Quality AssuranceProcess
Building a Pre-Launch QA Process That Scales
How to turn ad-hoc launch-day panic into a repeatable QA process: environments, a tiered checklist, automation, and a go/no-go gate.
Tiered Checklist
Real Environments
Catch Regressions
Go / No-Go Gate
Every store has a QA process. For most it is called "the founder clicks around for ten minutes and hopes." That works until it does not - usually on the highest-traffic day of the year. A repeatable QA process is boring on purpose, and it scales with your release cadence.
1Test Somewhere That Looks Like Production
Bugs hide in the gap between your test setup and the real store. Close that gap first.
- Use a duplicate theme or a dedicated staging store, not the live theme's edit mode.
- Seed realistic data: real product counts, real collection sizes, a discount code, an out-of-stock variant.
- Test with production-like apps installed - app conflicts are a top launch-day failure.
Keep a throwaway order path
A $0.01 hidden product plus a 100%-off code lets you complete real checkouts without touching payouts.
2A Checklist With Tiers, Not 200 Flat Items
Nobody runs a 200-line checklist honestly. Split it by risk so the right depth matches the change.
Smoke (every deploy)
Home, collection, product, cart, checkout complete. 5 minutes.
Core (weekly release)
Search, filters, account, discounts, email receipts, key landing pages.
Cross-device (theme change)
iOS Safari, Android Chrome, one small viewport, one desktop.
Full regression (big launch)
Everything, plus load test and analytics/pixel verification.
3Running the Pass
- 1
Pick the tier
Match it to the size of the change, decided by the person merging.
- 2
Run it on staging
Log every failure with a screenshot, URL, device, and steps to reproduce.
- 3
Triage
Blocker, major, or minor. Blockers stop the release; minors get tickets.
- 4
Re-test fixes only
Verify each fix and one thing around it - not the whole checklist again.
PRO TIP
A good bug report has all of these - missing one wastes a round-trip:
- What you did (steps)
- What you expected
- What happened
- Device / browser / URL
- Screenshot or recording
4Automate the Boring 20%
You will never automate everything, and you should not try. Automate the checks that are identical every time and expensive to miss.
| Check | Automate? | Why |
|---|---|---|
| Checkout completes | Yes | Highest cost of failure, stable steps |
| Add to cart from PDP | Yes | Runs on every deploy, fast |
| Visual layout of a campaign page | No | Changes constantly, human eye is faster |
| Broken-link scan | Yes | Cheap, catches content mistakes |
| Analytics events fire | Yes | Silent when broken, big blind spot |
5The Go / No-Go Gate
One named person makes the call, against written criteria, and records it. That is the whole ritual - it just has to happen every time.
Ship criteria
Zero open blockers, smoke suite green on staging and production, rollback plan written, owner on call for two hours after launch.
FAQ
Scale it down: just the smoke tier and a one-line go/no-go. The structure matters more than the size.
The checkout-complete path. It has the highest cost of failure and the most stable steps.
Review it after every incident. If a bug reached production, add or change one line so it cannot happen silently again.
Want a QA process built around your releases?
We set up practical QA and test automation for Shopify teams.
Book a Free Consultation
