Small mobile
Check around 360–390px, long titles, sale prices, sticky controls and the mobile keyboard.
Combine automated checks, engineering review and a real storefront preview before publishing.
A pre-deployment checklist for Liquid, HTML, schemas, assets, translations, performance and review.
Theme QA is not a single pass/fail test. It combines automated validation, storefront review and business-context checks. Use the short list for routine changes and the full checklist when a change affects product purchase behavior, navigation, tracking or several templates.
Responsive defects often appear only when content is unusually long, images are missing or browser controls reduce the available viewport. Test representative breakpoints and deliberately difficult content.
A visual review should include interaction states, focus states and overlays, not only the first static screen.
Check around 360–390px, long titles, sale prices, sticky controls and the mobile keyboard.
Check intermediate layouts where desktop and mobile rules can overlap unexpectedly.
Check wide and laptop-sized screens, not only a large design monitor.
Check missing media, long translations, many variants, long navigation labels and large account names.
| Evidence | Why it matters |
|---|---|
| Request and interpreted scope | Confirms the implementation solves the intended problem |
| Files changed | Makes the exact technical surface clear |
| Preview URL | Lets the merchant inspect the real staging state |
| Checks passed and warnings | Separates verified facts from remaining uncertainty |
| Approver and approval time | Creates attributable control over publishing |
| Deployment and verification record | Shows what reached production and whether it behaved as expected |
| Rollback conditions | Prevents a later reverse operation from overwriting newer work blindly |
Automation should remove repetitive validation and make evidence consistent. It should not produce false certainty about subjective design, business logic or integrations it cannot fully observe.
The best approval packet combines automated results with a short list of manual scenarios selected from the actual scope.
| Automated evidence can verify | Human review is still needed for |
|---|---|
| Liquid and JSON validity | Whether the requested experience is commercially correct |
| Missing files and prohibited targets | Visual hierarchy, brand quality and content accuracy |
| Known dangerous patterns and new console errors | Unusual app, market and customer-state behavior |
| The exact files included in the changeset | Whether remaining warnings are acceptable |
| That a staging write and preview were produced | Whether the merchant is ready to authorize production |
Testing every URL is rarely practical. Choose representative states that exercise the changed behavior and its dependencies. A product-page change might need one simple product, one multi-variant product, one sold-out product and one subscription product rather than fifty similar products.
Add markets, customer accounts or device states only when the implementation can affect them. The goal is broad risk coverage with a well-defined review plan.
| Dimension | Representative states |
|---|---|
| Product | Simple, variants, sold out, subscription or bundle where relevant |
| Viewport | Small mobile, tablet transition, laptop and wide desktop |
| Content | Long title, sale price, missing image, long translation |
| Customer | Guest, logged in, market or currency variation when affected |
| Interaction | Initial load, variant change, add to cart, section reload and back navigation |
The guide explains acceptance criteria, Liquid and JSON structure, HTML, schema, assets, translations, performance and interactions so the reader can distinguish a reliable workflow from a generic code suggestion.
Review acceptance criteria, Liquid and JSON structure, HTML, schema, assets, translations, performance and interactions, identify the affected templates and document dependencies or assumptions.
The implementation should be small enough to review, explicit enough to test and isolated from unrelated storefront work.
Automated checks catch defined failure classes; mobile, desktop, content extremes and interactions still require inspection.
Production should change only after the staged result and remaining risks are understandable.
The guide preserves uncertainty and routes complex work to experienced engineers instead of offering a false guarantee.
Capabilities change over time. These links point to the alternative provider’s own current documentation and should be reviewed during future updates.
Connect your store to compare the guidance with findings from your own theme.