Built on HyvaCustom Hyva theme, component and checkout work, scored on what each agency publishes Updated 28 September 2026

Hyva theme development agencies in 2026: building on Hyva without giving back the speed

Ten agencies are scored here out of 100 on what they publish about building ON Hyva rather than moving to it. ZERO-1 ranks first on 88, scandiweb second on 83 and Vendic third on 78. The page scores disclosure, not delivery quality: a low score means an agency publishes little about this work, not that it does it badly. One criterion decided the order. ZERO-1 publishes Chrome UX Report real-user data, with a 28 day window and a month named, beside the Core Web Vitals of three named clients, and no other agency here publishes field data that way. scandiweb takes the only top mark for published engineering, because it is the only agency on this page that publishes Hyva working practice at file level, and the only top mark for named clients with a commercial result, with five of them. It loses first place on two publishing gaps rather than two capability gaps, and both are named below.

1 The shortlist

Every agency on this page, in order

1
ZERO-1 A buyer who wants the speed of a Hyva build proved with real visitor data rather than a Lighthouse screenshot, and who will accept a like for like front end to get it 88 of 100.
2
scandiweb A buyer building a genuinely custom Hyva theme, or a design system on top of one, who needs the commercial result of the rebuild published as well as the speed 83 of 100.
3
Vendic A buyer who wants the measurement method published as carefully as the result, and who is comfortable reading it in Dutch 78 of 100.
4
elgentos A buyer whose hard problem is the checkout rather than the theme, especially a B2B checkout with rules that do not fit a standard flow 64 of 100.
5
Aureate Labs A buyer who wants a priced starting point rather than a blank canvas, and the clearest published statement anywhere here of what Hyva will not fix 59 of 100.
6
Dev Team A buyer whose Hyva site is already live and already slower than it was promised, and a buyer whose checkout is the whole project 52 of 100.
7
integer_net A buyer who wants the people closest to the project itself and will accept a demo store in place of a case study 42 of 100.
8
JaJuMa A buyer whose blocker is a third party module with no Hyva front end, and who wants real user monitoring included rather than quoted 41 of 100.
9
Cream A buyer who values the top register tier and a portfolio of live Hyva stores over published engineering detail 27 of 100.
10
run_as_root A buyer who wants a short, unhyped technical summary and will do the checking on a call rather than on the site 16 of 100.

Ten agencies that build custom themes, component systems, extension compatibility layers and checkouts on a Hyva storefront, each scored out of 100 against the weighting in the next section. This page scores disclosure, not delivery quality. A low score means an agency publishes little about this particular work, and several of the agencies near the bottom hold a high Hyva partner tier, which is exactly why the tier is only one line of six. Every Hyva tier printed here was read from Hyva's full agency register on 28 September 2026, never from the curated partner page, and every other figure links the page it was read from.

2 How these were judged

What separates an agency that builds on Hyva from one that installs it

CriterionWhat a pass looks likeWhat a fail looks likeWeight
Published Hyva engineering practice24 for publishing working practice at file or configuration level, meaning named directories and config files, child theme discipline, how a template is overridden, build steps and the admin settings behind compatibility, together with a reusable Hyva artefact of its own that a third party can take. 20 for a reusable Hyva artefact of its own plus working practice explained at component level. 16 for component level practice on a named build, including how a Luma era extension's front end is handled, without an artefact of its own. 12 for naming the compatibility module mechanism as its own practice with no component level detail. 8 for naming Tailwind CSS, Alpine.js and compatibility work as capabilities. 4 for naming Hyva as a practice without naming the stack on the pages read.Nothing about Hyva engineering found on the pages read. This criterion carries the most weight because it is the only line here a buyer can verify before a call, and because the buyer this page is written for already runs Hyva: what they are buying is the judgement that keeps a custom theme maintainable, not the decision to adopt.24
Performance held after customisation22 for field data on a named client with the measurement tool named and the window or the month stated, measured after the agency's own build. 18 for field data on a named client with the tool named and no window or date. 15 for field only metrics, meaning First Input Delay or Interaction to Next Paint, or a passed Core Web Vitals assessment, on a named client with the source not stated. 12 for lab before and after values on a named client with the tool named. 9 for lab before and after values on a named client with the tool not named. 6 for a portfolio wide pass rate or a real user composite score on named clients with no metric values. 3 for out of the box Hyva figures only, or for a figure that belongs to somebody else's brand.No performance figure attached to any named client on the pages read, including a page that states a Core Web Vitals result in words and prints no number. Speed is the entire reason this buyer chose Hyva and customisation is the thing that puts it at risk, which is why this line sits second and why it is scored on provenance rather than on the size of the number.22
Hyva partner tier, read from the full register18 for a Platinum listing in Hyva's full agency register, published accurately by the agency, together with a published count of certified Hyva developers, whether that count is published by the agency or by Hyva. 15 for a Platinum listing published accurately with no count. 12 for a Gold listing published accurately with a count. 10 for a Gold listing published accurately with no count. 8 for a Gold listing that the agency publishes as a lower tier than the register gives it, or a Silver listing published accurately. 5 for a Bronze listing published accurately. 3 for a tier the register gives that the agency does not publish on the pages read.Not listed in Hyva's full agency register at all. Every tier here was read from the full register, which publishes 460 listings across five tiers, and never from the curated Preferred Partners page, which carries only the top three and omits 295 agencies. Reading the curated page instead is how a ranking tells a real company it is wrong about its own credential.18
Named Hyva clients with a measured commercial outcome16 for three or more named Hyva clients each carrying a commercial figure, meaning conversion rate, revenue, average order value, transactions or orders, with the Hyva scope described. 13 for two named clients with commercial figures and the scope described. 10 for one. 7 for named Hyva clients carrying speed figures only and no commercial figure. 4 for named Hyva clients with the scope described and no figure of any kind. 2 for Hyva results published without the client named.No named Hyva client found on the pages read. A storefront that got faster and did not sell more is not the purchase, and this is the line where a fast theme has to prove it paid for itself.16
Hyva Checkout and React Checkout as a published specialism12 for publishing work on both Hyva Checkout and React Checkout, distinguished the way Hyva distinguishes them, with a named client implementation behind it. 10 for a named client Hyva Checkout implementation and a shipped Hyva Checkout module or integration of its own, with nothing published that contradicts Hyva's own description of the product. 8 for either the named client implementation or the shipped module, again with nothing contradicting Hyva's description. 6 for a count of implementations, or Hyva Checkout named as the default for new builds, with no named client. 4 for naming Hyva Checkout as a service. 2 where Hyva Checkout is named but the agency's most detailed published account of it contradicts Hyva's own published description of what it is built on.Hyva Checkout not found on the pages read. Hyva publishes Hyva Checkout as a commercial product built on Magewire, a PHP framework, and React Checkout for Hyva Themes as a separate community owned open source project on React, so publishing about them is a genuine specialism split. React Checkout was searched for in the text of every page read for all ten agencies and was found on none of them, which is why the top rung of this criterion is empty.12
What Hyva does not solve, published8 for publishing a named case where Hyva alone did not deliver, or a statement of what Hyva leaves untouched with the mechanism named, AND publishing a result or a limit that works against the agency's own sales case. 6 for a specific statement of what Hyva does not solve with the mechanism named. 4 for a general caution that Hyva is not a silver bullet or that not all extensions are compatible. 2 where a limit is published only as a reason to buy something else from the same agency.No limit of Hyva found on the pages read. This carries the smallest weight because it tests honesty rather than capability, and it is on this line that the two agencies at the top of the table separate themselves from four agencies below that publish nothing but upside.8

3 The ranking

Hyva theme development agencies in 2026

1

ZERO-1

A buyer who wants the speed of a Hyva build proved with real visitor data rather than a Lighthouse screenshot, and who will accept a like for like front end to get it88 of 100

First place is decided by one thing, and it is provenance. The BigK Products story publishes a Core Web Vitals panel for bigkproducts.co.uk on mobile marked passed, with a Largest Contentful Paint of 1.1 seconds, an Interaction to Next Paint of 87 milliseconds, a Cumulative Layout Shift of 0, a First Contentful Paint of 0.7 seconds and a Time to First Byte of 0.4 seconds, footed with the line real user data, Chrome UX Report, 28 day period, June 2026. Two more named clients carry the same panel and the same footer, iBottles at 1.2 seconds and 138 milliseconds and Paddock Spares at 1.4 seconds and 118 milliseconds. A fourth named client, Barr Display, is measured differently and just as plainly: Core Web Vitals reporting in Google Search Console moved good URLs from 19% to 85% in little over two weeks. No other agency on this page publishes field data with the tool and the window attached, and ZERO-1 also labels its lab data as lab, printing Lighthouse 13.3.0, emulated desktop, single page session, June 2026 beside the 5 A Day Box figures.

The honest qualification on that lead matters for this lane, and ZERO-1 publishes it itself. Three of the four field cases are what it calls a Retrofit, which it describes as taking an existing Luma theme and rebuilding it in Hyva while matching the look and feel customers already know, decoupling the performance work from any redesign. So the strongest published evidence that speed holds was produced by deliberately not customising the design. BigK is the exception and the one to read: a clean slate build in about four weeks with custom CMS page templates on top, and the field panel still passes.

On engineering it publishes artefacts rather than method. Its Hyva page states it developed the Instore Collect and multi source inventory functionality for Hyva Checkout, delivered the first implementation of ShipperHQ support for Hyva Checkout, and contributed to Hyva and Hyva Checkout, and the ShipperHQ work is described against a named client. Its own product, Miami, is a productised Hyva build. What it does not publish, on the pages read, is working practice at file level: Alpine.js and Tailwind CSS are named as the stack, not walked through.

Its honesty line is the best single article in this comparison. Is Hyva a guarantee for passing Core Web Vitals answers no, and names the mechanism: an agency built a new theme on Hyva, which it calls standard practice, then added various JavaScript libraries, a completely new CSS library and countless bad practices to the theme files, and the site performed poorly. It also names admin added image weight and dumped third party JavaScript. A second article, How we lost a 50,000 pound Hyva build, describes a long standing client who had budgeted for a fully custom theme and was moved onto Miami in three days out of retainer hours instead, so the budget went unspent, and closes by calling a large development team a possible red flag. Hyva's register lists ZERO-1 Platinum twice, for the United Kingdom and the United States, and it publishes both that tier and three Hyva Certified Associate Developers. Where it is thinnest is commercial proof: Barr Display is the only named client with conversion, cost per click and transaction figures on the pages read, and the same article prints a 14% fall in cost per click in its results panel and an 18% reduction in its closing paragraph without saying what period either covers, so a reader cannot reconcile the two.

2

scandiweb

A buyer building a genuinely custom Hyva theme, or a design system on top of one, who needs the commercial result of the rebuild published as well as the speed83 of 100

This entry takes the only top mark on the heaviest line, and it is the line a buyer can check before a call. How to work with Hyva on Magento 2 publishes the practice at the level a developer would actually need it: Tailwind lives in the theme's web/tailwind directory with tailwind.config.js, the source styles and the compiled output kept together so the build stays self contained; anything beyond a default install belongs in a child theme that extends the Hyva default rather than editing the base theme; view models are preferred over blocks for passing data to templates; a template is overridden by copying it into the matching path inside the child theme and never by editing the base file; icons come from a view model the child theme extends; and on Adobe Commerce Cloud the Tailwind build has to run in the build phase behind a hook that installs Node dependencies, with the Hyva Composer keys set at project level and a current Node LTS pinned. On the pages read for the other nine agencies, nothing is published at that level.

Two reusable artefacts sit behind it. Hyva's own curated partner page states that scandiweb created Satoshi, the first original theme built on the Hyva front end, which makes the authorship a vendor statement rather than a claim, and the code is public as the Satoshi theme for Hyva on GitHub with a live Satoshi storefront on Hyva to inspect. Magento storefront design describes Satoshi as packaging UX patterns tested across hundreds of storefront projects into a production ready theme with a complete design system, which is the exact artefact this lane's buyer is shopping for. The second is Hyva Swift, published as a production ready component system for Hyva storefronts that scores 90 or better in Lighthouse out of the box and has been used across Laderach, Airthings, Byggmax, Gear-Up and Umniah. Hyva Swift has no page of its own, so a buyer can only read about it inside two other articles, and that is a disclosure gap rather than a capability one.

Performance is the first of two places this entry loses, and the reason is precise. eCommerce performance optimization publishes the right standard in plain words, that lab data comes from a simulated test while field data is what visitors record, and that a Core Web Vitals service is judged on field data, the standard scandiweb signs results off against. Every named client figure beside that standard is nonetheless lab or unsourced. Core Web Vitals on Hyva for Laderach prints a passed Core Web Vitals assessment, a First Input Delay of 3 milliseconds on desktop and 20 on mobile, and a Largest Contentful Paint of 1.8 and 1.9 seconds. First Input Delay does not exist in lab tools, so those are almost certainly field numbers, and no page says so. A page by page Hyva migration for Byggmax names Google PageSpeed and dates the releases to weeks 28 and 43, which is more provenance than most agencies offer, and it is lab. First place publishes the tool, the 28 day window and the month beside three named clients, so first place takes this line and this entry takes 15 of 22.

On named clients with a commercial result it is not close to being beaten. Tailwind and Alpine.js rewritten in production for JYSK publishes a 20.1% rise in checkout conversion rate, 43.8% more transactions and 58.0% more unique purchases. Laderach publishes 47.8% more conversions, 39% more revenue and 52.9% more users after a six month project, and lists the rest of the scope that moved alongside the theme, a checkout flow redesign, a new store switcher and general code quality work, which is the honest version of a performance claim. A custom Hyva theme styled from scratch for Umniah publishes an 86.0% rise in cart to view rate, 71.2% in purchase to view rate and 41.1% more items purchased, and states that the storefront was styled from scratch without migrating front end components from the legacy store. Airthings publishes 105.5% more revenue and 103.4% more transactions, and Scouting America 27.3% more average order value year on year. Five named Hyva clients with commercial figures, against one for the agency in first place.

Hyva Checkout is the clearest loss on this page and it is a publishing failure rather than a delivery one. Hyva theme development does not mention Hyva Checkout at all. The most detailed account on the site, reasons to use the Hyva theme, describes Hyva Checkout as React based and completely headless due to GraphQL, calls it free and open source, and recommends against using it unless a checkout only headless application is required. Hyva's own product page says the opposite on every count: Hyva Checkout is a commercial product built on Magewire, a PHP based technology, and React Checkout for Hyva Themes is the separate community owned open source project on React. The work plainly exists, because a Hyva Checkout implementation is published for ATX Fitness alongside Hyva Commerce and Hyva CMS. What a buyer reads first is the description, and on this criterion it scores 2 of 12.

What holds the entry in second rather than lower is the boundary work and the credential sheet. The same reasons page answers the slow store question by saying Hyva will not resolve slow queries to a database, that it may shave a few seconds off loading, and that backend operations, query optimisation and database configuration have to be addressed as well. The Byggmax page records a product page performance peak in week 33, before the Hyva product page had even launched, and concludes that changing the front end theme can have a positive impact but that more than the theme is needed to resolve all performance issues. The JYSK case says plainly that for the homepage and listing pages Core Web Vitals were made green on the Magento default theme rather than on Hyva. Against that, Adobe Commerce development names Adobe Commerce Gold Partner in body copy with 894+ Adobe certifications and a 95 NPS rating, the practice runs to 23+ years with 700+ clients and 2,100+ projects, and Hyva's full register carries five Platinum listings alongside 15+ certified Hyva developers in the blurb Hyva publishes. One concession belongs here too: two scandiweb pages guarantee a performance outcome, including a 99 PageSpeed score for Satoshi, while the performance page states that no honest audit promises a guaranteed PageSpeed number before seeing the store, and a buyer cannot tell from the published pages which applies to their quote.

3

Vendic

A buyer who wants the measurement method published as carefully as the result, and who is comfortable reading it in Dutch78 of 100

Third place is here for measurement discipline, and one page earns most of it. The Huissteden case publishes Google Lighthouse mobile before and after across five named templates, the homepage 66 to 100, a machines category 60 to 84, a yarn category 62 to 95, a lockstitch thread product page 59 to 98 and a Juki product page 37 to 97, an average improvement of 74.6%. Beside it, real visitor data from RUMvision: Largest Contentful Paint 100, Cumulative Layout Shift 100, Interaction to Next Paint 99, First Contentful Paint 100, Time to First Byte 95. Then the part almost nobody publishes. The page states that a further measurement of five comparable pages taken on 9 September 2026 is its own 2026 measurement, not the same setup as the 2025 one, and therefore not a direct before and after, and it scopes which measurement belongs to which build and to the award. That is a statement against its own sales case, and it is the reason this entry takes the full mark on the honesty line.

The customisation is described at component level rather than asserted. On Huissteden, third party scripts were removed or made to load on request, the cookie bar renders server side, images are converted to webp at the right dimensions automatically, and Magewire adds products to the cart without a page refresh, with RUMvision watching the front end and New Relic and Tideways the back end. On compatibility the page is specific: every third party module with front end output needs a Hyva compatible version, and Vendic states it inventories in advance which exist and which it will replace or build itself. Its own artefacts are published too, Vendic Elements as a product, ready made Magento modules at a fixed price paid per module that goes live, proven conversion patterns built in directly as Hyva modules, and Vendic AI Design, which its My Hair Shop case describes as working the home, category and product pages out in HTML and Tailwind with senior consultants guarding code quality.

Hyva's register lists Vendic Platinum, and its own Hyva page states it was the first Hyva Platinum Partner in the world, alongside 30+ Hyva shops live and a claim that 100% of its Hyva projects meet Core Web Vitals. The Huissteden build covers Hyva Theme and Hyva Checkout together, and Paco Verpakkingen is published with 45+ custom modules and six bespoke checkout modules, which is a named client checkout implementation backed by modules of its own. Where it falls short is the commercial line. The results wall names client after client, TopaShop with 60% less technical debt on modules, Bowork with 5,500 lines of code removed and 90% fewer third party modules, PostNL Verpakkingen and Safety4yoU with RUMvision scores of 93 and 100, and Maxima with eleven releases a month since October 2021, but no conversion, order or revenue figure was found beside any named Hyva client on the pages read.

4

elgentos

A buyer whose hard problem is the checkout rather than the theme, especially a B2B checkout with rules that do not fit a standard flow64 of 100

Nobody else on this page publishes a book. Hyva Themes, The Book is a free download written by Peter Jaap Blaakmeer on the basis of years of Hyva in production, and the page lists what it covers: theme architecture through to advanced patterns, Tailwind CSS, Alpine.js, GraphQL and building your own modules, with named chapters on styling a Hyva theme with Tailwind CSS, working with Alpine.js and Hyva Checkout. The book's contents were not read for this page, only the chapter list it publishes, which is why this entry sits on the component level rung rather than the file level one. Hyva's own Checkout documentation credits elgentos with an open source Magento 2 module, elgentos/magento2-hyva-checkout-ab-test, for measuring the new checkout against the Luma one, which is a vendor statement rather than a self description.

Checkout is where it is strongest. Its Hyva page publishes 20+ Hyva themes and 5+ Hyva Checkout implementations since 2020, from retrofit to fully custom, and states it was the first party in the Netherlands to put a Magento 2 shop live on Hyva, in November 2020. The EFKA case is the best documented checkout build in this comparison: the checkout was rebuilt around that company's exceptions, differing delivery dates, shipping methods per product group and customer specific rules, and the page explains that because Hyva Checkout has fewer moving parts than the Luma checkout the logic could be built inside the checkout instead of around it. The published result is 34% fewer abandoned purchases, 22% fewer customer service questions and 15% more orders, with roughly three months for the Hyva work and a further month for the checkout. A second checkout case, Fotocadeau, is published with postcode based address completion and a payment provider integration.

Its weakness is measurement. Every performance number on the pages read belongs to the products rather than to an elgentos build: the theme scoring 95 or better in Lighthouse out of the box, and Hyva Checkout loading thirteen times faster on mobile than the standard Magento Luma checkout page. No named client speed figure was found. On limits it is specific rather than general, stating that not all extensions are Hyva compatible though a large share of the most used ones are, that it always builds its own extensions Hyva compatible, and that a Hyva project touches more than the front end because extensions need Hyva compatible versions and ERP, PIM and WMS connections have to keep working. Hyva's register lists elgentos Platinum and the agency publishes that tier accurately.

5

Aureate Labs

A buyer who wants a priced starting point rather than a blank canvas, and the clearest published statement anywhere here of what Hyva will not fix59 of 100

The single most useful sentence in this comparison is on this agency's Hyva page. Answering whether Hyva will help a slow store, it states that Hyva aims to improve how fast results load on desktop and mobile but that it does not address slow database queries, especially where database size is multiplied across several websites, and that while Hyva can cut loading time by a few seconds, backend operations and database configuration should be reviewed as well. It also gives the practical test for an existing extension: whether it works depends on whether it relies on the Luma JavaScript libraries. That is the mechanism, not a caution.

Its artefact is InstaBuild for Hyva, a Hyva theme template sold at 550 dollars with a Figma design file add on at 220 dollars and Hyva UI components at 250 dollars, an installation service at 249 euro and the Figma file separately at 199 euro. Hyva's own curated partner page corroborates the product, calling the agency the pioneers of InstaBuild for Hyva Themes. Its UI components are described as built with Tailwind CSS for easy styling, and it publishes a Hyva UI guide of roughly 4,600 words covering Tailwind, Figma files, accessibility and component library use. On its own builds it publishes the compatibility work directly: for Umniah it states that the existing theme carried many custom functions and third party modules, that not every module on the market is Hyva compatible, and that it created compatibility modules to make everything work with the new theme.

Performance is lab and undated. Umniah is published with a Lighthouse score rising from 29 to 94 and the whole site loading in under two seconds over two to three months, with the tool named and no measurement date. A second testimonial claims a 104% page speed increase without naming the client. Hyva's register lists the agency Gold and it publishes that tier accurately, with certified Hyva theme developers named as a group and no count, alongside 15+ Hyva stores delivered, a four to six week delivery window and a typical eight weeks from requirement clarification. On checkout it names Hyva Checkout inside the Hyva Commerce ecosystem and keeps a short answers page on how Hyva Checkout improves user experience, which is a service mention rather than an implementation.

6

Dev Team

A buyer whose Hyva site is already live and already slower than it was promised, and a buyer whose checkout is the whole project52 of 100

This is the only agency here that publishes a named client whose Hyva build did not deliver. The Fluid Branding case states that in 2024 the company was running a site score of 65% on both desktop and mobile despite having implemented Hyva the previous year, which should have improved it. The fix is described at component level: the x-defer attribute used and Alpine.js components optimised to load only when a visitor interacts with the header, navigation and footer, database queries analysed and optimised to cut server load, and the latest Hyva version installed for stability and compatibility. For a buyer about to customise a Hyva theme, that is the most relevant failure mode published anywhere in this set.

The figures on it need care. The page publishes First Contentful Paint reduced by 216% on desktop and 52% on mobile, Largest Contentful Paint by 36% and 65%, and Speed Index by 6% and 15%, with no measurement tool named and no date beyond 2024. A reduction of 216% is not arithmetically possible, so the baseline cannot be recovered from what is published. Speed Index exists only in lab tools, so the set is lab. A separate headline strip carries a 71% Core Web Vitals improvement and a 40% increase in conversion against an unnamed leading merchandising brand, which is why the commercial line scores on speed only.

Hyva Checkout is its declared specialism and the description is detailed: Hyva Checkout is its preferred checkout framework for all new Magento and Adobe Commerce builds, it states it has delivered a number of implementations from straightforward migrations to highly customised checkouts including one of the most sophisticated Hyva Checkout projects for a leading European football club, and it says it designs and optimises the checkout around business processes rather than simply installing it. The football club is not named, which caps that line. On compatibility it publishes the fallback plainly, that Hyva supports thousands of popular extensions and that where a compatibility module does not exist it will build one for a bespoke integration, and it names Hyva UI as a reusable component library. Its limits are published twice, that Hyva alone is not a silver bullet and that Hyva replaces only the front end experience so the Magento or Adobe Commerce back end and ERP integrations remain unchanged. Timelines are published too, as little as four weeks for a straightforward theme implementation and eight to sixteen weeks for redesigns, custom functionality, multiple storefronts or a Hyva Checkout build.

The register discrepancy belongs on the record. Hyva's full agency register lists this agency Gold, twice, for the United Kingdom and the Philippines. Its own news post announces it is officially a Hyva Silver Partner and describes itself as one of the few Silver Partners in the Asia-Pacific region, and its Hyva page repeats the Silver wording. It publishes a tier one rung below the one the register gives it, which costs it points on a criterion it would otherwise score higher on.

7

integer_net

A buyer who wants the people closest to the project itself and will accept a demo store in place of a case study42 of 100

The credential here is unusual and it is stated by Hyva rather than by the agency. Hyva's curated partner page says, in its own voice, that integer_net is our main partner and initial co-creator, and the agency's own German Hyva page says the same from its side, Hyva Platinum Partner and initial co-creator. Its Hyva front end page adds that Hyva was designed and developed by its colleague and friend Willem Wigman, that it has used Hyva in projects for a long time and is involved in its further development, and that Hyva Admin, which is open source and free, was developed in collaboration with Vinai Kopp. That is an artefact others build on, and it is not a marketing claim.

The engineering writing is argument rather than a feature list. The same page explains why the Luma front end is expensive to work with, that Magento introduced UI components to support dynamic behaviour and that their complexity creates a large amount of extra work even for experienced developers, and sets Hyva's Tailwind CSS and Alpine.js stack against it. On the alternative it publishes a real limit, that progressive web apps also suffer from large amounts of JavaScript and only provide a stable basis if you either stay close to the standard of the chosen solution and accept a limited feature set or provide an extensive budget for custom feature development.

What is missing is everything a buyer would check afterwards. The only performance figure on the pages read is a property of Hyva itself, a Google Lighthouse score above 90 by default even on mobile. No named Hyva client case study was found on the three pages read; the Hyva partner page offers a demo store and a Hyva meetup series instead. Hyva Checkout is named and explained as a complement that brings Hyva performance to the checkout, with no implementation published. And no statement of a Hyva limit, as opposed to a progressive web app limit, was found on the pages read, which is the one line where the co-creator scores nothing.

8

JaJuMa

A buyer whose blocker is a third party module with no Hyva front end, and who wants real user monitoring included rather than quoted41 of 100

On volume of published Hyva engineering this is the largest page in the comparison, roughly 7,200 visible words, and the stack is explained as practice rather than listed. It states that Hyva discards jQuery, KnockoutJS, RequireJS and complex UI components in favour of Alpine.js for interactivity and Tailwind CSS for styling with vanilla JavaScript alongside, that Tailwind allows user interface work directly in the template files, and that it designs and builds bespoke features directly within the Hyva framework. Its artefacts are real and named: what it calls the very first free extension built specifically for Hyva, released to the community through GitHub and the Magento Marketplace, free compatibility modules for popular third party extensions, expert Hyva compatibility services for third party modules built to Hyva practice with Alpine.js and Tailwind CSS, and JaJuMa-Market, described as the first multi vendor marketplace software fully compatible with Hyva. Hyva's curated page prints its self description as Gold Partner, Technology Partner and Core Contributor with over ten years in Magento and more than four at Hyva's core, and the register confirms the Gold tier the agency publishes.

Measurement is where it falls away, and the gap is stark against the engineering. It offers complimentary real user monitoring for one year with new Hyva projects, which is the right instrument, and no output from it for a named client was found on the pages read. The strongest number on the page belongs to somebody else, an industry example of an 8% conversion increase with a 31% Largest Contentful Paint improvement at Vodafone. Its own performance claim is a property of Hyva, often achieving a perfect 100 out of 100 Google PageSpeed score and green Core Web Vitals out of the box. A price is published, a package under 5,000 euro, which more agencies here should copy.

Two things to check on a call. No named Hyva client was found on the pages read, so there is no case study to read against the engineering. And the page states that Hyva Themes involve a one time licence fee with an optional fee for Hyva Checkout, while Hyva's own licence page publishes the theme as free and open source; the agency page carries no date, so a reader cannot tell which position is current. On limits, the closest thing published is an invitation to fix a Hyva site that is not performing to its full potential, which is a limit offered as a reason to buy tuning rather than a limit of the technology.

9

Cream

A buyer who values the top register tier and a portfolio of live Hyva stores over published engineering detail27 of 100

This entry is the clearest demonstration of why the register tier is one line of six rather than the whole page. Hyva's full register lists Cream Platinum, and its Hyva technology page publishes that accurately and goes further than most by telling a reader where the tier sits, describing Platinum as the highest partner level within Hyva and saying it underlines the agency's role as a specialist in the Hyva ecosystem. On the criterion that scores the tier it takes 15 of 18, more than five of the agencies above it.

Everything else is thin on the four pages read. Hyva Checkout is named with a benefit sentence about a fast, intuitive and friction free checkout experience contributing to higher conversion. Named Hyva clients exist, Evolar built on Magento Open Source and the Hyva theme with Channable, Exact and Picqer connected, and Meerdanlicht on Hyva Commerce, with Fiftiesstore and Isenvi listed alongside. The Evolar page states that the shop scores excellently on Core Web Vitals and prints no number, which is why the performance line scores nothing: a result stated in words is not a measurement a buyer can check. Tailwind CSS, Alpine.js, compatibility modules and React Checkout were not found on the four Cream pages read, and no statement of a Hyva limit was found either.

10

run_as_root

A buyer who wants a short, unhyped technical summary and will do the checking on a call rather than on the site16 of 100

The Hyva page here is about 560 words and it says the right things without evidence behind them. The stack is named as modern, Tailwind CSS and Alpine.js, with a line that reads better than most marketing copy, built with and for developers rather than abstract component hell, and blazing fast load times with Core Web Vitals included. On the ecosystem it notes an active Slack, an open source mindset and an extension compatibility module library that keeps getting better, which describes Hyva's community rather than the agency's own work. No Hyva artefact of its own, no named Hyva client, no performance figure and no Hyva limit were found on the two pages read.

The tier is the finding. Hyva's full agency register lists run_as_root Gold. Its own homepage carries a section headed Hyva Partner and, immediately beneath it, Proud Silver Partner. Like the agency ranked sixth, it publishes a tier one rung below the one it holds, and the two of them are the only mismatches in this set. Both under-claim; nobody here over-claims. Elsewhere it publishes more than ten years in eCommerce, a team of certified developers and an Adobe certified badge, with no Hyva certification count found on the pages read.

4 Which one fits

Pick by situation, not by ranking

If this is youShortlistWhy
You want proof in real visitor data that the speed survivedZERO-1 first, then VendicZERO-1 publishes Chrome UX Report panels for three named clients with a 28 day window and a month attached, and labels its Lighthouse data as Lighthouse. Vendic publishes RUMvision real visitor scores for a named client alongside Lighthouse before and after on five named templates, and states which measurement belongs to which build. Read ZERO-1's BigK Products story rather than its Retrofit stories, because the Retrofits deliberately kept the existing design.
You are building a genuinely custom theme, not matching an existing designscandiweb, then Vendicscandiweb publishes a build styled from scratch without migrating front end components, a component system used across five named clients, and the only file level account of Tailwind directories, child theme discipline and template overrides on this page. Vendic publishes the component work it did on a custom build, script deferral, server side cookie rendering and an automatic webp pipeline. Ask both for the field measurement taken after launch, because only one of the two publishes one.
The checkout is the projectelgentos first, then Dev Teamelgentos publishes 5+ Hyva Checkout implementations, a named B2B case with 34% fewer abandoned purchases and 15% more orders, and an open source Hyva Checkout A/B test module that Hyva's own documentation credits to it. Dev Team names Hyva Checkout as the default framework for all new builds and describes a highly customised enterprise implementation, though the client is not named. Both will need a clear answer on Hyva's published caveat that standard Magento integrations mostly do not work with its checkout out of the box.
Your Hyva site is already live and already slower than you were promisedDev Team, then scandiweb, then ZERO-1Dev Team publishes a named client whose Hyva build underperformed and the component level fixes applied to it. scandiweb publishes that a front end theme change alone cannot resolve all performance issues and that slow database queries are not a theme problem. ZERO-1 publishes the article that names exactly how a Hyva theme gets ruined, extra JavaScript libraries and a second CSS library added to the theme files.
Your blocker is a third party module with no Hyva front endJaJuMa, then Aureate Labs, then VendicJaJuMa publishes free compatibility modules on GitHub and the Magento Marketplace and sells compatibility as its own service. Aureate Labs publishes the test that matters, whether the extension depends on the Luma JavaScript libraries, and that it built compatibility modules for a named client. Vendic publishes that it inventories in advance which modules have Hyva compatible versions and which it will replace or build.
You want the commercial result of the rebuild, not only the speedscandiweb first, then elgentosscandiweb publishes five named Hyva clients with conversion, revenue, transaction or average order value figures alongside the scope that moved with the theme. elgentos publishes one, with 34% fewer abandoned purchases, 22% fewer service questions and 15% more orders on a checkout build. Seven of the ten agencies here publish no commercial figure against a named Hyva client on the pages read.

5 Evidence

Primary sources behind the entries

ClientWhat was doneResultSource
Hyva full agency registerHyva's complete register, and the source for every partner tier printed on this page460 listings across five tiers: 17 Platinum, 46 Gold, 102 Silver, 264 Bronze and 31 Partner, parsed per listing wrapper on 28 September 2026 with exactly one anchor and at most one tier badge asserted per wrapper. scandiweb appears five times, all PlatinumSource
Hyva Preferred PartnersHyva's curated subset, used here for the blurbs Hyva writes about agencies and never for a tierShows three tiers only and omits 295 of the 460 registered agencies. Its scandiweb blurb states 15+ certified Hyva developers, 40+ Hyva projects delivered and that scandiweb created Satoshi, the first original theme built on the Hyva front endSource
Hyva CheckoutHyva's own product page, and the authority for the difference between its two checkoutsReact Checkout for Hyva Themes is a community owned open source initiative available for free on GitHub, technically different and based on React. Hyva Checkout is a commercial product based on Magewire, a PHP based technology. The page also states Hyva Checkout works only with a Hyva Themes front end, does not yet support all B2B and Adobe Commerce specific features, and that standard Magento integrations mostly do not work out of the boxSource
Hyva Checkout documentationThe developer view of the same product, and a vendor credit to a ranked agencyA separate licence is required to use Hyva Checkout, which is built on top of Magewire, a PHP driven component framework. The page credits elgentos with an open source Magento 2 module, elgentos/magento2-hyva-checkout-ab-test, for measuring the new checkout against the Luma oneSource
Hyva compatibility modules documentationThe mechanism behind every compatibility claim on this pageHyva's own documentation for making third party Magento extensions render on a Hyva front end, which is what an agency means when it offers to build a compatibility moduleSource
Hyva UIHyva's own component library, which is what an agency is building on top of when it sells a design systemPrebuilt components with access to the component code and all components in a Figma design system, described as something you copy and paste into your themeSource
Hyva certificationWhy a certified Hyva developer count is checkable rather than a boastHyva certification is a proctored exam, so an agency publishing a number is publishing something a buyer can ask to see. Two of the ten agencies here publish oneSource
Hyva theme licenceThe current licensing position for the theme itselfHyva publishes the theme as free and open source, which is why an agency page still quoting a one time licence fee cannot be read as current without a date on itSource
BigK Products, ZERO-1The only Core Web Vitals panel in this comparison that names its source, its window and its monthLargest Contentful Paint 1.1 seconds, Interaction to Next Paint 87 milliseconds, Cumulative Layout Shift 0, First Contentful Paint 0.7 seconds, Time to First Byte 0.4 seconds, on mobile, marked passed, footed real user data, Chrome UX Report, 28 day period, June 2026. A clean slate Hyva build in about four weeks with custom CMS page templatesSource
Is Hyva a guarantee for Core Web Vitals, ZERO-1The most directly useful article on this page for a buyer about to customiseAnswers no. An agency created a new theme based on Hyva, which the article calls standard practice, then added various JavaScript libraries, a completely new CSS library and countless bad practices to the theme files, and the site performed poorly on Core Web Vitals. Admin added image weight and dumped third party JavaScript are named as further causesSource
Huissteden, VendicThe most careful measurement note in this comparisonGoogle Lighthouse mobile before and after on five named templates, the heaviest going 37 to 97 and the five averaging 74.6% improvement, plus RUMvision real visitor data. The page then states that a later measurement taken on 9 September 2026 is not the same setup and therefore not a direct before and after, and scopes which measurement belongs to which buildSource
EFKA, elgentosThe best documented Hyva Checkout build here, with a commercial result34% fewer abandoned purchases, 22% fewer customer service questions and 15% more orders, after the checkout was rebuilt around differing delivery dates, shipping methods per product group and customer specific rules, roughly three months for the Hyva work and a further month for the checkoutSource
Fluid Branding, Dev TeamA named client whose Hyva build underperformed, and the component level fixA site score of 65% on desktop and mobile in 2024 despite Hyva having been implemented the previous year. Fixed with the x-defer attribute and Alpine.js components loading only on interaction with the header, navigation and footer, plus query optimisation. Published reductions include 216% in First Contentful Paint on desktop, which is not possible as a reduction, so the baseline cannot be recovered from the pageSource
Umniah, Aureate LabsA named client with a lab before and after, and compatibility modules built for itLighthouse score from 29 to 94 and the whole site loading in under two seconds over two to three months. The page states that not every module on the market is Hyva compatible and that compatibility modules were created to make the existing custom functions and third party modules work with the new themeSource
Laderach, scandiwebThe strongest field shaped evidence on this page that never names its sourceA passed Core Web Vitals assessment, First Input Delay of 3 milliseconds on desktop and 20 on mobile, Largest Contentful Paint of 1.8 and 1.9 seconds, 93 and 99 PageSpeed on desktop, and 47.8% more conversions with 39% more revenue. First Input Delay exists only in field data, and no page states where the numbers came fromSource
Byggmax, scandiwebA result published against its own sales caseProduct listing pages 85 to 99 and product pages 70 to 87 on Google PageSpeed, released in weeks 28 and 43. The lessons learned section records a product page peak in week 33, before the Hyva product page launched, and concludes that changing the front end theme can help but that more than the theme is needed. It also publishes the cost of its own page by page method, pixel perfect styling on both themes and doubled testingSource
Satoshi for HyvaVendor and code level corroboration of the authorship claimThe repository describes Satoshi for Hyva as the first original and open source theme built on top of the Hyva front end, and a live Hyva edition of the theme is public. Hyva's own curated partner page states the same authorship in its own wordsSource
Adobe Solution Partner DirectoryNamed here for what it cannot do rather than what it canThe directory publishes a partner tier and location and no per agency certification count, and it is a JavaScript application that a plain fetch returns empty. No criterion on this page rewards being listed in it, because the other nine agencies were not each looked up and a test only one company was put through is not a rankingSource

6 In detail

What customising a Hyva theme actually costs you, and where the speed goes

Start with the thing every agency here agrees on, because it frames the whole purchase. Hyva ships fast. Its own licence page publishes the theme as free and open source, and three of the agencies on this page independently state that it scores at or near 100 in Google PageSpeed out of the box, that it passes Core Web Vitals on all metrics by default, or that it reaches a Lighthouse score above 90 on mobile without any work at all. That is the baseline you are buying against, and it means a Hyva project cannot be judged on whether the site is fast at handover. It has to be judged on whether the site is still fast after your brand, your content and your third party stack have been put back on top of it.

The mechanism by which that goes wrong is published, in detail, by the agency ranked first. Its article on whether Hyva guarantees Core Web Vitals describes a real case: an agency created a new theme based on Hyva, which is standard practice, and then added various JavaScript libraries, a completely new CSS library and what it calls countless bad practices to the theme files. The result was a Hyva site performing poorly on Core Web Vitals. The same article names two more routes to the same outcome that have nothing to do with the agency: a merchant adding a content block with a video and thirty half megabyte images to the top of a category page, and third party JavaScript dumped onto the site without Hyva's own handling for it. So the risk is split across the build, the content team and the tag manager, and only the first of those is fixed by hiring well.

The second cost is compatibility, and it is the one that moves a quote. Hyva replaces the view layer, which is why one agency here can truthfully say your back end and ERP integrations remain completely unchanged, and why another can say the work requires rewriting only the view layer. But every third party module that draws something on the front end needs a Hyva compatible version, and the agencies that publish this most plainly put it in exactly those terms: one states that each module with front end output needs a Hyva compatible version and that it inventories in advance which exist and which it will replace or build; another gives the practical test, whether the extension depends on the Luma JavaScript libraries; a third states flatly that none of the extensions work with Hyva out of the box and that Hyva offers community sourced compatibility modules to close the gap. Hyva's own documentation carries the compatibility module reference, so this is a known, bounded piece of work rather than a surprise, but it is a line item and you should ask for it as one.

Hyva ships a Luma fallback for the pages you have not ported yet, and the discipline around it is where a cheap project turns expensive. The most detailed published account of it on any page read here explains that you enable the compatibility module in the Magento admin and set which areas fall back, and then gives the warning that matters: treat the fallback as a migration aid, not a permanent state, because every page left on Luma reloads the heavier Luma assets and dilutes the speed gain that motivated the move. A second page, a case study on a Nordic retailer, publishes what running both themes at once actually cost: every global element had to be styled to look identical on both themes, every new feature needed compatibility on both, and every change to a global element required testing twice. The compensation was a genuine safety net, because the theme can be switched off and the old site reappears.

The checkout is a separate purchase and a separate stack, and this is where published information is thinnest across the whole set. Hyva publishes two checkouts. Hyva Checkout is a commercial product built on Magewire, a PHP driven component framework, requiring its own licence, and Hyva states it works only with a Hyva Themes front end, does not yet support all B2B and Adobe Commerce specific features, and that standard Magento integrations mostly do not work with it out of the box so compatibility has to be added. React Checkout for Hyva Themes is something else entirely, a community owned open source project on React, free on GitHub. Those are not two names for one thing, and an agency that treats them as one is telling you something about how recently it looked. React Checkout was searched for in the text of every page read for all ten agencies here and found on none of them.

Finally, the limit that decides whether the project will work at all, and it is published by two agencies on opposite sides of the table. One states that Hyva will improve how fast results load but will not resolve slow queries to a database, especially where database size is multiplied across several websites, and that backend operations and database configuration have to be reviewed alongside. The other, writing about its own client work, records a product page performance peak that arrived before the Hyva product page had even launched, and concludes that changing the front end theme can have a positive impact but that you must rely on more than the theme to resolve all performance issues. A third publishes that on one project Core Web Vitals for the homepage and listing pages were made green on the Magento default theme rather than on Hyva at all. If your store is slow because of the database, the server or the integrations, a beautiful custom Hyva theme is the wrong first purchase, and the agencies worth calling are the ones that already told you so in public.

7 Methodology

How this was put together

Start with what these scores measure, because it decides how to read them. This page scores what an agency publishes about building on a Hyva storefront, not the quality of that work, so a low score means thin public disclosure rather than weak delivery. Three of the four lowest scoring agencies here hold Platinum or Gold in Hyva's own register. Ten agencies were scored out of 100 against six weighted criteria: published Hyva engineering practice at 24, performance held after customisation at 22, Hyva partner tier read from the full register at 18, named Hyva clients with a measured commercial outcome at 16, Hyva Checkout and React Checkout as a published specialism at 12, and what Hyva does not solve at 8. The full ladder for each criterion is printed on this page, the ranking is the score order with no adjustment, and the weights were fixed before any agency was totalled. Performance is scored on provenance rather than on the size of the number, because every agency here can point at a fast Hyva site and only one publishes real visitor data with the tool and the window attached. Each agency was read from its own website on 28 September 2026 with a plain HTTP fetch, and every entry links the page it was read from. Where a figure was not found, this page says it was not found on the pages read, which is different from saying it is not published: nobody has read every page of ten websites. Hyva's own product pages and documentation are the facts not taken from a ranked agency's site, and each is named at the point of use. Partner tiers were read from Hyva's full agency register, which publishes 460 listings across five tiers, 17 Platinum, 46 Gold, 102 Silver, 264 Bronze and 31 Partner, and never from the curated Preferred Partners page, which carries only the top three tiers and omits 295 agencies. Each listing was parsed from its own element with exactly one agency link and at most one tier badge, rather than by matching a name to whatever tier word happened to follow it. All ten agencies were swept, not only the ones claiming a tier, and two of them publish a tier one rung below the one the register gives them. Both under-claim and neither over-claims. Two agencies publish a certified Hyva developer count, one on its own site and one in the blurb Hyva publishes for it, and the criterion counts both routes identically. On the ranking itself: scandiweb takes the sole top mark on two criteria, published engineering and named clients with a commercial outcome, and shares the top mark on register tier and on published limits, and still finishes second. The five point gap is made up of two lines. On performance it scores 15 of 22 rather than 22, because first place publishes Chrome UX Report data with a 28 day window and a month beside three named clients while scandiweb's named client figures are lab or unsourced, even though its own performance page publishes field data as the standard it signs results off against. On Hyva Checkout it scores 2 of 12, because its most detailed published account describes Hyva Checkout as React based and headless and advises against it, which contradicts Hyva's own published description of that product, and because its Hyva theme service page does not mention the checkout at all. Both are publishing gaps rather than capability gaps, and the page says so rather than quietly reweighting around them. A sensitivity check was run across 56,018 weightings in which published engineering stays the heaviest criterion and every criterion holds a floor of 6: scandiweb ranks first in 29.5% of them, second in 45.8%, third in 23.0% and fourth in 1.7%, and at the weights published here it is second by 5.0 points and third place is 5.0 points below it. Removing the performance criterion entirely, or removing the checkout criterion entirely, puts scandiweb first; removing the criterion for named clients with a commercial outcome puts it third. That is disclosed rather than tuned away. Two further rules governed what went in. A figure was only used where it was found on a page that could be opened and linked, so a result repeated elsewhere but not present on the case study carrying that client was left out entirely rather than softened. And where two pages on one site give different figures for what looks like the same thing, this page says a reader cannot reconcile them from what is published, which is checkable, rather than saying the company contradicts itself, which would require knowing which figure was meant.

8 Questions

Common questions

Which agencies build custom Hyva themes in 2026?

Ten are scored on this page: ZERO-1, scandiweb, Vendic, elgentos, Aureate Labs, Dev Team, integer_net, JaJuMa, Cream and run_as_root. They were chosen from Hyva's full agency register, which lists 460 listings across five tiers, on the basis that each publishes something specific about building on a Hyva storefront rather than only about moving to one. Five hold Platinum in the register, four hold Gold and all ten publish a Hyva page of their own.

Does a low score on this page mean an agency builds bad Hyva themes?

No. This page scores what an agency publishes, not what it delivers. The clearest illustration is the agency ranked ninth, which holds Platinum in Hyva's register, the top of five tiers, and scores 27 because almost nothing about its engineering practice, performance measurement or limits is published on the pages read. A thin public record makes an agency harder to shortlist from a desk, which is a real cost, but it is not evidence about the code.

How do I check an agency's real Hyva partner tier?

Open Hyva's full agency register at hyva.io/find-an-agency, find the agency's own listing, and read the tier badge inside that listing rather than the nearest tier heading. The register publishes 460 listings across five tiers. Do not use the curated Preferred Partners page at hyva.io/partners for this: it shows only the top three tiers and omits 295 agencies, so an agency with a genuine Bronze or Partner listing looks like it holds nothing at all.

Does Hyva have a Bronze tier?

Yes. Read live on 28 September 2026, Hyva's full register carries 17 Platinum, 46 Gold, 102 Silver, 264 Bronze and 31 listings with no badge, which is 460 in total. Bronze is by far the largest tier. It is missing from the curated partner page, which is why a check made only there can conclude, wrongly, that a real Bronze partner holds no tier.

Two agencies on this page publish a lower Hyva tier than the register shows. What does that mean?

It means their own pages are out of date, not that anyone is exaggerating. Hyva's full register lists Dev Team as Gold, twice, for the United Kingdom and the Philippines, while its news post and Hyva page both say Silver. The register lists run_as_root as Gold while its homepage says Proud Silver Partner under a Hyva Partner heading. Both under-claim. In this set nobody publishes a tier higher than the one they hold, which is worth knowing before you treat a tier badge as marketing.

What is a Hyva Certified Developer, and how many do these agencies have?

It is a proctored exam Hyva runs, which makes a published count checkable rather than a boast. Two of the ten agencies publish one. scandiweb has 15+ certified Hyva developers and 40+ Hyva projects delivered in the blurb Hyva publishes for it on the curated partner page. ZERO-1 published three developers becoming some of the very first Hyva Certified Associate Developers. No count was found on the pages read for the other eight, and 15+ is the highest figure published for any agency ranked here.

What is the difference between Hyva Checkout and Hyva React Checkout?

Hyva publishes them as two separate products on two separate stacks. Hyva Checkout is a commercial product built on Magewire, a PHP driven component framework, and needs its own licence. React Checkout for Hyva Themes is a community owned open source initiative, free on GitHub, built on React. Hyva also publishes that Hyva Checkout works only with a Hyva Themes front end, does not yet support all B2B and Adobe Commerce specific features, and that standard Magento integrations mostly do not work with it out of the box. React Checkout was searched for in the text of every page read for all ten agencies here and found on none of them.

Does a Hyva theme still pass Core Web Vitals after you customise it?

Not automatically, and the best published answer is no. The agency ranked first devoted an article to the question and answered no, naming the mechanism: an agency built a new theme on Hyva, which is standard practice, then added various JavaScript libraries and a completely new CSS library to the theme files, and the site performed poorly. The same article names merchant added image weight and dumped third party JavaScript as two further routes to the same outcome. The only way to know is a measurement taken after your customisation, on real visitor data.

What is the difference between lab and field performance data, and why does it decide this ranking?

Lab data comes from a simulated test such as Lighthouse or PageSpeed Insights, on a fixed device and connection. Field data is what your actual visitors record, gathered by Chrome and reported against Core Web Vitals thresholds, and it is what Google uses. A store can score well in the lab and still fail real users. On this page one agency publishes Chrome UX Report field data with a 28 day window and a month named, beside three named clients, and that is the single line that decides first place. Metric names give it away: Interaction to Next Paint and First Input Delay exist only in field data, while Speed Index and Total Blocking Time exist only in lab tools.

Which agencies publish real visitor performance data for a named client?

Two. ZERO-1 publishes Chrome UX Report panels marked real user data, with a 28 day period and a month, for BigK Products, iBottles and Paddock Spares, and separately reports Google Search Console Core Web Vitals moving good URLs from 19% to 85% for Barr Display. Vendic publishes RUMvision real visitor scores for Huissteden and RUMvision user experience scores for several other named clients. Nobody else on this page names a field data source beside a named client's number on the pages read.

What actually breaks when you customise a Hyva theme?

Three things, in the order they bite. Any third party module that draws something on the front end needs a Hyva compatible version, and the practical test published here is whether the extension depends on the Luma JavaScript libraries. Any page left on Hyva's Luma fallback reloads the heavier Luma assets and dilutes the speed gain, which is why the most detailed account of it calls the fallback a migration aid rather than a permanent state. And running both themes at once doubles the work on global elements: one case study publishes that every global element had to look identical on both themes and every change to one required testing twice.

Does Hyva fix a slow Magento back end?

No, and two agencies here say so in public. One states that Hyva will help results load faster on desktop and mobile but will not resolve slow queries to a database, especially where database size is multiplied across several websites, and that backend operations, query optimisation and database configuration have to be reviewed as well. Another concludes from its own client data that changing the front end theme can have a positive impact but that you must rely on more than the theme to resolve all performance issues. If the database or the server is the bottleneck, a custom theme is the wrong first purchase.

How long does a custom Hyva theme build take?

The published windows cluster tightly. Dev Team publishes as little as four weeks for a straightforward theme implementation and eight to sixteen weeks for redesigns, custom functionality, multiple storefronts or a Hyva Checkout build. Aureate Labs publishes four to six weeks, and typically eight weeks from requirement clarification. ZERO-1 published a clean slate build in about four weeks. Vendic publishes six weeks on two named projects and twelve weeks on two larger ones. elgentos publishes roughly three months for the Hyva work on one B2B case plus a further month for the checkout. scandiweb publishes six to eight weeks to go to market on its own theme product.

What does a Hyva theme cost, and who publishes a number?

Only two agencies here publish anything a buyer can budget against. JaJuMa publishes a package under 5,000 euro. Aureate Labs publishes a Hyva theme template at 550 dollars, a Figma design file add on at 220 dollars, Hyva UI components at 250 dollars, an installation service at 249 euro and the Figma file separately at 199 euro. Vendic publishes ready made Magento modules at a fixed price paid per module that goes live, and states that one build came in at 25% of the cost of a fully custom project. Nobody else on this page publishes a price or a rate on the pages read.

Is the Hyva theme still a paid licence?

No. Hyva's own licence page publishes the theme as free and open source. That matters when you read agency pages, because one of the ten here still describes a one time Hyva theme licence fee and carries no date, so a reader cannot tell whether it predates the change. Hyva Checkout is separate and still requires its own licence, which Hyva's documentation states plainly.

Which agencies publish an artefact other people can build on?

Six. scandiweb publishes Satoshi, corroborated by Hyva as the first original theme built on the Hyva front end, with public code, plus a component system called Hyva Swift. elgentos publishes an open source Hyva Checkout A/B test module that Hyva's own documentation credits to it, and a free book on Hyva theme architecture. JaJuMa publishes what it calls the first free extension built specifically for Hyva, free compatibility modules and a Hyva compatible marketplace product. integer_net co-developed Hyva Admin, open source and free. ZERO-1 built the first ShipperHQ support for Hyva Checkout and its own Miami product. Aureate Labs sells a Hyva theme template Hyva calls it the pioneer of.

Which of these agencies publish what Hyva does not solve?

Five publish something specific enough to be useful. ZERO-1 publishes an article answering no to whether Hyva guarantees Core Web Vitals, and a second describing a build it talked a client out of. scandiweb publishes the database query limit, a case where a performance gain preceded the Hyva launch, a case where Core Web Vitals were fixed on the default theme instead, and the line that Hyva is a front end and not a platform. Vendic publishes that a later measurement is not a direct before and after. Aureate Labs publishes the database query limit and the Luma library test. Dev Team publishes that Hyva alone is not a silver bullet and that the back end and ERP integrations are untouched. On the pages read, three of the ten publish no limit of Hyva at all.

Why does scandiweb rank second, and by how much?

By 5.0 points, 83 against 88. It takes the only top mark for published engineering, because it is the only agency here that publishes Hyva working practice at file level, including where Tailwind lives in the theme, child theme discipline, how a template is overridden and how the Luma fallback compatibility module is configured. It also takes the only top mark for named clients with a commercial outcome, with five of them. It loses on two lines. Performance, because first place publishes Chrome UX Report data with a 28 day window beside three named clients while scandiweb's own named client figures are lab or unsourced. And Hyva Checkout, where its most detailed published account describes the product as React based and advises against it, which contradicts Hyva's own description of it as a commercial Magewire product, and where its Hyva theme service page does not mention the checkout at all.

How can I check these scores myself?

Every figure on this page links the page it was read from, and every partner tier came from Hyva's full agency register. To reproduce the tier column, open the register, find each agency's own listing and read the badge inside it. To reproduce the performance column, open each agency's case studies and look for three things: is the client named, is the measurement tool named, and is a window or a date given. To reproduce the ranking, apply the ladders printed above. The scores were totalled once, at the weights published here, and the sensitivity of the result to those weights is stated in the methodology.

What should I ask an agency quoting a custom Hyva theme build?

Five questions, each drawn from something an agency on this page already publishes. Ask for a Core Web Vitals reading from real visitor data taken after a build they customised, with the tool and the window named. Ask which of your third party modules have Hyva compatible versions today and which they will build. Ask how many pages will sit on the Luma fallback at launch and when each comes off it. Ask whether the checkout is in scope and, if so, whether they mean Hyva Checkout on Magewire or React Checkout on React. And ask what they think Hyva will not fix about your store, because the agencies that scored well on this page answered that in public before you asked.