Software testing of the future needs testers with a T-shaped profile: besides testing, they understand requirements engineering, architecture, development and DevOps. On top of that comes the know-how to test data-intensive systems such as machine learning applications.
Michael is a computer scientist who founded QualityMinds by accident (or maybe not). He has around 20 years of experience in software engineering, testing and QA, and agile leadership. What he calls his “lifeblood” is building high-quality products, following technology trends and helping people grow. Over the past 10 years, he has founded QualityMinds and several startups in Germany and Poland. In his free time, he loves the mountains and motor racing.
The Challenges Facing Software Testing of the Future
We as a testing community certainly won’t run out of work. But the kind of work we know as testers is going to change. The biggest challenge is seeing the whole system. It’s not enough to think only about analytical quality assurance, and about testing in particular. What matters is approaching quality assurance as a whole. Knowing about and being integrated into all the other disciplines in software engineering, such as requirements engineering, software architecture, development and DevOps engineering, will be a must, at least in the form of a T-shaped profile for testers. That applies to methods as well as to the tools that go with them, which the open source movement has made extremely varied in recent years. It hopefully goes without saying by now that most projects and product developments follow agile approaches, with all the implications that has for testing.
As if that weren’t challenging enough, the kinds of systems we test will change too. Alongside web, mobile, cloud and IoT systems, more and more data-intensive applications (with machine learning components, for example) are coming into play. Testing them changes, in many ways, the paradigms our testing is based on today. That leaves industry and research alike with the task of looking for such approaches and developing them.
Why Testers Need a Growth Mindset and a T-Shaped Profile
To take on quality assurance as a whole (not just testing), we first of all need the willingness, the mindset, to see it as an opportunity rather than a threat. As testers, we need to develop a growth mindset and learn a lot. About software engineering in general (especially in the agile world), but also about methods and tools in detail. To get closer to software engineering, there are bridges through “technical topics” such as test automation (no record and playback, no solutions far removed from the code), load and performance testing, simple integration into CI/CD pipelines or even test data generation. All of these areas are indispensable in modern software development, and as testers we can add a lot of value there and learn a lot along the way. Syllabi (such as ISTQB’s) can help, but individual learning through the very active global testing community matters much more. There are tons of sources and content online for almost every testing topic.
New Approaches for Testing Machine Learning Systems
The challenge of new kinds of systems under test calls for new approaches. Just as we as a community spent years working the field of mobile and IoT testing, we now have to take on testing data-intensive systems, for example, from simple machine learning approaches like recommenders all the way to deep neural networks. In research, there are active efforts in the field of Safe AI looking for methods and tools to assure AI systems. In practice, though, we still lack those methods and tools to effectively test systems created through training on huge amounts of data. It’s also conceivable that we could use AI in testing tools to test classic or AI-based systems. Some first approaches are on the market now, but on closer inspection we’re still “in our infancy” here.
How We Will Test: More Automation, More Insight from Data
In my view, one thing is certain: in the future, we won’t solve challenges with very, very large testing teams doing lots of manual work, and we don’t want to. The degree of automation in the testing process will have to go up. That mirrors the general trend in digitalization. What I hope we’ll see more of is that we can automatically derive insights, or at least recommendations, from the data we have as testers (test plans, test cases, test data, test logs and so on). That lets us steer coverage of the risks we focus on more precisely and make decisions faster. I believe AI-based approaches in testing that learn from data can help us enormously. With luck, that will give us more time for the really exciting things in testing (exploratory testing, for example) instead of fighting with broken test automation or even manual regression tests.
How Testers and Test Managers Can Prepare Today
Learn, learn and learn some more! As testers in particular, we need to keep educating ourselves, dig deeper into techniques and tools now and then, and keep an eye on trends. Nostalgia for past decades and old-school images of test management won’t help us. A healthy dose of curiosity, courage and openness will. Besides the well-known certifications and training courses, I can warmly recommend the active testing communities, meetups, online learning platforms and even YouTube. Every good testing conference has talks or workshops presenting the latest trends. And thanks to open source, there are countless frameworks and methods for almost every testing topic that you can use for free.
Frequently Asked Questions
Why isn’t it enough for testers to focus solely on analytical quality assurance?
Because quality assurance must be approached holistically. This requires knowledge of related disciplines in software engineering (from requirements engineering and software architecture to development and even DevOps engineering), at a minimum as a T-shaped profile. This applies not only to methods but also to the tools, which have become highly heterogeneous due to the open-source movement.
What changes when testing data-intensive applications that incorporate machine learning?
Such systems challenge the paradigms on which traditional testing is based. They are created through training processes involving very large amounts of data, ranging from simple recommendation systems to deep neural networks. In 2022, there were active efforts in the field of Safe AI in research, but in practice, there was a lack of robust methods and tools to effectively test such systems.
Through which topics can testers get closer to software development?
Through technical bridge topics: test automation close to the code rather than “record and playback” or solutions far removed from the code; load and performance testing; easy integration into CI/CD pipelines; and test data generation. These areas are indispensable to modern software development, and testers provide measurable added value there while continuing to learn themselves.
Are certifications and curricula enough to keep up with the technical demands of testing?
No. Curricula like ISTQB’s may have been helpful in 2022, but individualized learning through the world’s highly active testing communities was even more important. There are vast amounts of resources and content available online for nearly every testing topic. What’s crucial is a growth mindset that views change as an opportunity rather than a threat.
Can AI-powered tools already take over testing today?
Only to a limited extent. The first attempts to use AI in testing tools appeared on the market in 2022, but upon closer inspection, they were still in their infancy. It is conceivable that such tools could be used to test both traditional and AI-based systems. At that time, however, they were not yet a viable solution for testing trained systems.
Will manual testing disappear as the level of automation increases?
No, but its focus is shifting. Challenges should no longer be solved with very large testing teams and a lot of manual work. Once the struggle with malfunctioning test automation and manual regression testing is eliminated, time is freed up for the exciting parts, such as exploratory testing. Recommendations can be automatically derived from test plans, test cases, test data, and test logs.
Where do testers reliably obtain knowledge about current trends in testing?
From multiple sources simultaneously: active testing communities, meetups, online learning platforms, and YouTube. Every good testing conference features talks or workshops on current trends. Thanks to open source, numerous frameworks and methods for nearly all testing topics were available for free in 2022. Nostalgia for classic notions of test management doesn’t help.
