Eric Siegel, founder of the Predictive Analytics World conference series and author of the award-winning, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die, joined me to discuss the hows and whys of predictive analytics. In an uncommonly approachable manner, Eric actually managed to make predictive analytics both accessible and captivating (and quite amusing too).
I started off by asking about his out-of-the-box educational rap music video, “Predict This!” (see www.PredictThis.org for 3 ½ minutes of informative fun). Enjoy the full interview below!
MG: So, first off, who writes your rap lyrics and how did they manage to find a rhyme for “pedagogical”?
ES: That’s obviously the most pertinent question that I would have expected to receive today, so I’m glad you jumped right to the meat of the matter. I wrote the song lyrics. The song is educational, but I feel people don’t pay enough attention to the lyrics because they get distracted by the kookiness of the music video. But be sure to pay attention to the educational message.
MG: I typically turn on the captions, so was glad to see them in this video as I was indeed following along.
MG: Ok. So, to really start, can you give us an overview of predictive analytics?
ES: Let’s start by distinguishing predictive analytics from forecasting. Of course, both are ways you’re using quants or math to predict the future. The difference is that, with predictive analytics, you’re predicting for each individual. In most cases, for sales, that would be each individual lead or customer. Whereas forecasting would be trying to determine an overall estimate across leads. “How many deals will we have next quarter?” would be forecasting. “Which individuals are most likely to convert?” would be predictive analytics. The main bottleneck to applying predictive analytics is the availability and accessibility of data. There’s lots of hype around data and big data, but data isn’t necessarily complicated in and of itself: it’s an encoding of experience across the organization, a list of prior events. It serves as experience from which to learn.
Learning from data to predict individual behavior is more actionable than any other kind of data analysis. And, when applied to sales, predictive modeling addresses a kind of “information overload”: The model’s outputted predictions help you triage and prioritize a mountain of leads. In general, a predictive question such as “Will this lead convert?” can be answered by the model. It approximates the answer to this predictive question, for each individual, by learning from historical data.
MG: Through your videos and writing, you seek to make data and predictive analytics more accessible. As it applies to businesses, where do most companies fall on the adoption scale in terms of understanding analytics and putting the full range of their capabilities to use?
ES: Overall, I would say that the technology is mature, its potential is there, and lots of companies are very advanced. But we’re still at the early phases of adoption. Just because companies are advanced doesn’t mean they’re far along in exploiting the potential. We have to look at it beyond just a core technology – it’s actually an organizational endeavor. You can’t just do the analysis and create the predictive scores. You need to act on them. For example, salespeople need to be trained on the meaning of scores and how to use them. Because of the way predictive analytics influence an organization, it’s no longer business as usual; you’re literally changing operations, the most wide-scale operations. So, I’d say that while the technology is mature and valuable, there’s still a lot more potential than is being realized – even within the most advanced enterprises.
MG: What’s the more valuable application of predictive analytics in business? Is it predicting what customers will do/buy, or does its deeper value lie in internal use and improving your organization?
ES: Both are critical for the organization, and there’s value to using the data to improve both sides. As far as which has greater potential, in general I think predicting leads does. There are more of them than there are internal employees, and more data about them that you can use to optimize the lead prediction process, prioritizing and flagging those leads who are most likely to convert. Whereas, with workforce applications, you’re dealing with smaller numbers – employees or staff members. Still, there is a huge amount to be gained there. In fact, we have an entire predictive analytics world conference every year focused solely on HR analytics, called PAW Workforce. Predicting which application is more valuable is sort of guesswork. It definitely makes sense to do both.
MG: In your book, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die, you use analytics to make some interesting correlations, such as one between ice cream and shark attacks. Any specific to sales? Or that stand out to you?
ES: In terms of entertainment value, there’s one that says people who buy little felt pads so they don’t scratch the floor are a safer credit risk, and people who go to bars are a higher credit risk. Areas of San Francisco that have higher instances of crime also have more demand for Uber rides. The diaper and beer one is a classic. Pop-Tart sales spike before a hurricane. More expensive hotels are booked by Mac users. So, there are lots of entertaining connections. There’s also a brand new one that says conservative politicians tend to be more attractive. I summarized all of these in one table in Chapter 3, and within that table there’s a column with suggested explanations as well. These things are connected and correlated, but that doesn’t mean there’s a direct causative relationship; the explanation for each connection is always just conjecture. The causative answer to “why” that people come up with is not scientifically conclusive.
MG: When you advise companies on data use, what’s something they should avoid? What are the pitfalls?
ES: Well, actual technical pitfalls are relatively arcane. There are always possible bugs, but you can mitigate risk in deployment. But the main pitfall turns out to be non-technical. It’s organizational. Often there’s lots of enthusiasm around diving in and triggering a project, but not enough of an organizational process to determine how it’s going to be used, what value it will add. Organizations that are churning out predictive scores need to have established the operational buy-in to realize those scores’ value in the field. Depending on what you’re actually trying to do, either salespeople will act on them, or other processes will be influenced. You’ve got to get the right buy-in and understanding of why you’re doing prediction and how it’s going to affect organizational operations. Otherwise, you might do a bunch of analysis and it’s DOA.
MG: You are a leader and influencer in your industry. We’re curious to find out who influences you and why?
ES: A great new example of the kind of popularly appealing coverage of data analysis is Seth Stephens-Davidowitz, who recently published, Everybody Lies: Big Data, New Data, and What the Internet Can Tell Us About Who We Really Are. He will be speaking at our Predictive Analytics World conference, October 29 – November 2 in New York (www.pawcon.com/newyork). Regarding ethics in data analysis, a great read is Stephen Jay Gould’s The Mismeasure of Man.
You can find out more about Eric Siegel’s leading conference series at Predictive Analytics World. Be sure to also take a look at his book, Predictive Analytics: The Power to Predict Who Will Click, Buy, Lie, or Die, at www.thepredictionbook.com.
Also, check out our expert interview series.
For the Win: How to Successfully Use Gamification in Sales Operations
Salespeople love to win and sales organizations are the perfect environment for gamification. Download this guide to learn how gamification can improve sales performance.