Is it possible to do accurate predictive modeling in-house, or does it require outside expertise?
Data analytics has been a great way for us to gain insight into our customers behavior, but it doesn't really tell us what they will do in the future. We are discussing a predictive modeling project, and there is some debate whether we can do this in-house, or whether we need to contract out for the project. Obviously, the downside of the second option is cost, which will run into six figures. On the other hand, using our internal resources only results in savings if the results are valid. How accurate has predictive modeling generally proven to be, and is it realistic to take on the task in-house, or does it really require outside experts to get it right?
Answers
I suspect that the answer will vary, depending on the company. It might be a good idea to find out what specifically the outside company can do for you in terms of services. Then sit down with your internal group and find out if they can replicate everything on that list.
If they can't then the outside company might be your better bet if you really need accuracy as reflected in their list of services. But if your internal people can do it then have a go at it and see what happens.
- Share this answer
- Permalink
Related Questions
I recently read a blog post from Dan Blacharski called "Vendor Financing Hard to Come By" which he posted in 2009 and I am curious how...
I've been hearing a lot about Big Data. What makes this any different than the large amounts of data we've been generating up to now?
What's standing in the way of the growth of big data analytics? How come we haven't seen more of it? Is there still a penalty for early...
Ask a question
White Papers & Webcasts
White Paper
Turn your information into enterprise value
See more White Papers | Webcasts
One of my degrees is a B.S. in anthropology, and in that field agent based modeling, which is a type of predictive modeling, is sometimes used to objectively measure behaviors within populations. I had a professor who utilized it extensively in his research and I gained some exposure to the methodology. Basically, you have to create a mathematical model that predicts how a system or population will behave based on prior data. It is not easy or simple in any way. I hate to admit it, but when my professor was explaining some of the finer points of his modeling, I had difficulty follow the process. Unless your company has some serious talent and experience with predictive modeling, I don't think it would be feasible to do it in house. I would expect to pay mid six figures to bring in someone and develop a project for a company. Maybe more. If they know what they are doing, however, the investment can pay off. Back at university, it was pretty amazing to see how closely actual behavior tracked the results of agent based modeling, but to get those sort of results you really have to know what you are doing.