We are devloping an application that is a kind of development tool for developing Line Of Buisness applications. The current applications that we build are windows desktop clients but we are looking into also targeting silverlight / cloud kind of applications.
What we are looking for is a "standard" way of measuring the time to build an application of medium to large complexity. The "easy" way would be to build two versions of an application, one using "standard" tools, like VS and components, and one using our platform, but I'm looking for a more efficient way to measure smaller parts and still be able to get some useful metrics of how much time could be saved using our product.
Do you guys have any pointers for me to look at, and what to test etc?
Everything that can be counted is a potential mesure. LOC, classes, components, dependencies etc, etc.
You specifically ask about "time measure" withouth actually masuring the time. Well, if you are going to document the time, then measuring the time is the only way to go. If you want to estimate or predict the time you think will be spent, then you could use some prediction tool. The COCOMO model is perhaps the best known model. This model uses LOC as the essential input, and some additional calibration parameters, like complexity, type of system, personell experience and historical calibration.
The original model has been redesigned (essentially including more parameters) and is refered to as COCOMO II. And the original model is renamed to COCOMO 81 (Boehm published the first model in 1981).
You'll find a lot of info if you google COCOMO.
The Wikipedia article is probably one of the first hits: http://en.wikipedia.org/wiki/COCOMO