Multi-Brand Digital QA Ecosystem

End-to-end QA strategy for a major restaurant group’s web, mobile, and loyalty platforms across multiple brands.

  • Web
  • Mobile
  • CI/CD
Production release cadence achieved
Every sprint
Full regression cycle time
< 1 day
Functional, Visual, Accessibility, Performance and Security
5 layers

Context

Early in my career, I joined a cross-functional team responsible for expanding the digital presence of a major restaurant group. The scope was significant: improve partially built native mobile apps, build a new website from scratch, including a CMS (Sitecore) and an integrated ordering experience, and eventually absorb additional brands under the same umbrella, including their loyalty platform.

The project demanded a QA strategy that could scale across multiple brands, multiple platforms, and a growing team.

What was built

The automation coverage spanned five dimensions, all running automatically on every push to the QA environment:

Functional testing via a Cypress-based web suite with Page Object architecture built for scale and multi-brand reuse. Visual regression via Percy to catch layout and component breakage on a CMS-heavy frontend. Accessibility checks via Axe-core covering WCAG compliance across page templates. Performance load testing via k6 targeting key user flows in staging. Security scanning via OWASP ZAP integrated directly into the CI/CD pipeline.

Tools used:

  • Cypress
  • Percy
  • Axe-core
  • k6
  • OWASP ZAP
  • Maestro
  • Robot Framework

Key decisions

  • Cypress over Java + Selenium

    I led two structured POCs and ruled out Java + Selenium despite it being the familiar choice. The onboarding complexity didn’t fit a team that included interns, and the verbosity would slow delivery. With multi-brand expansion already anticipated, I chose Cypress and designed the Page Object architecture with that scale explicitly in mind.

  • Percy over BackstopJS

    I identified that functional assertions alone wouldn’t catch visual regressions on a CMS-driven, image-heavy frontend. After evaluating both tools, Percy won on CI/CD integration quality and baseline management reliability.

  • k6 over JMeter

    Performance testing wasn’t on the original roadmap. I identified the gap, made the case for its inclusion, and chose k6 over JMeter for its code-first approach and native fit with our pipeline. JMeter’s GUI-oriented model would have created unnecessary friction.

  • Deferring Robot Framework rewrite — then migrating at the right time

    I spotted structural problems in the existing mobile suite early but assessed that an immediate rewrite would be more disruptive than valuable. I made the deliberate call to hold, and later led the architectural migration to Maestro when the project maturity and team capacity justified it.

  • OWASP ZAP as a pipeline gate

    I integrated ZAP directly into CI/CD so security scans ran automatically on every build, treating security as a continuous delivery concern rather than a last-minute check.

Outcomes

Production releases: no regression backlogs blocking delivery
Every sprint
Full regression cycle, replacing what used to take several days manually
< 1 day
Architecture scaled to additional brands with minimal rework
Multi-brand
Functional, visual, accessibility, performance and security, all automated
5 layers