To become predictive, to use predictive analytics to differentiate a software product, you need to identify the decisions that can be made more profitably, more accurately, more precisely or more reliably. Unless the predictive analytics embedded in your product improve some decision-making then they will offer no real value. At best such analytics will be relegated to checking off the “does the product use predictive analytics” check box.
- Be Specific
Not all decisions are going to be equally impacted by predictive analytics and it’s important to rapidly identify some clear opportunities. Quick wins, low hanging fruit, will establish the value of predictive analytics and show momentum. Suitable decisions are going to be action-oriented, valuable, measurable, data rich and repeatable.
- Manage or support
Having found a suitable decision the next step is to decide if it will be managed by the software or made by the user and supported by the software—decision management or decision support. Successful use of predictive analytics in decision management, decision support or blended use cases alike requires a clear understanding of the decision. Identifying the decision, breaking it down into its components and identifying the information and knowledge required to make the decision will be required in every case.
Predictive analytics can be complex to build and to understand but they should not be complex to consume. This means that regardless of the approach taken, whether decision management or decision support, the product must use predictive analytics to deliver a simple interface.
A valuable prediction is one that improves decision making over all. The costs and benefits of the actions that result from the decision must be considered.
Next week: Step 2: Focusing on automation