Grays Sports Almanac. For many of you, especially those in your 40s like me, that name probably rings a familiar bell as the book played a crucial role in “Back to the Future Part II,” when Marty McFly (Michael J. Fox) travels 30 years into the future to help his future son stay out of jail and along the way stumbles across the sports almanac. Listing the result of every major sporting event from 1950 through 2000, Marty instantly recognizes the value such a book holds and plans to secure his future betting on games he already knows the outcome of once he returns to his own time. The almanac is eventually stolen by antagonist, Biff Tannen (Thomas F. Wilson), who uses it to become obscenely wealthy but also turns their hometown of Hill Valley into a dystopian wasteland. Despite this seemingly insurmountable obstacle, the characters eventually solve their time traveling problems; the almanac is recovered, the world put back in its proper place, and everybody learns an important lesson.
As a kid, I remembered watching “Back to the Future Part II” and wishing I had something similar to Grays Sports Almanac. How cool would it be to foresee future events and plan accordingly? I’d like to think I’d be a little more magnanimous than Biff and not turn the world into something out of Dante’s Inferno, but the idea captivated me. Unfortunately, time travel is not yet possible, so we can’t visit the future and find out exactly how the present turns out. But we can get a glimpse of what the future holds by using analytics.
By now we’ve probably all heard about how analytics are changing the way business is done and that your company will be left behind if you don’t hop on the analytics bandwagon. But simply finding a starting point can feel like an overwhelming process, so I have provided a basic roadmap to help get your organization on the analytics bandwagon.
The Starting Line
First, we need to understand the different categories of analytics that exist and what’s most valuable to you and your organization — especially in the beginning. Lustig3 et al designed a useful classification system for analytics using three broad categories: descriptive, predictive, and prescriptive analytics.
Descriptive analytics examines “what happened” through reporting, visualization and data modeling. It prepares and analyzes historical data and identifies patterns from samples for reporting of trends. Predictive analytics, or “what could happen,” uses statistical and simulation forecasting, as well as data mining and pattern recognition. It helps to predict future probabilities and trends and seeks out relationships in data that aren’t readily apparent. Finally, prescriptive analytics tells you “what you should do” and makes actionable recommendations. It identifies and evaluates potential new ways to operate and targets business objectives, while balancing all constraints.
When looking to implement analytics into your organization, the first question you want to ask yourself is: do you start with the data or start with the decision? Data-centric analytics posits a philosophy of “letting the data speak for itself” and is actualized by gathering as much data as possible before analyzing said data for patterns and trends that lead to insights. These insights are then forwarded to the decision-makers to help them make better decisions. Decision-centric analytics, on the other hand, works backward from the problem. An understanding of the decision that needs to be made is the initial step, and data requirements are determined by the chosen analytical model(s).
The best approach is often a combination of the two. There are definite benefits to digging through your data and looking for patterns and trends, since, without a specific goal in mind, you can easily find yourself chasing information down the rabbit hole and find interesting, but not necessarily beneficial, information. At the same time, if you are solely focused on the end goal, you might generate an analytical model without any data to feed into it.
The INFORMS Analytics Body of Knowledge provides a greater framework off of which to build your analytics system, proposing five manageable tasks to guide you: selecting the target problem, assembling the team, preparing the data, selecting your tools and executing.
Selecting the target problem
First, frame your target problem or determine what it is you’re trying to achieve or accomplish with your analytical initiative. Identify a single or a few primary internal business or operations problems that could be solved or improved. Be sure to include a value proposition when you’re framing your target problem. How will the outcome be measured and how will it impact your organization? The size of the problem is important as well. It should be large enough to provide a tangible benefit to the organization without being so large that you don’t have the resources or data available.
Assemble the team
The first person you’ll need on your team is your executive sponsor. They’ll set the vision and tone for the project, while ensuring the problem you’re addressing is still within your company’s organizational goals. You’ll also want a project manager to manage the schedule, assignments and budget if necessary. A project manager who has more than a passing familiarity with analytics would be beneficial, but not required in my opinion. The next member of your team should be a domain expert from the department or group in your organization that’s directly involved in the problem you’re attempting to solve.
You’ll also want to include an IT expert in your analytics team. They’ll have the necessary systems knowledge to find out where the data you need is stored and how to access it. You will also need a data scientist or someone who understands statistics and mathematics, programming, and algorithms. Depending on the size of your organization, this might be a position for an outside consultant. Finally, you’ll want to ensure you identify and include all relevant stakeholders. It might be that your domain expert is one of your stakeholders, but you want to be sure that all who are involved and impacted by the project are kept updated on its progress.
Prepare the data
Arguably the most important, and almost always the lengthiest, step in any analytical project is gathering clean and accurate data. First and foremost, you need to ensure that the data you’re gathering is accurate. Where are you getting your data from? Is it a reliable source or a manually entered and non-audited value? And once you’ve verified your data is accurate, you’ll want to ensure it’s clean (i.e. formatting data, missing values, etc.). Although these steps are laborious and time-intensive, they are extremely important. There’s a common refrain in data-science: “garbage in, garbage out.” If you don’t have good, clean data, you’re just wasting your time or worse. Making decisions based on incorrect assumptions and or faulty data could potentially be catastrophic.
Select your tools
When your organization is first starting along its analytical journey, it’s likely that your data will be numerical (as opposed to unstructured text, photos/video, etc.) and the work will thus likely be in the form of statistical algorithms. If this is the case, you’ll want any program you select and use to have a breadth of techniques available to the user, including clustering, segmentation, decision trees, time series, classification and regression analysis. You’ll also want to ensure that the analysts working on your program have a background in the program they’ll be using. After all, the best program in the world is useless if no one understands how to use it.
Finally, the best part of the project for many people, including myself, is doing the work and digging into the data to find the answers to your problem. Unless you’re lucky (or selected a really easy problem), your first model will probably not be the final and complete answer. Instead, you’ll often find that your initial data set was not fulsome enough or your model won’t be sufficient for a variety of reasons, and you’ll need to change your approach. As you’re working on your project, you’ll want to ensure that you’re keeping all members of the analytical team and stakeholders updated and in the loop. The last thing you want is to have a fully fleshed out model, only to find out from your stakeholders that the requirements changed, and your model doesn’t provide any benefit.
The above is certainly not an exhaustive guide to a full analytical implementation in your business and probably can’t quite match that movie-magic exactness of Grays Sports Almanac. Hopefully however, it has helped to demystify how an organization can begin to create an analytical culture in their business and set them up for success in the future.