For a software company it is also important that this automation scales appropriately. In particular any automated analytic model development tool you use should provide flexibility in deployment to match your own, manage any variability between your customers’ data models and be easy to deploy, focused on the “last mile” of analytic model development. For most software companies, then, predictive analytic modeling solutions offering both on-premise and cloud-based solutions are going to be particularly valuable.
A predictive analytic modeling solution that can feed a “standard” model building process with customized data is particularly useful for those companies that offer this kind of customization. It is important that this kind of data model variation does not force a custom predictive analytic model development effort in every case. Some data models may be so customized that nothing less than a custom modeling effort will do but a solution that minimizes this has much to offer a software company.
If it is hard to integrate the predictive analytic models that result from your modeling solution into your software product you will spend time and money making your analytics actionable, time and money you could have spent elsewhere. A strong set of APIs, an awareness of how predictive analytics are being used in real-time and good integration tools are critical.
For more on this topic you can check out this recording of a webinar I gave for the PDMA with Predixion Software on The Power of Embedded Predictive Analytics, get the white paper “Becoming Predictive” (registration required) or check out the Predixion OEM microsite.
Next week: Step 4: Land and Expand