Analytics 101: Descriptive, Predictive, and Prescriptive Analytics
One thing I’ve learned in my time as a data scientist has been that the term “analytics” means something different to just about anyone you talk to. Answers I’ve gotten ranged from basic statistics in a spreadsheet to customized machinery that knows when it’s going to break. One of the best things about working in data science is guiding clients along their analytics journey, whether that’s streamlining an Excel report into Tableau or building out machine learning models that predict product demand. One of the most important parts on the analytics journey is understanding the different types of analytics and what questions they can answer.
Let’s use the example of a sandwich shop and how the different levels of analytics can help them drive better business decisions.
Descriptive Analytics: What happened?
Before you can walk or run, you have to learn how to crawl. Descriptive analytics is the most basic form of analytics, focusing on describing what happened in the past. Most often, this is done in basic reporting through spreadsheet tables or shiny dashboards. In the case of our sandwich shop, they can use descriptive analytics to answer some of the following questions:
What was the best selling sandwich this month? What was the worst?
What day did I sell the most this week? What about selling the least?
Am I getting a lot of returning customers?
Answering some of these simple, but extremely important questions are critical to making some key business decisions on how to manage the business. After running some basic analytics, the sandwich shop realizes that the turkey sandwich is very popular, but the tuna salad sandwich barely breaks even.
Predictive Analytics: What is going to happen?
We’ve graduated from crawling and now we’re learning to walk. Predictive analytics seeks to use mathematical models to figure out what is going to happen in the future. And since no one has a crystal ball, simple regression will do. Our sandwich shop can use predictive analytics to figure out some of the following:
What are my sales going to be for the next few weeks?
How likely is this customer going to return for another meal?
Because no one can predict the future, forecasting will never be exact. The business goal to achieve through forecasting is to get better insight to what we expect to happen in the future while trying to get as accurate as possible. Our sandwich shop was able to create a simple linear regression including day of the week, weather forecast, and quarter of year to predict how many sandwiches will be sold next week.
Predictive analytics can be as simple as a rolling average or as complex as a neural network. They are great tools that learn from the past and can incorporate other variables to help with business planning decisions.
Prescriptive Analytics: What should I do to get the best result?
Finally, we’re learning how to run! After a forecast is generated, prescriptive analytics goes one step further and optimizes on a variety of potential scenarios to recommend a plan of action to get the best result. In the case of our sandwich shop, prescriptive analytics can help answering some of the following questions:
If my forecast for next week is to sell 100 turkey sandwiches, where is the best place to source my ingredients from to save money? What about to increase freshness? What about both?
What marketing mix strategy should I do in order to increase sales by 10%?
The sandwich shop built out a great sales forecasting tool to predict what sandwiches will sell, and now they have built out an optimizer on top of the model which can guide them on which vendors to source from in order to minimize their ingredient costs. Leveraging this prescriptive model is not only helping the sandwich shop plan for the amount of sandwiches it will need to make, but to also better manage the costs from vendors.
Prescriptive analytics is a great tool to help guide business decision making because it takes into consideration multiple parts of a business problem, but often requires strong processing capabilities since optimization problems are rather complex to solve.
All types of analytics will add value to your organization, whether is a simple product ROI analysis in Excel or a complex network optimization model running in the cloud. Knowing the difference between the types of analytics and their applications will serve as a solid guide in your analytics journey.
I hope you enjoyed this article! If you have any further comments, questions, or feedback feel free to email me at firstname.lastname@example.org!