In the two previous posts (part 1 and part 2) in this series we looked several styles of BI we encounter in our work as business intelligence consultants. In this final installment, we’ll look at predictive analtyics and activity monitoring: two styles of BI that many companies find particularly challenging to implement successfully.
Predictive analytics is similar to advanced analytics, except its focus is on predicting future results, as opposed to understanding current and historical data. Predictive analytics require an organization to have relatively high BI maturity; typically advanced analytics to understand the past are a prerequisite.
Example analytical techniques
- Statistical forecasting (Box-Jenkins, many others)
- Monte Carlo simulation
- Market basket analysis
- Credit scoring
- Many others
Many of the techniques (examples: regression, data mining) used for advanced analytics also can be used for predictive analytics.
Example software tools
Typically, statistical packages like SAS and SPSS, although each predictive technique might require a specialized software product or product module optimized for that technique.
Activity/event monitoring and alerts
Activity/event monitoring refers to real-time or near-real-time monitoring of business processes. Such monitoring can feed process status displays (dashboards) and/or can trigger alerts when specified exception conditions occur. The alert could be delivered via an audible alarm, an e-mail, a page, a phone call, or other methods sent to the person or people who need to take action in response to it. This style includes information presentation (via the status display and/or the alert) but often no user interaction with the information.
Example software tools
Most mainstream BI software vendors have functionality and/or a module to do activity/event monitoring and alerts.
Applying the BI Styles
The reason we pay so much attention to the styles of BI is that they are a critical element of the requirements for a BI system. If the BI system provides the requested information, but doesn’t deliver that information to the user in a form that he/she can understand it, interact with it, and apply it to the business problem of interest to make a more informed/better decision, the business value the system was built to provide will remain latent.
Here’s the takeaway from this discussion of BI styles: when eliciting the requirements for BI from a user, pay close attention to both what information they want and how they want it presented to them and to interact with it. Ask probing/clarifying questions like:
- How would you like the information presented to you?
- When you get the information, what exactly will you do with it?
- What are you looking for in this information – specific details, an overall sense of how things are going, trends, anomalies and outliers?
- Do you want this information to be a fully-cooked meal that you just consume (in which case a report or an alert might be the appropriate BI style) or ingredients from which you will cook your own meal and then eat (in which case ad-hoc query, advanced analytics, or predictive analytics might be best)?
- Is the purpose of this information just to make you more informed, to help you decide between alternative courses of action, to cause you to take an immediate predetermined action, or something else?
The answers to such questions will help you determine the right style(s) of BI for that user.
Satisfying both the “what” and the “how” are key to a successful BI system that delivers business value.
by Bill Collins