Hi guys, I am starting into estimation and just am a little bit confused about things.
How do we actually do testing time estimation. Is it just by estimation the time for development and then consider that it would take around 40% effort of the total dev time.
If yes then what all testing activities would this 40% time cover. Would it also cover UT, test enviroment setup regression etc. as well?
Also what are techniques like TPA, UseCase Points etc. Like if we can simply estimate by using 40% of the dev time then why use TPA or UCP.
Looking for inputs from all
We do everything starting with gut-feeling and ending up searching the previous project statistics - depending on whom WE you ask [img]/images/graemlins/smile.gif[/img].
I suggest you to do a search for "developers to testers ratio" and you will find some answers on your question
In my opinion UseCase or Function Points are techniques to estimate the "size" of the application. You can then estimate the effort (man/day for instance) by applying a factor to each "size" unit. As far as testing is involved, you should be careful when you use numbers like "40%" out of the box without definition of what you mean by testing, etc.
I understand that we first calculate the size of the application using techniques like TPA or UCP and then calculate the time needed by using statistics from past experience.
But what about a situation where the organization has never used any of the techniques before and hence does not know how much effort is needed per TCP or UCP.
This question drives me nuts. I can estimate how long it will take to execute passing tests for identified/defined risk areas/test cases/etc. No one can estimate how long testing will take.
If there were no bugs (impossible) and the intent were to verify that though a set of predefined tests, that would be possible. Outside of that, the time needed to test depends directly on the number of defects in the system, how difficult they are to find, and the degree of confidence a stakeholder wants to have in the tested areas of the system. Unless someone has a formula for that, I see test estimation at a project level to be some cross between black-magic and a waste of time in almost all cases.