Skip to main content
What makes a good predictive analytics provider?

Whether you’re using scores for the first time or selecting enterprise- class modeling systems, choosing the right analytics provider is critical. A good analytics provider has both experience—in analytic technique and building models that work in practice—and the ability to deliver long-term business value. Here are five questions you could ask any provider.

  1. How experienced are you?
    How long have you been building predictive analytic models? What kinds of models have you built and for whom? Do you work with any consortia to build pooled models? How much of your business comes from predictive analytics work?
  2. How advanced is the technology you offer?
    Many analytic providers excel in one technique, which may not suit all problems. If your analytic provider has proprietary and innovative modeling technologies, from genetic algorithms and neural networks to Bayesian algorithms and experimental design, that they have successfully used on client projects that's a plus. If they just use the algorithms available in the general market, that's not. You want a vendor who can extract more value from multiple data sources.
  3. What kinds of problems do you specialize in? 
    While superior techniques and technology produce a sharper model—one that’s built to last and offers greater ROI—technology alone is not enough. Models must be built to meet business objectives. This requires a provider that has succeeded with a variety of industries and types of client implementations. Not only is experience in your industry useful, experience in other industries that might offer interesting insights is also useful.
  4. How will you improve my business?
    You want a vendor who can implement quickly, so you see results fast. They should develop solid implementation plans and have the resources to deliver in your time frame.They should be able to estimate business impact before deployment and should understand how to deploy the resulting models quickly into a decisioning environment such as business rules management systems.  They should offer post-implementation guidance so you continue to see improvements with your models, including clear procedures for evaluating model performance. And they should help you manage change across your enterprise, which can make or break success with analytics. Great analytics are no use unless you can get them implemented.
  5. Can you help organizations at my level of analytic experience? 
    Different organizations need different kinds of helps. Some companies need help to get started with predictive models. Some are very sophisticated and need an analytic partner .Your analytic needs will grow and change over time so make sure your vendor can help at each stage.

related posts