Rod Gill said:
Actually, there are some very good ones. But they tend to be pretty
specific to the organization that created them. A good estimating template
will be calibrated to the organization, and recalibrated frequently.
The approach outlined by Rod is fairly typical. These estimates tend to be
optimistic by a large amount, which is why he suggests roughly doubling the
estimate.
With little experience to guide you, probably the best approach is to size
the request ... i.e., estimate the number of function points. Then look on
the web to get a range of the typical capabilities of different
organizations. Capers Jones, among others, has quite a lot of data.
Productivity varies on the type of application, the tools you use, etc., so
understand the variables. Recognize that since you don't have your own
estimating model, you are probably among the less productive shops, so look
to the low end of productivity in making your estimate.
If you have some history (sounds like you don't) a better approach is to go
back and size projects you have already done and see what it costs you to
deliver a function point.
Another way, subject to greater subjectivity, is to use a method like
Cocomo. Cocomo is fairly widely used, but it depends on estimating lines of
code. This is harder than estimating function points, and likely to lead to
more optimistic estimates.
An intermediate approach would be to size the request, then use a conversion
table for your development environment to convert function points to lines
of code (there are tables on the web). Then you can use Cocomo with a
somewhat more objective size estimate.
Whatever method you choose, begin keeping detailed track of the actual size,
size estimates, and costs of doing your projects. Over time, you can get
quite good at it, but only if you keep the data and pay attention to it.
To come up with an estimating template in Excel requires quite a lot of
data. Good templates ask a lot of questions about the application that you
can know at requirements time. Each of those questions will result in some
sort of factor. You need data from a lot of projects, many hundreds at
least, to calibrate all those factors.
...
