Introducción

Automation in clinical software: good, bad and not so bad experiences

Manel Moreno Jaimez    
Systelab - Spain

Presentation abstract

My first experience with SW Test automation really impresed me, it seemed to be a kind of magic. It was 15 years ago in a communications Lab of the first company I worked for. There was a computer monitor showing the UI of an application moving the mouse pointer, navigating, opening and closing windows, all without human intervention!
Now, after my own experiences, what seems to be the magic is achieving automation without dedicating huge amounts of time and money and having the feeling of obtaining a true payback from the tests you automated.
In my current company, that develops Clinical SW, I had the opportunity to start working on test automation. One of the things I have learned is that even having the management support, having the best tool or having very good skills as a developer, none is a guarantee for success in SW test automation.

From each experience you can have different results and explain different things. Four experiences will be referred here, providing details about good/bad points and lessons learned.
1. My first experience started with internal training for a commercial automation tool. I was then assigned to a task with a clear target: automate the existing sanity test of a little application. It is interesting to explain how we translated literary the steps of the manual test case, the scenario, why it was unsuccessful.
2. The next experience relates to the automation of tasks in a web based application, always with clinical purposes in mind. This task was longer and for a larger application. It was included in the project schedule and shared by the two testers on the team. A great deal of code lines, functions and the data to feed them was created. It was a sanity test, covering the main features of the application. Apart from its use for regression testing, specific automation was designed to cover very specific features. For instance a dosage algorithm tested over more than 2000 cases, only possible due to the automatic test. It was successful for 5 milestones of the project (that is about 3 years), until it was discontinued and never restarted due to the high cost of maintenance. 
3. &4. The next two experiences gave a successful result. Two different applications, one web-based and the other stand alone, showed common themes. First of all, and that has been key for the satisfactory result is automation approach and simplicity.

For the first application dull actions were automated and unexpectedly one important bug was found!
For the second application, very few features were automated but the automation was designed to be compatible with many of the multiple platforms running the application. Therefore this test can be executed for each version over more than 100 platforms covering OS, language, instrument configuration, analytical layout and monitor resolution. The content of the distribution CDs for each release was also included as part of test automation and proved very useful.
From these and other experiences in the company we can conclude that automation is needed but must be accurately approached for success in saving time, improving the coverage and allowing your testers to be focused on their main verification tasks.

Speakers information

Manel Moreno Jaimez has an Electrical Engineering degree and is currently Systelab’s Test Manager after 12 years in the company. Has long experience as a test engineer working for internal projects and also as a consultant for external companies. Apart from functional testing, he has experience in SW Test Automation, performance test, usability, web-testing, risk management and requirements management. He also possesses a Scrum Manager Certificate.
He has been a contributor author in one of the chapters of the last “Experiences on Test Automation” book, by Dorothy Graham and Mark Fewster.

volver