Twin Health scales automation coverage to 60% in 3 months with Low Code Automation

Twin Health is a digital health company delivering AI-driven whole body health programmes to members across the globe. Its platform combines continuous data monitoring with personalised clinical guidance, making the reliability and consistency of its mobile application critical to member outcomes and engagement. With a fast-moving product and a QA team responsible for validating high-traffic, safety-sensitive user flows across Android and iOS, the pressure to scale automation without scaling headcount was significant.

Dhirendra Kumar Jha, QA Manager at Twin Health, leads quality engineering across this platform and was responsible for driving the shift from a code-heavy, engineering-dependent testing approach to one that the whole team could contribute to. Manual testing and hand-coded Appium and Selenium scripts had taken the team as far as they could go. By adopting BrowserStack Low Code Automation, Twin Health grew automation coverage from a limited baseline to close to 60% within three months, cut test authoring time by 75%, and reduced debugging effort by 60%, all without adding infrastructure or restricting automation to engineers alone.

Before BrowserStack Low Code Automation, Twin Health’s testing relied on a combination of manual regression and code-based automation written in Python using Appium and Selenium. While functional, the approach created compounding bottlenecks that became harder to absorb as the product and team grew.

1. Every UI change broke the automation. Twin Health’s application evolves quickly. Each time the UI changed, existing automation scripts broke, and engineers had to spend time hunting down and fixing locators before they could write any new tests. The team was effectively running to stand still, maintaining old coverage rather than expanding it.

2. Test authoring was slow and engineering-dependent. Writing a single end-to-end test required finding locators, writing code, handling synchronisation issues, adding validations, and then debugging. Each test case represented a significant time investment, and only automation engineers could contribute, leaving manual testers unable to add to coverage regardless of their expertise.

3. Automation coverage was limited to the highest-risk flows. Because building automation was expensive in time and effort, the team could only justify automating the most critical user paths. Medium-priority scenarios and edge cases remained manual, leaving meaningful coverage gaps that represented a genuine quality risk.

4. Manual testers were stuck on repetitive regression. With low automation coverage, manual testers spent the majority of their time running repetitive regression cycles rather than doing exploratory testing, which is where their skills add the most value. This was both an inefficient use of capacity and a source of frustration for the team.

5. Defects found late in the cycle pushed releases back. The reliance on manual regression meant that defects were sometimes discovered late in the release process, after significant testing effort had already been completed. Late-stage defects forced rework, delayed releases, and eroded confidence in the team’s ability to sign off on time.

After evaluating several low-code testing tools, Twin Health selected BrowserStack Low Code Automation. Three requirements drove the decision: it had to be accessible to manual testers, not just engineers; it had to offer real AI capabilities rather than basic record-and-playback; and it had to integrate natively with BrowserStack’s device cloud, which the team was already using. LCA met all three.

1. AI-assisted test creation that eliminated manual scripting. The Low-Code Authoring Agent allowed the team to generate test steps automatically from recorded interactions, rather than writing code from scratch. What previously took hours of scripting, debugging, and locator management was reduced to a fraction of the time, immediately freeing engineers to focus on new coverage rather than maintenance.

2. A test recorder that opened automation to the whole team. The visual test recorder made building tests fast and intuitive enough that manual testers could contribute directly to automation for the first time. Rather than being restricted to running manual regression, manual testers became active contributors to the automation suite, expanding capacity without adding headcount.

3. Self-healing tests that survived UI changes. One of the most persistent pain points before LCA was the time lost to fixing broken locators every time the UI was updated. BrowserStack LCA’s AI-powered self-healing capability automatically detected when a UI element had changed and updated the test accordingly, keeping the suite stable through routine product changes without requiring engineer intervention.

4. Reusable modules and variables to eliminate duplicated logic. Twin Health’s test suite covers flows that share common steps across multiple test cases. Reusable modules and parameterisation in LCA meant that shared logic was defined once and reused across tests, rather than being duplicated and maintained separately in each script. This significantly reduced the overhead of keeping the suite consistent as the product evolved.

5. Real device execution baked into the same platform. Because LCA sits natively within the BrowserStack platform, tests built in LCA run directly on real Android and iOS devices through App Automate, with no additional integration work. The team had no new vendor to onboard, no new infrastructure to configure, and no new integration to maintain.

6. Centralised reporting with screenshots, logs, and recordings out of the box. Every test run in LCA produced detailed execution reports including screenshots, device logs, and video recordings, all accessible from a single dashboard with Test Reporting & Analytics product. Debugging a failure no longer meant digging through raw logs; the evidence needed to identify and act on an issue was already there.

7. Chosen over competing tools for its AI depth and native BrowserStack integration. Other tools evaluated offered record-and-playback but were web-only or lacked genuine AI capabilities behind the interface. Since Twin Health was already deep into BrowserStack for device testing, LCA was a natural extension rather than a new vendor relationship, removing procurement, integration, and onboarding overhead entirely.

Adopting BrowserStack Low Code Automation changed both what Twin Health’s QA team could achieve and how they spent their time. The results were measurable across coverage, speed, debugging efficiency, and team confidence.

1. Close to 60% automation coverage reached within 3 months. Starting from a limited automation baseline, Twin Health grew coverage to close to 60% within three months of adopting LCA. The speed of that growth was made possible by both the efficiency of AI-assisted test creation and the ability of manual testers to contribute to the automation suite directly for the first time.

2. 75% reduction in test authoring time. Building a new test, which previously required writing code, handling synchronisation, managing locators, and debugging, now takes roughly 75% less time using LCA’s recorder and AI-assisted creation tools. Engineers who previously spent hours on a single test case can now build and validate tests in a fraction of that time.

3. 60% less time spent on debugging and failure triage. Because BrowserStack LCA produces screenshots, device logs, and video recordings automatically for every test run, the time spent investigating failures dropped by 60%. Engineers no longer needed to dig through raw logs or attempt to manually reproduce issues; the evidence was already available in the execution report.

4. Manual testers freed for exploratory testing. With repetitive regression now handled by automation, manual testers were able to redirect their time toward exploratory testing, which is where they add the most value. This was both a productivity improvement and a quality improvement, as exploratory testing surfaces the kinds of edge-case issues that scripted regression typically misses.

5. Critical user flows automated reliably and kept stable. Twin Health’s member registration flow, described as a critical, high-traffic path for the business, was automated quickly using LCA and has remained stable through subsequent UI changes thanks to the self-healing capability. This represents exactly the kind of high-value, high-risk coverage that was previously too expensive to build and maintain with hand-coded scripts.

6. Leadership visibility into automation progress and release readiness. The dashboards and execution reports produced by LCA gave leadership a clear, real-time view of automation coverage and whether the team was release-ready, without requiring a manual status update from the QA team. This transparency extended confidence in the testing process beyond the engineering team and made release decisions easier and faster.

7. Automation no longer restricted to engineers. Perhaps the most significant cultural shift was that automation became a shared responsibility rather than an engineering-only discipline. With manual testers now able to build and contribute tests, the team’s overall automation capacity increased without any change in headcount.