One of the biggest hurdles to modeling business processes is taking reliable snapshots of what really goes on, as opposed to what's supposed to occur. Fujitsu, whose enterprise BPM product that isn't well known outside its Japanese market (and used to be actively marketed by Software AG prior to their webMethods acquistion), is introducing a deceptively simple idea for finding process needles in proverbial haystacks. They sample log files from all systems touched by end users, whether it be databases, applications, or even standalone Excel spreadsheets or emails.

Fujitsu has developed a data mining approach to process modeling, where it scans log files of a variety of systems to tease out patterns. It stumbled upon the idea when it was facing a problem with its own business: lots of standalone, often undocumented applications were used and not connected to Interstage. In other words, lots of activity that was falling through the cracks.

Fujitsu’s BPM mining approach takes dumps from databases and snapshots of applications in use to identify events, analyze data schemas to generate process maps. The reporting tool provides a layered approach; for instance, at first glance, your process flows might look like spaghetti. Then filter for most frequently used process flows, or other factors such as recursive loops, ideal cases, exceptions, and so on. At that point, you can start seeing the forest through the trees. In our view of the demo, the graphics were still a bit primitive – the mining and reporting tool could use some cosmetic improvements that would make the reports even more useful. But the filters did indeed reveal patterns that you could use to make some real conclusions.

The process mining is being offered by Fujitsu as a consulting service, and although there’s no requirement for the customer to follow through and implement Interstage (the analysis is BPM tool independent), Fujitsu wouldn’t mind getting its foot in the door.

---