Rexer Analytics have just released the results of their 2011 survey – the 5th annual one, answered by over 1,300 data miners from 60 countries in the first have of 2011. The survey continued to show that CRM/Marketing, Financial and Insurance are the major commercial focus areas for data mining. It also reiterated the top three algorithms – regression, decision trees and cluster analysis while showing a significant increase in text mining with about 1/3 already using it and another 1/3 planning to.
Rexer Analytics have just released the results of their 2011 survey – the 5th annual one, answered by over 1,300 data miners from 60 countries in the first have of 2011. The survey continued to show that CRM/Marketing, Financial and Insurance are the major commercial focus areas for data mining. It also reiterated the top three algorithms – regression, decision trees and cluster analysis while showing a significant increase in text mining with about 1/3 already using it and another 1/3 planning to. Cloud-based platforms are still only 12% with most work done locally on desktops or laptops and open source data mining tools like R are growing rapidly.
A couple of decision management centric observations:
- Customer centric goals dominated with 6 of the top 8 goals being directly related to customers. Improving understanding of customers was top (a somewhat passive goal) but retaining them, selling them more and improving their experience all scored highly and these are classic goals also for Decision Management (which is why the Decision Management Summit this year will focus on decisions about customers).
- Key drivers for satisfaction with analytic modeling tools were variable discover/profiling, ease of interpretation, model quality metrics and visualization. Sadly ease of model deployment didn’t make the top 10. Analytic modelers are still much more concerned with perfecting their models rather than making sure they can be deployed and used which is a pity as I have noted before.
- Model performance was still the top measure of analytic success, beating out financial performance/ROI. Another sign that too many models are concerned only with their model, not on the business performance being improved by their model (the focus of Decision Management).
- Reasons for non deployment of models were dominated by effort/cost too high and result not understood but were closely followed by model failures, politics/lack of support and changing business situations (a consequence of taking too long, at least sometimes). Modelers who want to get more of their models deployed and used should perhaps spend more time picking tools that handle deployment well and deploy quickly (see 2) and more time focused on business results to build management support (see 3)!
You can find more about the survey on the Rexer Analytics site.
