Descriptive? Diagnostic? Predictive? Prescriptive? Here’s how to design your own analytics agenda.
Originally published with Entrepreneur.
A Boeing 787 aircraft generates half a terabyte of data on an average flight. That’s an enormous amount. Though it’s filled with meaningful insights, the sheer quantity can make it seem like an endless maze.
The best way through the maze is to determine a path to follow. A recent survey by NewVantage Partners reveals that 97 percent of executives are investing in analytics projects. What this figure does not reveal is that each of those projects is unique, incorporating different tools in pursuit of individual goals. Business analytics are a powerful tool, but only if you embrace the type that is right for you.
Analytics come in four distinct types, and each builds off the other. Imagine a pyramid, with each level supporting the next: descriptive, diagnostic, predictive, and prescriptive.
The base of the pyramid is descriptive analytics, which report what happened and allow analysis of past performance data in order to identify known strengths and weaknesses. A descriptive report for United Airlines, for example, could show how many tickets the airline sold last month in different major markets.
The next level is diagnostic analytics, which report on why something happened and reveal what factors drive positive and negative performance. If United’s descriptive report shows that sales are down in one market, a diagnostic report might show that it’s because of reduced marketing spend.
With those as our foundation, we can use predictive analytics to report on what could happen, based on past performance. If United wanted to increase revenue, its predictive report could show where higher marketing spend would go farthest to increase revenue.
At the very top we have prescriptive analytics, a report on what should happen. Here, artificial intelligence and machine learning mine past data to inform future decisions. Prescriptive data could tell United’s marketers exactly which price points to push in their latest sales copy.
Analytics from the bottom up.
Companies often prioritize a single tier of the analytics pyramid without first securing more primary tiers.
The process starts with stabilizing your operational data. Only once operational data has been properly structured and organized is it possible to think about more advanced analytics. Companies that fail to build an analytic foundation are often saddled with huge tech costs, business intelligence software that can’t meet its full potential and weak overall insights.
After mastering descriptive analytics, companies can explore whether the higher tiers are even necessary for them. Having access to high-volume, high-velocity data is essential for some enterprises, but many others can get by with “small data.” As with all aspects of analytics, the key is to be realistic about your needs and capabilities.
Building your own analytics agenda.
Developing your own analytics agenda starts with asking the right questions. This helps you fully understand your needs and wants, your strengths and weaknesses. Begin with these important inquiries:
1. Do you have a single source of truth?
The first step of any analytics agenda is to clean up and integrate your data so that it’s consistent and reliable, because a company isn’t ready for even diagnostic reporting if it relies on conflicting data sources.
Access to high-quality data is a common problem in the medical world. Prudential Financial’s vice president of employee benefits, Andrew Gregg, said at U.S. News and World Report‘s Healthcare of Tomorrow conference that healthcare analytics across the U.S. is a “data dump” in a “sad state.” Care providers routinely store their information in disjointed systems, which can come to carry conflicting information because of inconsistencies in user input and a lack of standardization in data collection. That disjointed data is one reason the American healthcare system has been surpassed by other nations, Dr. Lisa Ishee of The Johns Hopkins University School of Medicine said at the same conference.
Step one is cleaning up your data so that it’s consistent and reliable. Decide on a comprehensive source of information to draw on. Without that, your insights will be inevitably incomplete and inaccurate.
2. Do you have a data analyst?
Data analysts are trained in statistical analysis and data modeling. Having one on staff is important for keeping data organized and actionable and for identifying and interpreting insights.
In a conversation, data consultant Lillian Taylor once described to me a business confusing correlation with causation. A longitudinal database study she conducted on K-12 attendance information showed a high dropout rate for students of one particular third-grade teacher. It would be easy to conclude, Taylor said, that this was an awful teacher, but there are additional data points to consider, including that the teacher may have been in an especially high-risk district.
Until you have an analyst on board, you risk drawing false conclusions from your data. Making someone specifically responsible for analytics ensures your efforts are productive and efficient.
3. Do you have dynamic reporting?
Dynamic reporting generates reports and visualizations essentially on demand. That way, they incorporate the most complete and accurate information possible and give you the most relevant insights available.
Say you have a board meeting tomorrow and need to prepare a deck. You go to your well-managed database or data source, export the required information, save it as a spreadsheet, and then build the visualizations you want in Excel or something similar. By the time your meeting rolls around, your insights have gone stale.
Instead, use one of the myriad reporting softwares to source your reports directly from the database and create real-time analysis and visualizations. Dynamic reporting will give you up-to-date information you can be confident in and will pave the way for using diagnostic analytics.
4. Do you have a business plan?
High-level analytics require high-level data. Predictive and prescriptive analytics are both possible, but the data and technology required are both expensive. Without the right financial incentive, the ROI on high-level analytics may not justify the investment. Companies must carefully weigh costs and benefits before pursuing more advanced analytics.
The first step is to make sure you have a way to put your data insights to productive use. Otherwise, the cost of acquiring those insights could be prohibitive.
The potential of analytics is endless, so it’s tempting to push projects forward and try to learn as you go. The better approach is to exercise caution and prioritize planning. The more that companies calibrate analytics up front, the more valuable they will be in the long run.