Liquid is the right default
Headless is a real architecture and a real fit for some brands. It is also a tax. The tax shows up as developer time, infrastructure cost, and operational drag. For most Shopify Plus brands doing 5 to 50 million in GMV, Liquid wins on every axis that matters.
We have built both. The bar we apply is simple. If the brand needs storefronts in multiple frameworks, or needs to plug a Shopify backend into an existing CMS, headless is the path. Otherwise, Liquid is faster to build, faster to maintain, and faster to load.
Dawn is the floor and not the ceiling
Dawn is a great starting point and a terrible ending point. The sections architecture is right. The performance defaults are reasonable. The visual language is too generic for a brand that wants to look like itself.
Our six week build starts from Dawn and rebuilds the sections that need brand presence. Cart drawer, PDP gallery, PLP filtering, homepage hero, footer. The rest stays close to Dawn because the rest does not differentiate the brand.
Performance budgets have to be written down
LCP under 2.0 seconds on a mid-tier mobile device on 4G. CLS under 0.05. INP under 200ms. These are the numbers we hold every theme we ship to. They are also the numbers most themes in the App Store fail.
The wins come from boring places. Self-host the font. Compress every image. Lazy-load anything below the fold. Skip the marketing-pixel bundle until interaction. We do not optimize for a Lighthouse score, we optimize for the LCP on the slowest device in the analytics report.
Sections we never compromise on
The PDP gallery, the cart drawer, and the checkout flow get the most engineering time on every project. The PDP gallery because it is the most-viewed asset on the site. The cart drawer because it is the most measurable revenue surface. The checkout because Plus customers get a checkout extensibility budget and we use it.
Everything else gets a sensible default and a sections editor. Marketing teams need to ship landing pages without a developer. The job of the theme is to make that easy without letting them break the brand.
The six week cadence
Week one: design system and tokens. Week two: foundation, homepage, footer. Week three: PLP and search. Week four: PDP and gallery. Week five: cart, checkout, and account. Week six: QA, migration, and ship.
Most theme builds slip because the design system is not locked at the end of week one. We treat tokens as the contract for the rest of the build. Colors, type, spacing, and motion are decided once and never re-litigated. That is the discipline that keeps six weeks from becoming twelve.