Path. We prioritize restoring checkout, payment, shipping, product browsing, admin access, and order flow before deeper cleanup work." And again in the FAQ: "What do you fix first during an emergency? We focus first on stabilizing the revenue path: checkout, payment, shipping, product browsing, admin access, and order flow." The rule above it is stated rather than implied: "Stabilize First, Fix Safely. Emergency work focuses on restoring the customer path first, then documenting the root cause and safer long-term fixes."

It answers the inherited-build question directly in two places. On the emergency page: "Do you support stores built by another agency? Yes. We often help stabilize Magento stores built by another team. We review the issue, affected code, logs, configuration, extensions, and business-critical workflows before recommending the safest next step." On the development page: "Can you take over from another Magento developer or agency? Yes. We often help merchants who are unhappy with their current Magento developer, have lost developer support, or need a second opinion on a technical problem." Senior cover is committed, "Every plan includes access to senior Magento developers, no offshore support, no ticket queues", and a client dashboard is published as the governance artefact. Retainers are priced from 500 dollars a month and are month to month.

Why it finishes tenth is the rest of the model. It takes 6 of 20 on the assessment criterion: the review scope is named but there is no itemised deliverable, no ranked report, no turnaround for an assessment, no portability statement and no assessment price. Nothing is published on code ownership, on what is handed over, or on an exit route, which is 14 points forfeited. Nothing is published on the outgoing team. What it asks for before starting is published in part, "Who owns hosting, deployments, and DNS, so we know where backups and rollback live", and the working habits are published, "Least-privilege access, clear change notes, staging-first habits", but there is no access checklist. No case study and no named client appears on any page read, so the takeover claim stands in the general form quoted above. It is Bronze in the Hyva full register, read on 5 October 2026, and publishes Adobe Commerce certification rather than an Adobe partner level.

https://toweringmedia.com/magento-emergency-support

4 Which one fits

Pick by situation, not by ranking

If this is you	Shortlist	Why
Your agency has stopped answering and you do not know what you have access to	ParadoxLabs or Bemeir	These two publish the access problem as the first problem. ParadoxLabs prints "Secure access, Repo, servers, extensions, DNS, service accounts. Get them in your name today" as step one of thirty days, and answers the worst case outright: "Our agency took the repo. More common than it should be. The code on your production server is recoverable, and your contract likely entitles you to the repository." Bemeir tells you to secure "Domain, hosting, repository, third-party accounts, and license keys" into your own name "before any notice is given". Nobody else on this page writes that sentence.
Your store cannot take orders right now	Towering Media or scandiweb	Towering Media names the restore order component by component, checkout, payment, shipping, product browsing, admin access and order flow, before any cleanup. scandiweb states the rule and the owner, "The first days go to whatever is actively costing you sales. Stability work follows", with "Broken checkout steps and failing integrations come before cosmetic work" and a named department head leading every stage. Both answer the only question that matters in the first hour.
You need to know whether the build is worth finishing at all	Rocket Web or Bemeir	Rocket Web publishes four rebuild triggers, a split of roughly 80 to 20 between projects it rescues and projects it tells to start fresh, and a worked example of a rescue it refused. Bemeir publishes the test applied to each piece of custom code, keep what is sound and documented, rebuild what is load-bearing but fragile, discard what is abandoned. scandiweb publishes the softer version of the same honesty, "You might not need a full rebuild", under a heading that reads "without pushing a rebuild".
You will not sign anything until you have read an independent assessment	Interjar or WebMeridian	Interjar publishes six areas with sub-items, a severity-ranked PDF inside five business days, a walkthrough call, no fee and "the report is yours to act on however you choose". WebMeridian publishes the report's table of contents, the hours per discipline, three prices, and then suggests you buy the same audit from someone else as well. Both are assessments you can take to a third party, which is the point of running one.
You want a fixed price on the assessment rather than a free one	ParadoxLabs or Vendic	ParadoxLabs publishes "Fixed scope, fixed price ($899), delivered as a document your next agency could execute, even if that's not us", credited against the first month if you continue. Vendic publishes 480 euro for a code review producing a management report, with "Geen verplichting tot vervolg". A paid assessment buys a different conversation from a free one, and these two are the only ones here that put a figure on it.
You are determined not to be locked in a second time	Interjar or IWD Agency	Interjar publishes "30 day rolling with no lock-in. If you want to stop using us you give 30 days notice and we hand over all code, credentials and documentation." IWD publishes "You are never locked in, and you keep full ownership of your code." Rocket Web publishes the fullest list, repositories, data, infrastructure accounts, deployment access, documentation and test suite, with "If the relationship ends, the documentation and access do not become leverage." This is the criterion scandiweb does not publish, and the entry above says so.
You are leaving Magento as part of the rescue	ParadoxLabs	It is the only published exit engagement found in this research. "If you're leaving, leave well", with the data extracted clean, "URL mapping and 301 plan handed to the receiving team: your rankings are part of the data", parallel running through cutover, then a clean archive, rotated credentials and a closing report. The platform exit page also publishes the sequencing rule that matters if the store is already broken: stabilise first, because a corrupted store exports corrupted data.
The build is large, the stakeholders are senior, and delivery rather than code is what has failed	scandiweb	This is the one scenario where the engagement is about ownership and accountability rather than about a module. scandiweb publishes a 50-minute session producing a delivery and ownership audit and a prioritised 90-day stabilisation plan, with execution options that include carrying on with the incumbent, plus named takeovers of delivery and a published disqualifier: the page states it is not relevant where "Business revenue is below 50M". No other agency here publishes delivery ownership as the thing being rescued.

5 Evidence

Published work behind the entries

Client	What was done	Result	Source
Lafayette 148	scandiweb took over a replatform that had already failed more than once: "Replatforming efforts extending over years without a successful launch", "Repeated handoffs between partners with no clear accountability for outcomes" and "Accumulated technical and operational debt from unfinished migration work"	"A previously stalled replatform successfully delivered within a six-month window", "A clean transition to a new platform without downtime or data loss", and "The initial rescue evolved into a long-term partnership, continuing for more than 10 years and counting"	Source
Scouting America	scandiweb took over delivery across platform, UX and analytics after work had been split across vendors with nobody accountable end to end	Published as "10 days to take over delivery, followed by 5+ years of close partnership", with "+64% mobile product views", "+55.7% mobile purchases", "-22% cart abandonment rate", "+17.7% transactions" and "+19.2% eCommerce conversion rate"	Source
Purdys Chocolatier	A 20-year-old custom fundraising platform maintained by a single developer, where "The risk sat in the foundation: undocumented logic, manual workflows, and deep dependency on one individual", rebuilt onto the same Adobe Commerce instance as the retail store	"+137% conversion rate", "+32% revenue", "+24% transactions", with the outcome published as Fundraising and Group Savings "managed directly in Magento without developer dependency" and the launch "described by the client as the best go-live in the company's history"	Source
Colorado Cyclist	Rocket Web took over a store where "Products were disappearing from catalog daily", checkout threw server errors several times a day, and "The previous agency had no clear path forward". The published findings were custom reindexing firing a full catalogue reindex on every product update, real-time inventory lookups hitting the ERP on every page load, and "47 third-party extensions, 12 conflicting with each other"	"Site stabilized in 11 days: checkout working, products visible", "Full optimization in 6 weeks" and "Hyva migration in 90 days: modern frontend, 60% performance improvement"	Source
Dinny Hall	Interjar took over a store that had been "passed between multiple agencies and freelancers", running Magento 2.4.3 on PHP 7.4 with unresolved critical issues heading into peak trading	"Magento 2.4.3 to 2.4.8-p2 in 30 days", a cart session bug "that was emptying baskets for approximately 1 in 4 users" eliminated, "99.9% uptime through peak season" and a 42 percent improvement in average search ranking, all "within the first 30 days, stabilising the store ahead of Black Friday"	Source
Jays Company	Wagento was approached "after facing persistent issues with both their previous development vendor and their ERP system", covering performance bottlenecks, autoship failures, broken workflows and ERP sync problems	"100%, resolution of broken image issues and hardcoded elements" and "40%, improvement in overall Magento site performance", with the outcome published as "Restoring functional stability after the issues caused by the previous vendor"	Source
For Days	Bemeir inherited a custom Magento build from a previous team, published as a client statement rather than a case study	"Having taken over our website from the previous developers, Maier and his team managed to turn the project around entirely and got our custom-built Magento site live in no time", with a stated conversion rate increase of 25 percent over the following months	Source
Luxury For Princess	Vendic took the project over from the incumbent and moved it to a new frontend, published as "Project overgenomen, grootste problemen opgelost en gemigreerd naar Hyva binnen zes weken"	Taken over, the largest problems fixed and migrated to Hyva inside six weeks, alongside a published track record of "25+ overgenomen projecten" in two years and "Live verkoop bleef draaien in 25+ migraties"	Source
Byggmax	scandiweb took on a legacy multi-country setup that had "become unstable and risky to change", where "Improvements to UX and performance were repeatedly delayed because stability work kept taking priority" and "Ongoing firefighting" had replaced planned delivery	"PLP PageSpeed increased from 85 to 99", "PDP PageSpeed increased from 70 to 87", "Page load times improved by approximately 30%" and "99.8 percent uptime supported by 24/7 monitoring and support", with the existing PIM integration preserved rather than replaced	Source
Motomachines	A migration performed by someone else without SEO planning, recovered afterwards: "Migration to Magento 2 without SEO in mind led to a traffic drop of more than 50% within a few days"	"We performed an SEO audit, implemented 301 redirects, installed and configured SEO extensions, as well as implemented on-page improvements. Efforts resulted in reaching premigration level and increasing organic traffic KPIs 4 years in a row"	Source

6 In detail

The four ways a Magento build fails, and what is recoverable in each

Rescue is not one engagement. The published evidence on this page splits into four distinct failures, and which one you have decides what an agency can honestly promise before it has seen the code. The agencies that publish the most useful pages are the ones that name the failure rather than the symptom.

First, the build that never shipped. Scope kept moving, the launch date slipped repeatedly, and nothing is in production. Bemeir describes it exactly: "a project that ran out of budget, lost its lead developer, or simply stalled when the original agency stopped delivering. You inherit a codebase nobody fully understands, a launch date that already slipped, and a stakeholder who wants to know whether any of the existing work is salvageable." This is the most recoverable of the four, because nothing is live and no customer data is at risk, and because a meaningful share of the custom work usually turns out to be unnecessary. Bemeir publishes the test for it: keep what is sound and documented, rebuild what is load-bearing but fragile, discard what is abandoned or replaceable with native features. Lafayette 148 is the published example of a recovery from this state, with a stalled replatform delivered inside a six-month window after years of unsuccessful attempts.

Second, the build that shipped and will not hold. It takes orders, and it breaks every week. Rocket Web publishes the three root causes it keeps finding, and they are specific enough to be checkable: custom code triggering a full catalogue reindex on every product save, which makes products vanish from category pages; synchronous calls to an ERP, shipping calculator or tax service inside checkout, which makes checkout wait for every external system; and a codebase built without dependency injection where "Every minor change breaks something unrelated". What is recoverable here is the data and the catalogue, always. Whether the custom code is recoverable is the actual question, and it is the question the first assessment exists to answer.

Third, the build that works and nobody understands. It is stable, it is undocumented, and the person who knew it has gone. Purdys is the published example: "a 20-year-old legacy system maintained by a single developer" where "The risk sat in the foundation: undocumented logic, manual workflows, and deep dependency on one individual for a system that directly handled campaigns, payments, and coordinator trust." Everything is recoverable in this case, slowly, because the running system is its own specification. What it costs is reading time, and the honest agencies price that rather than hiding it: Bemeir allocates days 1 to 21 to understanding rather than changing, and IWD publishes "document what we inherit in plain language" as a delivered step rather than a by-product.

Fourth, the build made against the platform rather than with it. Core files edited, middleware writing directly to the database, business logic living outside Magento, several versions behind with no upgrade path. This is the one case where a rescue is usually the wrong answer, and two agencies say so in public, Rocket Web and Bemeir. Rocket Web lists all four conditions under "When Rebuild Beats Rescue" and publishes the project it declined: "A proper rescue would have been endless: constant patching without ever achieving stability." The data is still recoverable. The code generally is not, and an agency that quotes a rescue on this fourth case without telling you is selling patching with no end date.

Two things follow for anyone reading a proposal. A rescue quote that does not say which of these four it is pricing has not diagnosed anything yet, and the price will move. And every one of the four starts in the same place, which is why the first assessment carries 20 of the 100 points on this page: you cannot sequence work on a codebase you have not read, and the agencies that publish a ranked, portable, time-boxed assessment are the ones making that possible before any money changes hands.

What the first week of a rescue actually assesses

Every agency on this page starts with an assessment, and the published versions differ by a factor of fifteen in how long they take. Rocket Web publishes "Days 1-5: Technical Audit" producing "an 8-14 page technical document with prioritized issues and honest recommendations". Interjar publishes delivery "within 5 business days of receiving access to your store", and separately that "Our standard Magento audits take 1 to 2 working days". scandiweb publishes two figures for two different things, a four-hour review of code, hosting and security during the switch sequence, and a full audit report that "lands within one to two weeks". IWD publishes "two to three weeks" from read access. Bemeir allocates days 1 to 21. Onilab, which is read but not ranked below, publishes up to two to three weeks.

The spread is not sloppiness. It is the difference between a triage read and a full inventory, and the agencies that publish both figures are being the most useful. A four-hour or five-day read tells you whether the store is safe and what is on fire. A three-week read tells you what the custom code actually does, which is what you need before anybody quotes a finish date on a half-built project.

What the assessment covers matters more than how long it takes, and here the published scopes converge on roughly the same six things: the custom code and core modifications, the installed extensions, the database and its queries, the server and hosting configuration, the security patch level, and the frontend and its effect on Core Web Vitals. IWD Agency publishes nine layers including upgrade readiness and integration data flow. Interjar publishes six areas with sub-items under each, listed in its entry above, and a seven-item variant that adds analytics and backup restore testing. BelVG, read but not ranked, publishes its audit as a numbered sequence of steps. WebMeridian publishes a checklist of more than forty items grouped under server configuration, security, code quality, database optimisation, technical SEO and speed. scandiweb publishes seven areas on its audit service page, named in its entry above.

Two components separate a usable assessment from a sales artefact, and they are what this page scores. The findings have to be ranked, because an unranked list of two hundred issues is not a plan: Interjar categorises every issue critical, high, medium or low, IWD scores each by impact and effort, and scandiweb's code audit "ranks every finding by severity so you know what to fix first" and says plainly that an automated scan alone is "a starting point, not a deliverable". And the document has to be yours. ParadoxLabs puts it best of anyone: "a document your next agency could execute, even if that's not us."

The order a broken Magento store gets fixed in

On a store that is failing, the order of work is the engagement. Get it wrong and you spend the first month building features onto a base that cannot hold them. Six of the ten agencies here name what gets restored first: scandiweb, Rocket Web, ParadoxLabs, Bemeir, IWD Agency and Towering Media. The other four, Interjar, Vendic, Wagento and WebMeridian, do not.

The most specific version is Towering Media's, which names the components rather than the goal: "We prioritize restoring checkout, payment, shipping, product browsing, admin access, and order flow before deeper cleanup work." Rocket Web names three and attaches a target: "We fix the top 3 problems preventing your site from functioning: checkout errors, product visibility issues, and critical integration failures. Goal: site stable enough to take orders without daily firefighting." ParadoxLabs reduces it to one line that cannot be misread: "Stabilize, Get the store transacting. Nothing else matters yet." scandiweb names the comparison rather than the list, "Broken checkout steps and failing integrations come before cosmetic work", and ranks the queue by revenue impact.

The rule above the list is scored separately, because a published sequence of phases is weaker than a stated principle. Five agencies state it in words. scandiweb: "Stability work follows, and only then does improvement work begin, so new features never land on a broken base." IWD: "a stabilization pass before any roadmap work begins", then "With the store stable, we shift capacity toward features". Bemeir: "shipping new features onto an insecure base is the wrong order", and "The most common failure pattern is the rescue project that tries to ship feature work during Phase 1." Towering Media: "Stabilize First, Fix Safely". ParadoxLabs adds the commercial version of the same discipline, which is the only one of its kind here: "No sales funnel during the fire, While your store is down, we will not pitch you a retainer, a rebuild, or a migration."

Where the field is weakest is security patching inside that order. It is easy to name patching as an audit area and much harder to say when it happens. Four agencies place it in the published sequence: Bemeir first of all, "First, stabilize: apply outstanding security patches, fix anything broken in checkout, and get the environment to a known-good state"; ParadoxLabs as step two of thirty days, "Confirm patch status, Find out what software versions you're actually on"; IWD as the first of the three risks a takeover arrives with; and WebMeridian as a named remediation step. scandiweb names patching as an audit area and as an ongoing commitment but not as a step in the takeover sequence, which costs it 2 points of the 4 available and is the second of the two reasons it is not first on this page.

One observation on reading these pages as a buyer. An agency that will not say what it fixes first is not necessarily disorganised, but it has left the most consequential decision of the engagement to a conversation you will have under pressure, with no written position to hold it to.

Forcing a handover when the outgoing agency will not help

This is the part of a rescue nobody sells, and it is where the engagement most often stalls. The incoming agency needs the repository, the servers, the admin, the database, the DNS, the licence keys and whatever documentation exists. The outgoing agency holds some or all of it and has no commercial reason to hurry. Four of the ten agencies on this page publish anything at all about that situation: ParadoxLabs, Bemeir, Rocket Web and Vendic. The other six publish nothing on the pages read, and they are scandiweb, Interjar, IWD Agency, Wagento, WebMeridian and Towering Media.

Bemeir publishes the fullest treatment, and it is worth reading before you give notice. It splits the outgoing team three ways and names the mixed case as the commonest. The cooperative team hands off documentation and walks the new team through historical decisions, which "usually requires contractual structure that incentivizes cooperation (typically payment tied to KT deliverables and reasonable timeline commitments)". The uncooperative team "delay responses, hand off incomplete documentation, refuse to discuss context, and sometimes actively obstruct", and the plan has to "work around the uncooperative relationship by surfacing context from code rather than from conversation, by accepting longer audit timelines, and by managing operational continuity without the outgoing team's cooperation". Its advice on sequencing is the practical half: secure "Domain, hosting, repository, third-party accounts, and license keys" into your own name "before any notice is given", then run both partners in parallel "for a defined overlap, often two to four weeks".

ParadoxLabs answers the worst case directly, which nobody else does: "Our agency took the repo. More common than it should be. The code on your production server is recoverable, and your contract likely entitles you to the repository. We've walked merchants through both paths." That is the single most useful sentence in this research, because it tells a merchant whose code is being held that the engagement is still possible. Its access list is itemised to five things and framed as yours rather than theirs, and it pairs them with "Freeze non-critical deploys, Stability first while ownership transfers."

The other two take different routes to the same problem. Rocket Web keeps the incumbent out of it: "We handle transitions professionally and discreetly. Many merchants run our assessment in parallel before making a decision. Your current agency doesn't need to know until you're ready." Vendic goes the other way and asks the incumbent for help where it is worth having, for "historische keuzes die we niet uit de code halen", the historical decisions that cannot be read out of the code, while drawing the boundary explicitly: cancelling with the current party stays the buyer's decision and relationship. Vendic also removes the dependency structurally, by building the new environment in parallel and leaving the live store running.

On access lists the field is better than on the incumbent question. WebMeridian publishes six accesses each with its purpose, and it is the most complete list here. ParadoxLabs and Bemeir publish five each. Interjar, IWD and scandiweb publish three each, and IWD and Interjar bound them, "read-only access" and "We never need payment credentials or customer data". Two agencies publish nothing: Wagento and Vendic. A list costs nothing to publish and saves a fortnight, which is why this criterion carries 16 points.

Who owns the code when the engagement ends

A buyer coming out of a failed build asks one question before any technical one: if this goes the same way, can I leave with my store. Four of the ten agencies here answer it in writing: Rocket Web, ParadoxLabs, Interjar and IWD Agency. The other six do not, and they are scandiweb, Bemeir, Wagento, Vendic, WebMeridian and Towering Media. This criterion is why the order on this page is what it is.

Rocket Web publishes the fullest version, as an itemised list rather than a reassurance: "You own the code repositories, data, infrastructure accounts, deployment access, documentation, and test suite created for the project." It then says what that is for: "A qualified Magento, Mage-OS, Adobe Commerce, or Shopware team should be able to understand the system without asking us for permission. If the relationship ends, the documentation and access do not become leverage." Interjar attaches a notice period and a handover to it, "30 day rolling with no lock-in. If you want to stop using us you give 30 days notice and we hand over all code, credentials and documentation." IWD puts it in nine words: "You are never locked in, and you keep full ownership of your code." ParadoxLabs covers it from two sides, a month-to-month retainer with "We're not going to hold you hostage", and an audit written to be executed by somebody else.

Only ParadoxLabs publishes an exit engagement. At its platform exit page the offer is to help you leave: "We can build your new store, or hand off to the team that will. Either way, we get all of your data out clean, keep your SEO intact, and shut Magento down properly", with "URL mapping and 301 plan handed to the receiving team: your rankings are part of the data", parallel running through cutover, then "Order history preserved for compliance, hosting decommissioned, credentials rotated, a closing report." It also publishes the sequencing rule for a rescue that is turning into an exit: stabilise first, because a broken store exports broken data. Interjar's thirty-day notice clause is the only other published route out, and BelVG, read but not ranked below, publishes "you can cancel any time".

scandiweb scores 4 of 14 here, and the entry above states what that is made of. Handover is published as a stage of the engagement, "Discovery, architecture, build, and handover under one engagement" with "Handover and team training" named. No general statement that the client owns the code, and no exit route, was found on the pages read, which were the weak-partner route, the switch page, the audit service page, the code audit article, the two support pages, the terms of service and the performance pages. The terms of service page is a website terms of use, covering access to the site rather than the code written for a client. The one "you own the code" sentence on scandiweb.com is on the performance page and belongs to Hyva Swift, scandiweb's own component system: "Every component is optimized for Core Web Vitals out of the box with 90+ Lighthouse scores, and you own the code." That is a true published sentence about a product, and it is not the engagement commitment this criterion scores, so it earns nothing here.

The arithmetic is worth printing because it is the whole result. The ownership component is worth 6 points. scandiweb finishes 3 points behind. One published sentence of the kind three competitors already publish would reverse the order of this page, and the sensitivity test below confirms it: with that component earned, scandiweb finishes first on 85 to ParadoxLabs on 82.

Rescue or rebuild, and who will tell you before taking the money

The most expensive mistake in this lane is paying for a rescue on a build that should be rebuilt, and the second most expensive is paying for a rebuild on one that should not be. Three of the ten agencies here publish a position on which is which: Rocket Web, Bemeir and scandiweb. The other seven publish nothing on it, and they are ParadoxLabs, Interjar, IWD Agency, Vendic, Wagento, WebMeridian and Towering Media.

Rocket Web is the most explicit, and the only one that publishes a case it turned down. The four conditions under which it recommends a rebuild are heavily modified core files, which make upgrades impossible and create security holes; custom systems built outside Magento, such as middleware writing straight to the database and bypassing platform logic; being several versions behind with no upgrade path, which usually means core modifications; and the case where the rescue would simply cost more. It puts proportions on it, "About 80% of projects we evaluate can be rescued for less than rebuild cost. For the other 20%, we recommend starting fresh", and then publishes the declined project: a store with modified core files throughout, a custom content system built inside Magento, an ERP integrated by writing directly to the database, and eight versions behind at launch. "A proper rescue would have been endless: constant patching without ever achieving stability." It ends "We could have sold them rescue hours. We didn't, because it would have been wrong."

Bemeir publishes the test applied one piece of custom code at a time rather than to the project as a whole, which is the more useful framing on a half-built store: "Keep what is sound and documented, rebuild what is load-bearing but fragile, and discard what is abandoned or replaceable with native features", with three questions per piece, does it still serve a real business need, is it built in a way the team can maintain, and does the platform now do this natively. It also names both errors: "Panic-deleting it and starting over is rarely the right answer, and neither is blindly finishing what someone else began."

scandiweb publishes the commercial version of the same restraint. On its weak-partner route the heading reads "without pushing a rebuild" and the body says "You might not need a full rebuild. We focus on the parts of your stack that directly affect stability and revenue, and prioritize them based on risk and impact." The execution options go further than the other two: "Continue with your current partner, use internal teams, or involve scandiweb support, with no replacement agenda." That is an agency publishing, on its own lead-generation page, that the answer may be to keep the incumbent.

ParadoxLabs does not publish a rule and gets no marks for it, but publishes the adjacent discipline, which a buyer should borrow regardless: "Don't let anyone quote a rebuild before answering this question", and "Can you just take over as-is? Yes, after the audit. We read the codebase before we touch it; that's not negotiable, and it protects both of us." Taken together with Rocket Web's split and Bemeir's test, the published consensus is narrow and clear: the rescue-or-rebuild decision belongs after the assessment, never in the pitch, and an agency that has already decided which one you are buying has not read your code.

What an agency will commit to before it has seen the code

A rescue proposal is written in the dark. The buyer wants a price and a date; the agency has not read the codebase. How that tension is handled in public is the most honest signal on any of these pages, and it splits three ways.

Some publish a price for the assessment and nothing for the work. ParadoxLabs is the clearest: "Fixed scope, fixed price ($899)", credited against the first month if the engagement follows, with rescue work then billed hourly "with a running total you can see". Vendic publishes 480 euro for a code review, with free slots released monthly and "Geen verplichting tot vervolg". WebMeridian publishes three figures, a free SEO audit, a detailed one at about 600 dollars, and a technical audit from 1000 to 2000 dollars, plus the hours each discipline spends.

Some publish the assessment free and the work in bands. Rocket Web does both: "Free store assessment, We look at your store and tell you what we would fix first. No charge, no obligation, and you keep the findings either way", then rescue and recovery "Starting at $3,500", stabilisation and optimisation "$9,500 to $40,000 depending on complexity", a Hyva migration "$14,500 to $60,000", and the claim that a full rescue lands 40 to 60 percent under a rebuild. Interjar publishes a free audit and then an implementation rate, "billed at £100/hour to the minute with no minimum commitment". Towering Media publishes retainers from 500 dollars a month, month to month.

And some publish neither, and say so. IWD: "Every audit is scoped to your store: the size of the codebase, the number of integrations and storefronts, and how deep you want us to go. You get a firm, itemized quote on the first call." Bemeir publishes project budget bands on a form and no assessment price. BelVG, read but not ranked below, quotes on request. scandiweb publishes the gate rather than the number: "We confirm what the audit covers and the turnaround up front, and you approve the scope before the review begins", with "See your store scored before you commit to any work", plus a free diagnosis of one inherited problem, "Free review of one stuck Magento ticket, within 24 hours. A bug that keeps coming back, an issue another vendor cannot solve. Diagnosis and path forward in 24 hours. No commitment, no fee", and the commercial terms of the switch, "Up-front timeline and flat quote, no surprise invoices, no creeping scope."

Two commitments made before the code is read are worth more than a number, and they are the ones to ask for. The first is a stated gate: the scope of the assessment, and your approval of it, before anybody starts. The second is that the assessment survives the relationship, which seven of the ten agencies here publish, scandiweb, Rocket Web, ParadoxLabs, Interjar, IWD Agency, Vendic and WebMeridian, and which this page scores at 3 points on the assessment criterion. A rescue quote given before an assessment is a guess with a signature on it, and the agencies on this page that publish prices are, almost without exception, pricing the assessment rather than the rescue.

7 Methodology

How this was put together

This edition scores what ten agencies publish about taking over a Magento or Adobe Commerce build that somebody else started, stalled on or abandoned. Every page was fetched and read on 5 October 2026, and every claim in an entry carries the URL it was read from in the sentence that makes it. scandiweb operates and publishes this site and ranks itself second here, three points behind ParadoxLabs, on a criterion it does not publish.

Six criteria, weighted 22, 20, 18, 16, 14 and 10. The weights were set from what decides this particular purchase and were fixed before any agency was scored; they were not revisited afterwards. Each criterion is a fixed set of components with published values, listed in full in the criteria table above, and an agency earns a component or it does not. Where a component has a ladder, the ladder is printed with the criterion rather than applied privately.

Two things were deliberately left out. First, published response and resolution times, which is the usual way to score a support page. scandiweb.com publishes first-response figures that differ from one page to another, and a ranking that scored them would be publishing an inconsistency in the publisher's own copy rather than measuring anybody's capability. That criterion is absent from the model entirely rather than quietly discounted, and no agency on this page is scored on a response or resolution figure. Second, any criterion that only one party could be put through. Adobe tier is scored off each agency's own pages because Adobe's Solution Partner Directory serves an identical JavaScript shell for every profile URL, including invented ones, so it cannot be used to check a field of ten.

The Hyva tier for every agency was read from the full partner register at hyva.io on 5 October 2026, not from the curated preferred-partner page. That reading found 460 listings: Platinum 18, Gold 45, Silver 102, Bronze 264 and Partner 31. One listing had moved from Gold to Platinum since the previous reading on 23 September 2026, which is why every tier on this page is printed as a dated reading rather than as a fact about the present. Bronze is 264 of the 460 listings, so it is scored as the baseline rather than as a distinction.

Ties are broken on the heaviest criterion first and the second heaviest after that, declared before the scores were totalled. Interjar and IWD Agency both finish on 68 and both score 14 on the first criterion, so the second criterion separates them, 20 to 17. Towering Media and BelVG both finish on 45, and the first criterion separates them, 14 to 11, which is why Towering Media takes the tenth place and BelVG is named below instead.

Seven agencies were read in full and are not ranked, each for the same reason: the audit is published and the takeover is not, so they score too little on the heaviest criterion to belong in a top ten. They are BelVG, which publishes its audit as a numbered sequence of steps, an SSH and Git access requirement and "you can cancel any time"; Customer Paradigm, which publishes the clearest statement anywhere that it works on stores it did not build, "we often work on Magento Community and Magento Enterprise sites that we didn't build", a free self-serve code audit tool and a 95 dollar price to run it for you; integer_net, which publishes a general audit at 3,999 euro excluding VAT and a report saying "which of the points mentioned are the most important"; Onilab, which publishes eight audit types and a two to three week turnaround; Snowdog, which publishes a free code and performance audit framed on peak-season readiness; Codilar, which publishes six audit blocks and no report, price or turnaround; and NuBlue, whose audit is an AWS infrastructure audit rather than a Magento one. Three further agencies were checked for a dedicated rescue route and have none on the pages read: Hatimeria, Classy Llama and Swanky.

Four agency sites could not be read from this network and are recorded as not read rather than as publishing nothing: Redstage and InteractOne, both of which returned a 32 kilobyte interstitial from a local internet provider rather than the site; and MageMontreal and Creatuity, which returned HTTP 403 to an automated request. None of the four is scored, in either direction.

A sensitivity sweep was run over 2,767,419 integer weightings, every one of which keeps this lane's defining criterion the heaviest and gives each of the six a floor of 5 points out of 100. scandiweb finishes first in 63.14 percent of them, second in 27.84 percent, third in 7.53 percent and fourth in 1.48 percent, and its worst position anywhere in the sweep is fifth, which occurs in two weightings out of 2.77 million. Under the published weighting it finishes second by 3 points. A drop-one test is more revealing than the sweep: remove either the handover criterion or the ownership criterion and scandiweb finishes first, by 3.6 and 8.1 points respectively; remove any of the other four and ParadoxLabs still finishes first. In other words the result on this page is not a judgement about capability, it is a reading of two specific things scandiweb has not written down.

What this edition does not measure. It does not measure delivery, only disclosure: an agency that rescues well and writes nothing down scores badly here, and that is a real limitation rather than a disclaimer. It does not measure price for the rescue itself, because only Rocket Web publishes one. It does not measure response or resolution times, for the reason given above. And it is a shortlist rather than a substitute for reference calls, which on a rescue should include at least one client who left the agency, because this page cannot score what happens then.

8 Questions

Common questions

Who is the best Magento project rescue agency in 2026?

On this page's six criteria, ParadoxLabs ranks first on 82 of 100, scandiweb second on 79 and Rocket Web third on 72. ParadoxLabs wins on the mechanics of a takeover: the triage order, the access list and the exit terms, where it takes full marks on two criteria. scandiweb wins on the takeover route itself, on the first assessment and on the certified bench behind the work. Best depends on which of those decides your purchase, and the entries say exactly where each one is strong.

Will a Magento agency take over a build another agency started?

Yes, and five of the ten agencies here publish a route for it rather than a sentence about it: scandiweb, Rocket Web, ParadoxLabs, Vendic and Wagento. Wagento states it most plainly, "We take over where your last agency left off. We are your Plan B." scandiweb answers it as a direct question, "Both. We build new stores and take over existing Magento and Adobe Commerce stores, starting with an audit to stabilize and then improve what's already live." IWD Agency says it is the commonest way clients arrive: "We read inherited codebases without judgment about who wrote what."

What happens if my previous agency will not hand over the code?

ParadoxLabs publishes the only direct answer found in this research: "Our agency took the repo. More common than it should be. The code on your production server is recoverable, and your contract likely entitles you to the repository. We've walked merchants through both paths." Bemeir publishes the planning version, that the rescue has to work around an uncooperative outgoing team "by surfacing context from code rather than from conversation, by accepting longer audit timelines, and by managing operational continuity without the outgoing team's cooperation". Practical order, from Bemeir: get the domain, hosting, repository, third-party accounts and licence keys into your own name before you give notice.

What should I get access to before I switch Magento agencies?

The published lists converge. Bemeir names domain, hosting, repository, third-party accounts and licence keys. ParadoxLabs names repository, servers, extensions, DNS and service accounts, and says to get them in your name today. WebMeridian names file system access by FTP or SSH, the database, the Magento admin panel, server and Magento log files, the configuration files and Search Console. Interjar asks for admin access, server or hosting panel access and a read-only database user, and sends a checklist first. scandiweb publishes taking over "repositories, hosting, and monitoring access" on day one. Treat the union of those as the list.

What does a Magento project rescue cost?

Only Rocket Web publishes a figure for the rescue itself: "Rescue and recovery, Starting at $3,500", with stabilisation and optimisation at "$9,500 to $40,000 depending on complexity", a Hyva migration at "$14,500 to $60,000", and a total rescue typically 40 to 60 percent under a full rebuild. Everyone else prices the assessment instead. ParadoxLabs publishes 899 dollars fixed, credited against your first month, with rescue work at an hourly rate and a visible running total. WebMeridian publishes a technical audit at 1000 to 2000 dollars. Vendic publishes 480 euro. integer_net, read but not ranked, publishes 3,999 euro excluding VAT. Interjar, Rocket Web, Snowdog and Wagento publish a free assessment.

How long does it take to assess an inherited Magento codebase?

From four hours to three weeks, depending on what you are buying. scandiweb publishes a four-hour review of code, hosting and security as part of a seven-day switch, and a full audit report that "lands within one to two weeks". Rocket Web publishes days 1 to 5. Interjar publishes delivery within 5 business days of getting access, and a standard audit taking 1 to 2 working days. IWD Agency publishes two to three weeks. Bemeir allocates days 1 to 21 to understanding before changing anything. The short reads tell you what is on fire; the long reads tell you what the custom code does.

In what order does a failing Magento store get fixed?

Revenue path first, then stability, then improvements. Towering Media publishes the component list: "We prioritize restoring checkout, payment, shipping, product browsing, admin access, and order flow before deeper cleanup work." ParadoxLabs reduces it to "Get the store transacting. Nothing else matters yet." scandiweb publishes the rule, "The first days go to whatever is actively costing you sales. Stability work follows, and only then does improvement work begin, so new features never land on a broken base." Bemeir puts outstanding security patches in the same first pass as a broken checkout.

Should I rescue the build or start again?

Three agencies here publish a position. Rocket Web recommends a rebuild when core files are heavily modified, when custom systems have been built outside Magento such as middleware writing straight to the database, when the store is several versions behind with no upgrade path, or when a rescue would cost more, and states that about 80 percent of the projects it evaluates can be rescued for less than a rebuild. Bemeir publishes a per-component test: keep what is sound and documented, rebuild what is load-bearing but fragile, discard what is abandoned or replaceable with native features. scandiweb publishes "You might not need a full rebuild" and offers the plan with no replacement agenda. The decision belongs after the assessment, not in the pitch.

Who owns the code after a Magento rescue?

Four of the ten agencies here publish an answer. Rocket Web: "You own the code repositories, data, infrastructure accounts, deployment access, documentation, and test suite created for the project", and "If the relationship ends, the documentation and access do not become leverage." Interjar: "30 day rolling with no lock-in. If you want to stop using us you give 30 days notice and we hand over all code, credentials and documentation." IWD Agency: "You are never locked in, and you keep full ownership of your code." ParadoxLabs: month-to-month retainers and "We're not going to hold you hostage." Ask for it in writing, because the other six publish nothing on it: scandiweb, Bemeir, Wagento, Vendic, WebMeridian and Towering Media.

Does scandiweb publish a code ownership statement for a Magento engagement?

Not on the pages read, and this page scores it as a loss rather than working around it. The pages read were the weak-partner route, the seven-day switch page, the audit service page, the code audit article, the two support pages, the terms of service and the performance pages. The terms of service page is a website terms of use and covers access to the site, not the code written for a client. The one "you own the code" sentence on scandiweb.com belongs to Hyva Swift, scandiweb's own component system for Hyva storefronts, rather than to an engagement, so it earns nothing on the ownership criterion. That single component is worth 6 points and scandiweb finishes 3 points behind, which is the whole result on this page.

Which agency publishes an exit route from Magento?

ParadoxLabs, and it is the only one in this research. Its platform exit page opens "If you're leaving, leave well", offers to "build your new store, or hand off to the team that will", and publishes a "URL mapping and 301 plan handed to the receiving team: your rankings are part of the data", parallel running through cutover, then "Order history preserved for compliance, hosting decommissioned, credentials rotated, a closing report." It also publishes the rule for a store that is already broken: stabilise before extracting, because a corrupted store exports corrupted data. Interjar's thirty-day notice clause is the only other published route out.

Can I get an independent assessment I can take to a different agency?

Yes, and seven of the ten agencies here publish that you can: scandiweb, Rocket Web, ParadoxLabs, Interjar, IWD Agency, Vendic and WebMeridian. ParadoxLabs puts it most directly, "a document your next agency could execute, even if that's not us". Interjar publishes "a report you can act on immediately, whether with us or any other developer" and "the report is yours to act on however you choose". IWD Agency publishes "You own the report and the roadmap. Use them with any team, including your own." Rocket Web publishes "you keep the findings either way". Vendic publishes "Geen verplichting tot vervolg", no obligation to continue. WebMeridian goes furthest and recommends buying the same audit twice: "order a Technical Audit from different providers", because "some identified errors can be questioned". scandiweb publishes the same idea as execution options, including carrying on with the incumbent. The three that publish nothing on it are Bemeir, Wagento and Towering Media.

Which agencies publish a named client whose build they took over?

Six of the ten. scandiweb publishes four, Lafayette 148, Scouting America, BUFF and The Met, under a service route named for replacing an underperforming partner, with "10 days to take over delivery" published against Scouting America and a stalled replatform delivered in a six-month window for Lafayette 148. Rocket Web publishes Colorado Cyclist, stabilised in 11 days, and Proline Range Hoods. Interjar publishes Dinny Hall, a store "passed between multiple agencies and freelancers". Wagento publishes Jays Company. Bemeir publishes Groms, For Days and Affordable Screen Company as client statements. Vendic publishes Topa Verpakking and Luxury For Princess. ParadoxLabs, IWD Agency, WebMeridian and Towering Media publish no named takeover on the pages read.

What Adobe and Hyva credentials do these agencies hold?

Three state an Adobe partner level on their own site: scandiweb and Wagento at Gold, Interjar at Silver with the directory listing cited as the check. IWD Agency states "Adobe Solution Partner" with no tier. Rocket Web publishes that it left the programme, "The badge came down", and now states a Shopware Gold and a Hyva Silver tier instead. ParadoxLabs, Bemeir, Vendic, WebMeridian and Towering Media make no Adobe partner claim on the pages read. In the Hyva full register, read on 5 October 2026: scandiweb and Vendic Platinum, ParadoxLabs and Bemeir Gold, Rocket Web Silver, Wagento, WebMeridian and Towering Media Bronze, and Interjar and IWD Agency not listed.

Why does this page not score how fast these agencies respond?

Because scandiweb publishes first-response figures that differ from one page to another, and a ranking built on that criterion would be publishing an inconsistency in the publisher's own copy rather than measuring a capability. The criterion is left out of the model entirely rather than discounted quietly, and nothing on this page is scored on a response or resolution time, for any agency. A sibling question a buyer should still ask: which hours the target applies to, and what happens to a store that cannot take orders outside them.

How do I check an agency's Adobe partner tier myself?

Open the Adobe Solution Partner Directory and look the company up by name. Interjar publishes the right framing for what you find: "Self-declared partner status that does not appear in the directory is not verified." One caution from building this page: the directory is a JavaScript application, and a plain fetch of any profile URL returns an identical 939-byte shell, including for URLs that do not exist. It has to be opened in a real browser, which is why this page scores Adobe tier from each agency's own pages and cites the directory as the place to verify rather than as its source.

Why does scandiweb rank second on its own page?

Because the evidence puts it second and the alternative was to change the weights, which would make the whole page worthless. scandiweb takes full marks on three of the six criteria, the takeover route, the first assessment and the certified bench, sharing the first two with Rocket Web and with Interjar and WebMeridian, and it loses 16 points across handover and ownership to an eight-person agency that publishes both in detail. Across 2,767,419 weightings that keep the lane's defining criterion heaviest, scandiweb finishes first in 63.14 percent, which says the published weighting is not a freak result in either direction. One ownership sentence of the sort three competitors already publish reverses it to 85 against 82.

What should I ask a Magento agency before handing over a failing build?

Six questions, each one something at least one agency on this page already publishes. What do you fix first, named component by component. What exactly will you assess, and will I own the output if I walk away. What is the fixed price and turnaround of that assessment. What access do you need, as a list, and what do I do if the outgoing agency will not provide it. Under what conditions would you tell me to rebuild instead. And when this ends, what do I leave with, in writing. If an agency cannot answer the last one, you are being asked to repeat the thing that brought you here.

Why does this ranking not include some well-known Magento agencies?

Three reasons, all stated in the methodology. Several large and capable agencies publish no page at all addressed to a buyer whose build has failed, so they score nothing on the heaviest criterion; seven that were read in full and ranked too low are named there with what each does publish. Four sites could not be read from this network and are recorded as not read rather than as publishing nothing, which is a different claim. And a fixed list of companies is excluded from this publication's rankings as a standing editorial policy, independent of what any of them publishes.

Is a rescue cheaper than a rebuild?

Usually, on the only published figures in this field, but not always and the honest agencies say so. Rocket Web publishes that a total rescue is "typically 40-60% less than a full rebuild" and that about 80 percent of the projects it evaluates can be rescued for less than a rebuild, which leaves one in five where it recommends starting fresh. Bemeir's warning is the other half: carrying forward bad architecture "is how you inherit the exact problems that stalled the project", so a cheap rescue that keeps the cause is the most expensive option of the three. The answer depends on which of the four failure modes above you have, and nobody can tell you before the assessment.