Last week I traveled to copenhagen on a trip sponsored by our Denmark office. Originally the trip was to meet with a few customers so they could learn enough about SQL Server Data Mining to start using our betas and then do a few sales and marketing types of meetings with a few more to more "higher-ups" to push the platform.
What was intended to be a small training session evolved into a full-blown two day data mining course with about 20 people from 11 different companies will skill sets ranging from strong MS platform skills with no data mining to no MS platform with strong data mining. The turnout was so great that the local office wasn't big enough and we had to hire out a dedicated facility north of Copenhagen. The attendees were from many different verticals ranging from banking, telecom, manufacturing, retail, and consulting.
A common theme among many attendees was "Can I adopt SQL Server Data Mining so I don't have to pay for SAS?" I believe that by the end of the two days, the common belief was "Yes!" In fact one customer stated that they were going to start moving to SQL Server Data Mining right away. It was a very exciting course where I was able to bring together people with platform knowledge with people with data mining knowledge so they can make predictive analyrical solutions happen that were simply impossible in the pre SQL Server DM era. We went over virtually every bit of DM user interface, DMX, Nested Tables, Algorithms, programming object models, and integration with Reporting Services, Integration Services and OLAP. We managed to get through all the material despite losing my voice and wierd laptop behavior. One of the programming examples I gave turned into my weekly tip or trick and very serendipitously I found this article relating Clay Cristiensen's Innovator's Dilemma to Microsoft BI the day before I spoke about developing data mining applications
In working with these customers and talking to the two additional organizations I met with the following day, it really has made me realize much of the value proposition in what we are offering in the data mining space. Despite the cost advantages and ease of use we present in SQL Server Data Mining, one of the biggest values we have is what I'm starting to call "Total Model Lifetime Management." Many data mining tools are just that - tools, and they are designed to perform data mining operations and they stop once those operations are finished. Microsoft's data mining, less by intention and more by a general overarching philosophy, never stops. Our vision doesn't end when an analyst completes and verifies a model. Our vision continues through to the operationalization of the model and the continued maintenance and management of the model. Customers have complained to me that with their existing data mining solutions it can take easily four or more weeks to put models into production once the models are complete and it takes specialized skills. For SSDM, it's four lines of code. Or it's even no code, since you can use Integration Services, Reporting Services or even OLAP to leverage the analytical results. And then, once the model's in production, what do you do with it then? Being part of a "complete data platform" means you do what you would do normally with anything else. You apply permissions, reprocess as necessary, use the profiler and perfmon counters to determine the usage and health of the model, back it up, restore it, sync servers, etc. etc. etc. - the data mining world is your oyster.
One thing that I find interesting is that I talk to database people and all of this is simply so obvious - of course you have integrated security, of course you have a standard data access model, of course you have basic management capabilities. But to the data mining world this is very very new - and it's the kind of thing that has kept data mining out of the mainstream for so long. But now that it's here....