Dr. Athanasios (Thanos) Gentimis, Assistant Professor of Math and Analytics at Florida Polytechnic University sees the future of analytics as a team effort, where subject matter experts (SMEs) collaborate in teams with Data Scientists and each team member plays to his or her strengths. He also sees a reliance on Deep Learning for creation of relevant questions and answers about data.
Dr. Gentimis’ field of expertise is Algebraic Topology, with a special interest in Geometric Group Theory. In a recent DATAVERSITY® interview, Dr. Gentimis discussed how analytics is changing and what the future might look like.
Today’s Data is Multi-Dimensional
“If you go back into the 80s, analytics were performed on small data sets where somebody would come up with a survey,” says Dr. Gentimis. “You would ask about 1000 people, and then try to extrapolate to give you some information about the society at large.”
It’s a process that involved starting with small samples and attempting to extrapolate for the whole population. This process required more mathematical intuition than computing power, and the tools used were created for that specific purpose. Simple questions like “Who is the average Amazon customer?” have no meaning with the sheer volume and variety of data available now. It’s no longer possible to take a few data points and draw significant conclusions, he says.
The presentation methods need to be more sophisticated now that the breadth of data is also more sophisticated, so using simple graphs or PowerPoint presentations, he says:
“That’s not going to cut it anymore. The complexity of those data sets, as you start doing things in more than one dimension, you have to think of products now that have four or five or six attributes, so that doesn’t fit anymore into a nice presentation.”
While simple presentation methods are no longer able to adequately show the story within the data, business executives still need that story to use data effectively for predictive and prescriptive purposes. A 2015 DATAVERISTY research report by Kelle O’Neal and Charles Roe entitled Business Intelligence versus Data Science, found that nearly half of respondents cited “improved understanding of customers” as the primary reason for investing in Data Science:
“Better decision-making and improved understanding of business results ranked as the most important measurements of Predictive Analytics/Data Science program effectiveness. Increased operational efficiency, better impact analyses, cost reductions, and more accurate forecasting were some of the top outputs that senior executives want to see from analytics .”
Making Meaning from Data: Telling the Story
How is it possible for the Data Analyst to share complex data with end users in a way that has meaning? “You have to collaborate with the SMEs,” Dr. Gentimis says, and that may require some added communication and social skills. He sees positions like Data Analyst now requiring collaboration or team building skills, so that a group that collectively has a more varied skill set “can take the knowledge of the market and see how it is applied.”
Dr. Gentimis confesses that, as a mathematician, “I can tell you what the data says, but I cannot explain what it means, so I really need to partner up with a SME and create the story.” Although it’s certainly possible for an individual to have the perspective and expertise of the end user as well as the analyst, he says, the “best way to do things would be to partner them up, create teams with analysts” and SMEs.
Knowing the story behind the data allows a company to understand what happened in the past (Descriptive Analytics), make projections about the future (Predictive Analytics) and make plans to influence future outcomes (Prescriptive Analytics). Although 80% of respondents in the DATAVERSITY study said that they plan to invest in Predictive Analytics over the next five years, more than 40% remarked that their biggest concern about investing in a Data Science program was unclear benefits or outcomes. They don’t necessarily know how to tell the story and/or what the story actually means. That is the point of the collaboration between Data Analysts and Data Scientists.
Terminology Can be a Barrier to Understanding
Some of the confusion about the value of analytics may stem from a lack of understanding about the terminology used. There are differences in perception about common terms used, and according to O’Neal & Roe “definitions are all over the map.”
They provide some helpful definitions, paraphrased here:
- Data Science is the process of organizing data for analysis – including measurement, collection, cleaning, application of tools, and analysis – while Business Intelligence is the use of all that data for business decision-making processes and reporting.
- Business Intelligence takes raw data stored in Data Warehouses and Data Marts and transforms it into useable knowledge/ information assets.
- Descriptive Analytics answers the questions, “What has happened?” or “Where are we right now?” and includes standard end-user reporting practices such as ad hoc reports, basic queries, dashboards, graphs, and charts.
- Data Science allows enterprises the ability to turn their data assets into a narrative. Descriptive Analytics certainly allows the telling of a particular type of data story, but one with only a specific temporal viewpoint, where the story answers a singular data question. Data Science allows that narrative to be expanded across timelines, in different data spaces that trace from the past into the future, with much more involved questions and answers.
- Predictive Analytics essentially asks the question “What could/might happen?” or “Where will we be?” using forecasting, decision analysis, text analysis, transaction profiling, trend lines, sentiment analysis, optimizations, geo-location data, Machine Learning, Natural Language Processing techniques, and numerous other approaches to provide answers to “possible” outcomes.
- The purpose of Prescriptive Analytics is to answer the question, “What should happen?” to reach a desired future outcome. There is some debate about whether Prescriptive Analytics is simply a subset of Predictive Analytics, but its purpose is the same regardless.
On the Horizon for Analytics with Experiential Learning
Another key to clarity about the value of Business Intelligence may be in a change of perspective gained by the experience of working with analytics, according to the analytics software company, River Logic:
“As companies progress their use of Business Analytics, they derive exponentially more value – especially as their orientation shifts from historical to forward-looking. With this frame of mind, they are able to leverage Prescriptive Analytics to translate data into actionable, optimal plans.”
Dr. Gentimis sees the next logical step as Deep Learning. He illustrates Deep Learning as: “you train a machine or a robot to just do whatever you want it to do, without having external help to begin with, [and] afterwards, you have some sort of supervised learning.” He credits IBM with being a pioneer in Deep Learning by designing algorithms, then feeding machines big datasets and teaching the computers to figure out what questions to ask.
He sees the next stage of Deep Learning as a “meta” stage, “where the algorithms themselves are going to be generated automatically.” He uses Facebook as an example of this stage, because Facebook finds and promotes associations between people, using algorithms that “went beyond whatever the original people intended. And there are new rules being created by the computer itself.”
Analytics as an Essential Tool for Success
Whether the approach is to pair Data Analysts with SMEs, to gain perspective by working with your data over time, or to use machines as partners in the story-making process, a smart business can’t afford to be without a BI/A program. According to O’Neal and Roe:
“The establishment of a comprehensive BI/A program that includes traditional Descriptive Analytics along with next generation categories such as Predictive or Prescriptive Analytics is indispensable for business success. The competitive advantages realized from a dependable Business Intelligence and Analytics (BI/A) program are well documented. Everything from reduced business costs and increased customer retention to better decision-making and the ability to forecast opportunities have been observed outcomes in response to such programs. The implementation of such a program remains a necessity for any growing or mature enterprise.”