I was getting too many questions on when and how to automate testing process. Instead of answering them individually I thought it would be better to have some discussion here. I will put my thoughts about when to automate, how to automate or should we automate our testing work? I know there some of our readers are smarter than me. So it would be always a good idea to start a meaningful discussion on such vast topic to get in-depth idea and thoughts from experts from different areas and their experience in automation testing.
Why Automation testing?
1) You have some new releases and bug fixes in working module. So how will you ensure that the new bug fixes have not introduced any new bug in previous working functionality? You need to test the previous functionality also. So will you test manually all the module functionality every time you have some bug fixes or new functionality addition? Well you might do it manually but then you are not doing testing effectively. Effective in terms of company cost, resources, Time etc. Here comes need of Automation.
– So automate your testing procedure when you have lot of regression work.
2) You are testing a web application where there might be thousands of users interacting with your application simultaneously. How will you test such a web application? How will you create those many users manually and simultaneously? Well very difficult task if done manually.
– Automate your load testing work for creating virtual users to check load capacity of your application.
3) You are testing application where code is changing frequently. You have almost same GUI but functional changes are more so testing rework is more.
– Automate your testing work when your GUI is almost frozen but you have lot of frequently functional changes.
What are the Risks associated in Automation Testing?
There are some distinct situations where you can think of automating your testing work. I have covered some risks of automation testing here. If you have taken decision of automation or are going to take sooner then think of following scenarios first.
1) Do you have skilled resources?
For automation you need to have persons having some programming knowledge. Think of your resources. Do they have sufficient programming knowledge for automation testing? If not do they have technical capabilities or programming background that they can easily adapt to the new technologies? Are you going to invest money to build a good automation team? If your answer is yes then only think to automate your work.
2) Initial cost for Automation is very high:
I agree that manual testing has too much cost associated to hire skilled manual testers. And if you are thinking automation will be the solution for you, Think twice. Automation cost is too high for initial setup i.e. cost associated to automation tool purchase, training and maintenance of test scripts is very high.
There are many unsatisfied customers regretting on their decision to automate their work. If you are spending too much and getting merely some good looking testing tools and some basic automation scripts then what is the use of automation?
3) Do not think to automate your UI if it is not fixed:
Beware before automating user interface. If user interface is changing extensively, cost associated with script maintenance will be very high. Basic UI automation is sufficient in such cases.
4) Is your application is stable enough to automate further testing work?
It would be bad idea to automate testing work in early development cycle (Unless it is agile environment). Script maintenance cost will be very high in such cases.
5) Are you thinking of 100% automation?
Please stop dreaming. You cannot 100% automate your testing work. Certainly you have areas like performance testing, regression testing, load/stress testing where you can have chance of reaching near to 100% automation. Areas like User interface, documentation, installation, compatibility and recovery where testing must be done manually.
6) Do not automate tests that run once:
Identify application areas and test cases that might be running once and not included in regression. Avoid automating such modules or test cases.
7) Will your automation suite be having long lifetime?
Every automation script suite should have enough life time that its building cost should be definitely less than that of manual execution cost. This is bit difficult to analyze the effective cost of each automation script suite. Approximately your automation suite should be used or run at least 15 to 20 times for separate builds (General assumption. depends on specific application complexity) to have good ROI.
Here is the conclusion:
Automation testing is the best way to accomplish most of the testing goals and effective use of resources and time. But you should be cautious before choosing the automation tool. Be sure to have skilled staff before deciding to automate your testing work. Otherwise your tool will remain on the shelf giving you no ROI. Handing over the expensive automation tools to unskilled staff will lead to frustration. Before purchasing the automation tools make sure that tool is a best fit to your requirements. You cannot have the tool that will 100% match with your requirements. So find out the limitations of the tool that is best match with your requirements and then use manual testing techniques to overcome those testing tool limitations. Open source tool is also a good option to start with automation. To know more on choosing automation tools read my previous posts here and here.
Instead of relying 100% on either manual or automation use the best combination of manual and automation testing. This is the best solution (I think) for every project. Automation suite will not find all the bugs and cannot be a replacement for real testers. Ad-hoc testing is also necessary in many cases.
Over to you. I would like to here your experience about automation testing. Any practical experience will always be helpful for our readers.