It's interesting to argue the case for your software product, for and against. And whilst there are devotees for the function approach, there are serious detractors out there too. So lets consider the arguments. After all, if you are going to use BF its quite likely that you will use it for some pretty important stuff - it’s a important decision.
- There are some functions you just couldn't model yourself with the sane degree of accuracy. S curves, property leases - if you want "accountant" quality accuracy, we don't see how you can not used BF or something similar.
- BF has been around over 12 years and been actively managed. Its still got bugs (;o!) but is substantially debugged and reliable.
- Its fast, documented, loads of examples etc.
- If you are at all familiar with the library, picking up someone else's model is much easier than deciphering freeform Excel formulae.
Is it a panacea? Is there a downside? Well, BF has its detractors who claim, with some legitimacy:
- "It's over-complicated": This is possible. We are trying to put a lot of capability into a single simple component, a function. This even results in a handful of functions that actually have 29 arguments! We try to address this with documentation and examples and many defaultable arguments. But the simple answer is if you want to avoid complexity, don’t use the more complex functions with the long argument lists. Many short functions like the Con… family and the Pmts… family don't have many arguments and are very powerful and will greatly improve your model (in our view).
- "It's a black box". Of course, decluttering spreadsheets - by abstracting calculations away to a library - does to some extent black-boxify your models. We address this by making the source code partially available in the help files. We also have a Trace facility that outputs debug messages from all the functions. Also, as a design methodology, the functions tend to call each other and use a quite small kernel of core functions, thereby reducing bugs. It is perhaps more a set of building blocks, each of which is a black box, rather than a single black box.
The key argument really is to what extent you are happy with black-boxing some really important formulae. Well, ever since the stone-age we have been black-boxing tools into things we use everyday, so its not so evil to do this in a spreadsheet, is it? And the key benefits are fast, accurate, compact and flexible models that your competitors simply cannot build!