Jumbo2010 - Fotolia

Evaluate Weigh the pros and cons of technologies, products and projects you are considering.

Can automated application testing be better than manual QA?

Vendors have been itching towards automated application testing for a long time, yet there is still room for growth. Amy Reichert offers her insights.

I remember automated application testing vendors swearing that manual QA would be going away. That particular conversation has happened over the course of 20 years.

Automated application testing has been at the forefront of the testing discussion for ages, but manual QA is still alive and well. Automation is still a work in progress, as early tools only worked on relatively simple applications. I don't know about you, but how often do customers need, use or want a simple application? Granted, I do use the calculator tool, but I wouldn't purchase an automation tool for it.

Another reason automated application testing development isn't as strong as you'd think it'd be by now is because it frequently fails to deliver value. It doesn't find significant defects and it's usually abandoned due to the high cost of maintenance.

That's not to say it can't be successful. It can, but only if it's planned on the basis of realistic expectations and factual data. There's no replacement for facts. The reality is it's going to take time to plan out which test cases, in what areas, to automate and in what priority order. Gathering input from developers and testers is essential to come up with a test automation plan that is realistic and maintainable. It's also important to do test runs and verify that the automation path you've chosen works for both the simple and complex application functions.

You may consider doing a pilot project on a mix of functional test cases that broadly cover the application. Determine which areas of the application cause trouble or where your scripts are running into problems. Run the pilot for a decent amount of time to detect any maintainability issues over a period of three to six months. It's better to find out earlier than later in the event that replanning or significant adjustments are needed.

Automated application testing development is viable when planned with realistic goals based on fact. I don't believe manual testing is replaceable, but I also believe automation has value when done appropriately.

Next Steps

We have automated dinosaurs but what about automated testing?

Today's top automated functional testing tools

Is test data management a better career choice than test automation?

Dig Deeper on Topics Archive

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Should automated testing eliminate manual testing? Why?
That's an interesting question. I can offer a few other.
  • Can automated management eliminate manual management?
  • Can automated recruitment eliminate manual recruitment?
  • Can automated parenting eliminate manual parenting?
  • Can automated kissing eliminate...

Well, you got it. It's not about "manual" way. It's about human way.

For further amusement I suggest the following article: http://www.satisfice.com/articles/cdt-automation.pdf

This question feels like a misunderstanding of definitions.

What is manual testing?

What is automated?

Programmers used to use machines to do program for them automatically on chips, later the got disk drives and RAM, and then had compilers to convert higher level languages to lower language which machines understand.

The answer is no, because all testing is automated to a degree when it involves a computer.   And unless you are testing dead tree manuals (which I doubt) then you probably aren't doing manual testing at all.
Very valid question in the current DevOps scenario.
Automation is a buzz word used in testing to attract developers into testing team. Assume if no manual testing and the code quality is bad, the tool cannot be blamed. This has to be addressed in manual testing phase to decide whether to automate or not. Also, proper feasibility analysis will help cost saving on tools and effort. 
It’s not really a question of manual vs. automated. It’s more a question of to what extent is your testing is automated?
Yes, manual testing could not be replaced by the automation testing. The automation testing shall rerun the scripted test n-number of time. But in manual testing we tend to explore the application based on the current change and identify more scenario and unveils defects. More over automation could be done only on fairly stable application other wise effort would be spent on making the code work.
You need both. Automated can test the bulk of the conditions and parameters set much quicker. Manual is going to catch more of the human error situations that the automated process may not have been looking for. When we assume certain things when designing the automated process it the obviously wrong exceptions that will make it fail.
I agree, automated testing can't replace manual testing, but it can complement it. And it can be better in some ways. For instance, performance testing and load testing is definitely best automated, if possible. In a lot of cases, it's too difficult for a tester to manually create the volume of data/connections/transactions (whatever the case may be) to accurately & efficiently test those aspects.
@Abby - but it's impossible for automation to design workload models and to analyze execution results. From this perspective, it's not automated testing but automation-assisted or tool-assisted testing.
And we use tools all the time. In general, all testing - and all programming - is tool assisted.
I do not think it's better, just faster. There is still to much human error that automation does not take into effect to solely rely on just one method.
Automated to what degree? Let's face it we don't live in an age of Ghost in the Shell and we don't have wide spread use of androids to catch many visual problems. But are that many people really testing manuals? (I kid)

There is tremendous low hanging fruit in Unit and API testing. Its when he UI gets into play that thins get tricky and a little bit of wisdom is needed.

Certainly. You can have a very unproductive commodity test effort that misses every single bug in the system, and you can have a well-planned and implemented automation suite. You can also have a totally useless, haphazard automation suite and a well-planned and implemented manual test effort. It’s not whether the effort is manual or automated, but rather how well thought-out and executed it is.