We talk a lot about the power of predictive analytics*. Data-driven decision making is the goal, but to get there, organizations need to learn how to extract actionable information from their data. We also talk a lot about how this is different from traditional business intelligence (BI), where the focus is on historical reporting. But what does actionable information really mean?
Actionable information means finding useful patterns in data that can be used to predict likely future outcomes. It’s not a “Minority Report” sci-fi precognition, but a mathematical probability that can be acted upon. Many organizations struggle to understand what this means in practice, so looking at examples is a great way to translate this concept into concrete steps.
An excellent example of how actionable information from data science can be integrated into BI was shown in a recent webinar by Vijay Kotu, Advisory Board member of RapidMiner and VP Analytics, Yahoo (@VijayKotu) and RapidMiner, an open source data science platform. Vijay presented several examples of BI dashboards integrated with actionable information from data science. The wide range of roles and operational areas underscores the business performance improvement amplification achievable by integrating data science more broadly across the organization. The examples included:
- Customer Dashboard: Churn alert for a specific customer
- Product Dashboard: Cross sell opportunities
- Help Desk Dashboard: Sentiment analysis of text comments
- Operational Dashboard: BI operational metrics with data science alerts and calls to action.
The operational dashboard is comprehensive so I have included a screen shot, below. Front line business operations now have the usual set of historical BI metrics, plus actionable information that can be acted on to improve business performance. Data science results from the data science platform have been integrated with a BI/Visualization tool.
Source: RapidMiner
An operational dashboard also provides me with a great example of where decision modeling can be of tremendous benefit. Dashboards are visual. Decision makers don’t have to pore through figures on a spreadsheet. But many dashboards are a collection of metrics that are too busy, undermining their usefulness.
Cluttered dashboards can be cleaned up and focused on decision-making by using a decision model like a wireframe. Instead of thinking of the dashboard as a collection of metrics where data science can be appended, organizations can think instead of what decisions they are trying to improve and what actionable information will improve those decisions.
The goal is better data-driven decisions, so being explicit about the decision making using the decision modeling approach makes sense.
The webinar goes on to describe four different analytical architecture options to achieve different levels of integration and is well worth watching. I want to thank Vijay and RapidMiner for the excellent webinar. Vijay also has a book, Predictive Analytics and Data Mining: Concepts and Practice with RapidMiner, that I haven’t had a chance to read but based on the clear presentation of concepts in the webinar, I predict the book will be very useful.
To learn more about decision modeling for integrating BI and Data Science, check out:
- Decision Modeling for Dashboard Projects
- 5 Steps to Streamline Reporting for Better Data-Driven Decisions
—————————————
*Predictive analytics are a set of data science techniques that, thanks to the advances in software and hardware, are available now through a wide range of software solutions. Many terms are tossed about in the “advanced analytics” space: machine learning, AI, deep learning, cognitive, data science, prescriptive analytics, predictive analytics and more. Gartners’ 2016 Magic Quadrant for Advanced Analytics used this definition, “Gartner defines advanced analytics as the analysis of all kinds of data using sophisticated quantitative methods (such as statistics, descriptive and predictive data mining, machine learning, simulation and optimization) to produce insights that traditional approaches to business intelligence (BI) — such as query and reporting — are unlikely to discover.” There are differences in when and why one or more of these techniques would be applied, but for understanding the importance of actionable information it’s not important.