I remmember a while ago I was consulting a company that had some trouble budgeting its small QA department the methodology that they used to favor for time estimation was "percentage of development", and by doing so were completely ignoring two major factors needed for good QA estimation:
1.Complexity of the SUT (system under test)
2.Average number of Test Runs
Those two factor will allow us to use more accurate, though more time consuming, estimation methodology called "Internal overview". By moving to this new approach the company improved significantly its budget planing and had been able to provide more fitting and accurate pricing for their projects.
To emphasize the shift that we did with that's company QA department lets look closely on the two methodologies:
Percentage of Development
This methodology assumes that the amount of time invested in QA effort its always a subset of the development work, usually around 50-60 percent. So if developer requires X number of hours to develop the new feature then Qa work estimated as 0.5*X=(QA ESTIMATE)
Sometime there are more generous departments which will estimate QA effort as equals to development which will bring us to the formula of X=QA ESTIMATE.
Advantages
- Really easy to estimate
- Doesn't requires QA involvement in the estimation process
Disadvantages
- Ignores different complexity levels
- Ignores the need for new test cases and planing
- Ignores the need for different number of test runs
- Highly inaccurate
Internal Overview
To get a good estimation for QA activities we need to understand QA's involvement in all stages of SDLC. QA department should begin its work alongside with development in order to plan and prepare proper set ot test cases. Estimation for this part done by investigation and understanding of the system requirements.
Usually this task is shorter than the development effort on big projects and can be longer than development on small new features requests. It is extremely important to be aware of that when planing a formal release.
Sometime ago I used to work for a company as a QA manager and part of my responsibilities was to provide time estimations for my department. I can tell you that it always was hard to justify QA activities for small feature releases then for big new projects. If for example a new web site required, it will take much longer for development to crate this website then for QA to prepare a solid test suite for it, but when a new button requested on all pages of the website its much faster for the development to create that button, then for QA to plan all possible scenarios of how that button could be used.
Later on when development is done QA execution should begin. The estimaton for that can be done by taking into account the number of tests to be executed in one full test cycle (TEST SUITE TIME) and multiply it by the number of test runs. Now, here my recommendation will be to take 3 test runs as average. This number can change based on the testing environment and the actual SUT, but for average software release 3 test cycles its a save bet. That will give us following QA estimation formula:
(TEST SUITE TIME)*3= (QA ESTIMATE).
A note will be appropriate at this point, as I had such a confusion before, the (QA ESTIMATE) time will be the net value of QA activities. Usual practice is to involve development for fixing rounds in-between testing cycles so the total time for a specific release increasing based on development work.
Advantages
- More aqurate time estimation
- Better quality results
- The estimation includes all QA activities
- The estimation provides better overview of all QA tasks.
Disadvantages
- This sort of estimation more time consuming
- Requires investigation and QA involvement in the estimation process
My recommendation for every company that I'm consulting is to use the "Internal Overview" methodology for QA time estimation on the regular basis and to leave the "Percentage of Development" approach only for extreme cases.
Comments
Post a Comment