Early FPA

[Frank Vogelezang]

Today, IFPUG has published their UTip #3 on Early FPA and Consistent Cost Estimating. It’s good to see that they have adopted a lot of ideas that are common practice within Nesma for a number of years now. Since Sogeti published a research paper in 2009 showing that the high level FPA method was nearly as good as the detailed FPA method but far quicker I have only done detailed analysis when I was contractually obliged to do so.

How many of your FPA analyses is still by the detailed method?

Of the last 100+ analyses I have performed, for only 3 I used the detailed method.

[Micha Verboeket]

For me it is the same, at Mediaan (software development/IT consultancy company in The Netherlands) the latest ~50 functional size measurements I worked with or created were all Nesma high level (also known as Estimated/Global counts). Sometimes we use Indicative detail level but in an informal way, e.g. in initial discussions with our clients about applications or ideas.
We mostly use FPA for budgeting, offering, planning purposes where requirement details are mostly not sufficient enough for a detailed measurement. And even when software is delivered and documented, we mostly do high level Nesma measurements instead of Detailed measurements (easier to measure and document, including the maintenance on the measurements itself).

And finally, I personally think that the inaccuracy of the less detailed measurements are acceptable, taking into account the other uncertainties and inaccuracies of measurements or input for measurements (like documentation) in software/IT projects.

[Luigi Lavazza]

A couple of years ago, I published a paper that compares several methods for early estimation of FPA size.

If you are interested, you can download the paper from http://www.iariajournals.org/software
Look for vol 6 no 1 & 2, year 2013. The paper is titled “An Empirical Evaluation of Simplified Function Point Measurement Processes”, by myself and Geng Liu.

We found that there are a few methods that can approximate the detailed measure quite well.

[EdwinvanGorp]

Since the publication of our aforementioned research paper in 2009, we at Sogeti only used the detailed methode once, because we were contractually obliged to do so.

For al our other measurements we mostly used the Nesma high level variant and in some cases the Nesma indicative variant.

One of the reasons not to use the detailed method is that the functional documentation that we receive to perform the measurements on, don’t provide the necessary information for a detailed analysis. Mostly the relationship between the transaction and the data model (e.g. in a CRUD matrix) is not described, which makes it hard (or even impossible) to determine the number of RET’s and DET’s involved.

Another reason for not using the detailed method is a lack of time. The the Nesma high level variant is far less time consuming and gives comparable and (for our goals) reliable results.

[Author]

Posts