Saturday, June 09, 2012

Outsourcing Prediction Models

Higher education can benefit from many sorts of predictors. One of the most common is modeling attrition: trying to figure out how to identify students who are going to leave before they actually do. Because you might be able to do something about it if you understand the problem, right?

Over the years, I've been underwhelmed by consultants who cater to higher education, and it seems to me that we get second-rate services.  I have seen only a few consultants that are worth the money, and I'll just mention one: SEM Works out of Greenville, NC. Jim Black seems to know everything there is about admissions, and I've seen direct benefits from his expertise.

There are various attractions to going outside for expertise. Maybe you want to make cuts, and need a hatchetman. Or maybe the decision makers are convinced that a "secret ingredient" is required, and the consulting firm has enough mystique to sell it. I don't believe much in secret ingredients. 


What's particularly galling is that the advanced methods of building predictive models come from universities. Then they get used at and Netflix and what have you, while our administrations still chug along on 50 year old technology. So we go out and buy consultants. Who also use 50 year old technology.

You may not be suspecting this twist, but I actually came across a solution for this problem. That's the whole reason I'm writing this post. I'll just let free market talk now, from

So instead of contracting up front for who-knows what sort of analysis, you can actually bid out the work and get a solution before you pay! Of course, you have to think hard about what data to submit, and spend some time cleaning it and prepping for such a submission. But judging from the rates on the site, it would be a lot cheaper, with an almost guaranteed better result, than buying a big name consultant. I'll leave you with this quote from their About page:
The motivation behind Kaggle is simple: most organizations don't have access to the advanced machine learning and statistical techniques that would allow them to extract maximum value from their data. Meanwhile, data scientists crave real-world data to develop and refine their techniques. Kaggle corrects this mismatch by offering companies a cost-effective way to harness the 'cognitive surplus' of the world's best data scientists.
Update: Also see "Not so Expert" in The Economist.

1 comment:

  1. I admire the idea of your secret ingredients of your article here. I also like the video you used as an example of your post.