Costing the segmenter

June 8th, 2010 by donnek Leave a reply »

Just for interest, I registered the Swahili verb segmenter at Ohloh. This is quite a clever setup, because they analyse the various bits of code in the repo and come up with a nice set of tables on the analysis page. The number of code lines comes out at around 2,400 (not counting comment lines or blank lines, which I like to strew liberally around my stuff, so that I have at least a chance of remembering what it’s supposed to do!), and on the little widget I’ve added to the site, it concludes that the segmenter would cost around $27,000 (£19,000) to write from scratch.

However, this is of course a little optimistic. Firstly, the Basic COCOMO model they use tends to overestimate the value of small projects. Secondly, about 38% of the code consists of the very nice CSS stuff from Blueprint – not really “mine”. Lastly, the average salary is assumed to be $55,000 per year, which is probably unlikely around here – I think between $35,000 (£24,000) and $45,000 (£31,000) would be more realistic. So if we put the lower figure in, and assume only 62% of the code is my own, and take a bit off for being a small project, that comes to around £9,000 ($13,000), which is probably a closer reflection of the monetary value.

It’s still quite a significant amount – the equivalent, assuming the lower salary figure, of around 4.5 months of work. Since the actual coding time was probably only about half that, it suggests that around 50% of “standard” costs go on things like overheads, meetings, etc. Perhaps that’s another argument for the free software development model – more emphasis on the code than on the organisational framework for it.

Leave a Reply