This is a guest blog post by Meta S. Brown who is a consultant, speaker and writer promoting the use of business analytics. She is a hands-on analyst who has tackled projects with up to $900 million at stake, and is a recognized expert in cutting-edge business analytics. We recently sat down with Meta to ask her a few questions about predictive analytics.
Q: Thanks for spending some time with us today. Let’s start with your views on why companies should be considering predictive analytics.
Meta: I’d never approach an executive by starting the conversation with predictive analytics. Executives want to know “how can I solve this problem?” or “how can I remove the obstacles that are standing in the way of this business opportunity?” It’s much better to understand what business problem that executive wants to solve, and then connect the dots to relevant applications.
Q: Is there a common business problem that predictive analytics can solve?
Meta: Well, math is math. In many cases, business problems can be solved using some kind of math. Predictive analytics encompasses many of those branches. Statistics is the mathematics of uncertainty, so predictive analytics can be used to address business problems where there’s an element of uncertainty.
For example, say the business problem is that the company wants to sell more product. Look for opportunities to know more about individual potential buyers. Look for indicators on who would buy the product, how much they would buy, and when. Relevant data and methods will help answer that question. And it doesn’t have to be a perfect answer – in real life there are no perfect answers. But you can get the information you need to figure out how to make more money.
Q: Are there specific types of businesses for whom predictive analytics provides a big advantage?
Meta: There are more obvious matches than others. Any business selling directly to a large number of prospects is almost certainly a very good match for predictive analytics. This is a classic scenario of asking “will this person take an action?” That action can be buying a product, requesting a brochure, asking to speak with someone for more information, or recommending a product or service to their peers. The business wants to better understand how it can better predict and influence that behavior. Multiply that incremental difference times many interactions and many people, and you can yield significant returns.
Q: Are businesses required to have a data scientist or other analytics expert in-house?
Meta: If you are a large business, it does make sense to have your own analytics talent in-house. However, I would never tell anyone to be thinking in terms of the data scientist job title or the ideas that come with that today.
Instead, I recommend they look at what talent they have in the organization and what talent they can develop. Focus on capabilities that are relevant and integrated with the context of everyday business.
There’s nothing inappropriate about hiring in new talent, but you can’t expect someone who’s disconnected with the everyday business to provide you with analytics value. You need subject matter experts on the team.
In real life, companies that are successful with analytics often develop the analytics competency within people who may not be viewed as data analysts. Yes, you need people to be trained on analytics, but that’s something any intelligent adult can do. When I was a software trainer I worked with an organization that lost their full-time data analyst. The boss told a secretary that she could take over that function. She knew the business very well, was methodical in how she did things, but she did not have a college education. They sent her off to training and she became a very competent data analyst. The takeaway is that organizations should draw their analytics talent from a wide pool of people. Don’t get wrapped up in the data scientist title or the outrageous skill mixes that are often associated with that title.
What I see in the data scientist community is a school of analytics thinking that has grown out of the computer science programming/IT world – folks tend to use too much stuff to solve a problem. The quantities of data, complexity of hardware and software, and amount of labor used to solve problems is often excessive for the business purpose. That’s the nature of their training and what they value.
I once worked with a software firm that had a gentleman who took great pride in having what he thought was the world’s largest cluster of a particular platform. His pride in that drove the way he did everything – resulting in spending millions of dollars more than it took to solve the business problem at hand.
This is a prime example of what I see going on in the data scientist community – they are too focused on the technology and not focused enough on effective problem solving.
Q: If you had to boil success with analytics down to one thing, what would that be?
Meta: Start with the problem and work backwards to find the solution.
Q: What do you see as the biggest innovation five years from now? What will have the biggest impact?
Meta: Nobody, really nobody, is taking full advantage of what’s already available today. They should be looking for opportunities to improve, and make better, more consistent use of what they already have.
However, take a look at what’s in current academic research, and that will give you a hint of what’s to come. What I see as very promising is video analysis; I expect it will be in use within a five-year horizon. It won’t eliminate human involvement, but it will make people more productive, and enable them to perform analysis more easily, faster and more effectively. It will have a great impact in some businesses – most obviously security and law enforcement. But it will also trickle into other uses, such as consumer research – for example, video analysis can be used to analyze how people move through stores.