Case Study · E-Commerce
API and UI test automation for Fabletics and Savage X in nine regions, with an engineer reviewing every AI-written test.
The Result
What Changed for Fabletics
Overview
Fabletics sells activewear online across North America and Europe, and Savage X, a lingerie brand, runs on the same platform. SprintSynergy has worked on e-commerce test automation for Fabletics for more than four years, and today covers both brands. We write and maintain 400+ API tests in PactumJS and 650+ UI tests for the internal tools behind the stores. Smoke and regression pipelines in Jenkins and GitHub Actions run them. To move the UI tests from Robot Framework to Playwright, we built an AI agent that converts them. It moved 450+ tests, and an engineer reviewed each one before it joined the suite.
*** Test Cases ***
Agent Finds Member By Email
Open Service Console ${BRAND} ${REGION}
Search Member ${MEMBER_EMAIL}
Member Profile Should Be Visibletest('agent finds member by email', async ({ serviceConsole }) => {
await serviceConsole.open(env.BRAND, env.REGION);
await serviceConsole.searchMember(env.MEMBER_EMAIL);
await expect(serviceConsole.memberProfile).toBeVisible();
});The Problem
Two brands share one platform. Every change has to work for each brand, in each region and in more than one environment.
Much of the risk sits behind the shopping pages. Support agents look up members and act for them. Merchandisers create products and pick their images. Marketing teams change what the stores show. A bug in any of these internal tools reaches shoppers without touching the storefront code.
The UI tests were written in Robot Framework. We proposed moving them to Playwright, mainly because it works better with AI agents, and Fabletics approved. That left 450+ tests to move without rewriting each one by hand.
What We Tested
The API tests cover member accounts, membership, products, cart, search, promotions and marketing content.
Environment variables choose the brand, region and environment for each run, so one suite can target either brand and each region.
How We Built It
What We Found
Three bugs our regression runs in Jenkins caught before release.
Support agents could not cancel an order with the status Placed, although Placed orders can be cancelled once a reason is given. Had it gone live, agents could not have cancelled eligible orders for customers.
Finalizing a new return failed with an error saying the customer ID was missing, although the order already had one. Had it gone live, agents could not have completed returns for the affected orders, and customers would have waited longer.
After a product was edited, the publish window stayed stuck on screen and blocked any further action on that product. Had it gone live, product updates would have stalled.
In Place Now
Questions
Yes. We built an AI agent with GitHub Copilot and Claude that converts a Robot Framework test into a Playwright test in TypeScript. An engineer reviews every converted test in a pull request before it is merged. More than 450 tests moved this way.
Book a free 30-minute QA strategy call. A written QA plan follows within 48 hours.
Book a Free QA Strategy Call