What happened
Their flagship product couldn't be automated with any testing tool on the market
The company manufactures a diagnostic blood analyzer used in hospitals worldwide. It's their most important product. And it had zero automated test coverage.
The problem wasn't a lack of trying. The analyzer's user interface uses virtualized elements and remote sharing. Tools that rely on locators, XPaths, IDs, or other identifiers in the code simply can't see what's on the screen. A limited number of testers were doing regression testing manually, and the analyzer wasn't tested as much as it should have been.
In a field where a software error can affect someone's health, that's not a comfortable position.
How they fixed it
Screen-based testing that doesn't need access to the code or installation on the device
TestResults interacts with applications the way a person does: by looking at the screen. It connects via VNC or RDP, both of which were already part of the analyzer's existing setup. Nothing gets installed on the device itself.
Virtualized elements, remote sharing, unusual UI technology, none of that matters. If a human can see it and click it, TestResults can test it.
The team automated the entire regression testing for the blood analyzer and freed up their testers for exploratory testing, where human judgment actually adds value.
Because TestResults is built for regulated industries, FDA compliance came without additional effort. No revalidation needed with every update, thanks to the Frozen Solution approach designed for MedTech and pharma environments.
e2e test cases automated in 102 days
software versions covered with each test case
average effort per automated test case
What changed
120 test cases. 3 software versions. 102 days.
The team automated 120 end-to-end test cases covering three different software versions in 102 days. That's less than 7 hours of engineering time per test case. Each test case covers the previous, current, and upcoming software version, all actively used by customers in the field.
Within weeks, the automated tests found bugs in the newer version that manual regression had missed in previous runs.
