10 Years in Testing, 10 lessons - Lesson 2

Welcome to Lesson 2: Testing cannot be automated

This is another pet peeve of mine.

Why on earth do we talk about test automation as if we can really replicate testing? We don’t talk about programming automation now, do we?

To me, people who say “all testing can be automated” have no fucking clue what they’re talking about. Do they have the ability to automate thinking then?! Because that’s what testing is: THINKING.

Let me explain.

Testing is a cognitive activity, just like programming. Your brain is the most important tool in order to do it well. Programming isn’t about typing the code into your editor of choice, it’s about knowing what to type and knowing how to structure the code effectively so you don’t fuck everything up for your future self and you can still refactor it, keep up with changing requirements etcetera.

Testing is just as complex as that! I’m not doing shit at random (people often seem to think testing is easy), I have a strategy and I’m applying test design. I’m trying to think of what to test at which layer, where the risks are and what bits in my testing are repeatable (so automation can actually help me). I’m also trying to figure out what would be terrible to automate and good to explore. In short: I’m thinking, I’m using my brain.

So please tell me you no longer believe you can automate this entire testing process. That’s just bollocks.

That’s why I appreciate what Mark Winteringham and Richard Bradshaw have done with this topic.

They renamed test automation to what it actually is: automation in testing. Because yes, there are certainly tons of tasks in testing where automation can be extremely helpful. Test automation doesn’t replicate testing though, it supports testing. It supports our eternal quest for more useful information.

The wording matters, incredibly much.

You cannot automate testing as a whole. Saying things like “all testing can be automated” is incredibly disrespectful and ignorant of what testing can do. I always feel like someone who says this is suffering from the Dunning-Kruger effect.

So please check out Mark and Richard’s work and I hope you don’t believe that testing can be automated (if you ever did).

Thank you for reading, you can find the rest of the blogs here.