There’s a certain joy to be had in working for a software vendor. The usual position is that the client has purchased your software, and they ask for assistance in implementing it into their business. You provide this assistance on a project basis, during which you charge a small fortune to display your expertise. Ideally, you manage to move them from a position of complete ignorance to near-independence. They should know enough about your product to understand how wonderful it is, and how much they need it. They should be able to use it well enough to do simple things for themselves. But their knowledge should always fall just a little short of being able to do the really complicated things – for in such things are millions to be made in consulting.
I worked in that exact position for five years, and it was a pleasant experience. You’re regarded as the expert, and what you say is usually taken as gospel. The client defers to your knowledge and decision-making prowess. But while I never took advantage of them, I don’t think I fully understood what it must be like from their point of view.
Until now. Last week, I was in a meeting discussing how we can perform some checks on our brand-new valuation model (provided by Ye Large Software Vendor). The key item we wanted was a runlog, which would give all the information in terms of what inputs went into a run of the model, and how the run was set up. This is a standard output of any actuarial model, and is vital for a Solvency II world, where everything has to be validated. Some software does it well (Prophet is pretty good), some less so (MoSes can be a bit sparse unless you design it to be verbose enough for your needs). My personal favourite is VP/MS, where you can take the runlog, and convert it into an input file in order to exactly reproduce the run.
And so we asked our vendor to produce a runlog, and were told “Well, the runlog exists, but you can’t see it.”
Only a software vendor employee would even think that they’d get away with such a comment. I think, though, that I was alone in the room in understanding their position, and where they were coming from, and so I didn’t ridicule them in public. Well, not too much. Not until now, anyway.
Because this idea of an existential runlog is just too good to be true. If I am at one with the model, will I be able to divine it? Or do I have to have faith in the potential of the runlog to answer my questions before it will reveal itself to have answered them? Perhaps some ascetic pilgrimage is required, or maybe I need to turn my computer in a certain direction in order to be rewarded with a result. Could it be that the runlog has to be deleted before it is revealed anew in a more visible form?
Who can tell. ‘Tis verily a mystery.