I will begin this by stating clearly: you should never get to the point where you need estimate how long it will take a big task to be completed. You should never have big tasks in the first place. There are a lot of reasons why having big tasks is a seriously bad idea.
Firstly, big tasks have big requirements and by the time the developer is halfway there he will most likely have forgotten parts of the requirements. A developer working on a big task needs to keep the requirement fresh inside his mind each day. Otherwise if he forgets some things he won't implement them at all. This is an issue even where I work. So extra time is wasted just refreshing those requirements.
Secondly, big tasks tend to affect a lot of systems inside an application. Most developers don't know all the systems in the application so he might spend a lot of time figuring out how those systems work and how does his task affect those systems. A series of smaller tasks can be worked on in parallel and you get more flexibility when assigning developers to those tasks. You can choose for example developers that are experienced working on areas and systems related to those tasks. On bigger tasks it's almost impossible to find a developer that knows all the affected areas. Having more developers with better understanding of the implications of the changes will reduce the development time and number of bugs.
Thirdly, it takes a longer time for a big task to reach the testing period. A small task reaches the testing period faster which means that the testers will never run out of work.
And finally, smaller tasks are easy to fix because you have a smaller portion of code that you need to analyze for bugs.
No wait, there's more. Smaller task are easier to accurately estimate. Even if some of them takes more to finish, you can keep a ratio of 50/50 of tasks that are finished earlier and tasks that are finished later. If you keep that ratio and sum up all the time it took to complete them you will have a better change of being on track because everything will balance out.
All in all, I think that the extra time wasted to split a big task into smaller tasks is recovered later due to the things I mentioned above.
Firstly, big tasks have big requirements and by the time the developer is halfway there he will most likely have forgotten parts of the requirements. A developer working on a big task needs to keep the requirement fresh inside his mind each day. Otherwise if he forgets some things he won't implement them at all. This is an issue even where I work. So extra time is wasted just refreshing those requirements.
Secondly, big tasks tend to affect a lot of systems inside an application. Most developers don't know all the systems in the application so he might spend a lot of time figuring out how those systems work and how does his task affect those systems. A series of smaller tasks can be worked on in parallel and you get more flexibility when assigning developers to those tasks. You can choose for example developers that are experienced working on areas and systems related to those tasks. On bigger tasks it's almost impossible to find a developer that knows all the affected areas. Having more developers with better understanding of the implications of the changes will reduce the development time and number of bugs.
Thirdly, it takes a longer time for a big task to reach the testing period. A small task reaches the testing period faster which means that the testers will never run out of work.
And finally, smaller tasks are easy to fix because you have a smaller portion of code that you need to analyze for bugs.
No wait, there's more. Smaller task are easier to accurately estimate. Even if some of them takes more to finish, you can keep a ratio of 50/50 of tasks that are finished earlier and tasks that are finished later. If you keep that ratio and sum up all the time it took to complete them you will have a better change of being on track because everything will balance out.
All in all, I think that the extra time wasted to split a big task into smaller tasks is recovered later due to the things I mentioned above.
Comments
Post a Comment