Listen Now

Large companies today are quite happy to use analytics for a wide range of tactical decisions, such as product assortment and identifying the most efficient distribution channel. But when it comes to bigger strategic decisions, C-level executives are more tentative. Their gut judgments are still their preferred decision-making mechanism. David Dittman is the director of business intelligence and analytics services for Procter & Gamble, which reported $65.1 billion in revenue last year. Dittman argues that C-levels must start using analytics for more strategic decisions, and they must do it now. David, why are so many executives so hesitant to trust analytics for the big decisions?

IT professionals, in general, are excellent at answering direct questions. Volunteering information that the questioner doesn't know about.... well, that's very different. That is one of the reasons why marketers tend to get frustrated when asking analytics questions of their IT partners. For some on the IT side, a datalake is a very easy and convenient answer to almost any analytics question. Todd Cullen is the Customer Insight & Analytics Practice Leader at KPMG. Todd argues that marketers need to get up to speed on these issues so that they can ask better questions. And therefore get more useful answers.

Few things in business analytics have gotten the excitement and, yes, the hype as Machine Learning. It fires up the imagination of every sci-fi loving data scientist, with images of true thinking computers adapting and learning new tasks on their own. The good news is that Machine Learning works and it can indeed do all of that. The bad news is that it can’t do those things as well, as fast or as easily as the hype suggests. There are quite a few things that Machine Learning can do quite well, such as signal identification and noise removal based on both structured and unstructured data. But there is also a list of things that Machine Learning in 2017 CANNOT do, things such as automated complete data cleaning. Another thing Machine Learning can’t yet do is replace humans. And yet, far too many data scientists today insist on using Machine Learning in areas where it makes little sense. And that frustrates today's guest. Jitendra Papneja has worked in analytics for multiple Fortune 100 companies and today serves as an analytics leader for a Fortune 100 consumer goods company.

Direct Marketing focuses on potential customers that are predicted as likely to buy. But is that in fact the best approach? Is it not a better approach to instead target those who are the most likely to be persuaded to buy? Is the point of a marketing campaign to increase purchases or to change minds—and THEN increase purchases. To figure this out, we have with us Eric Siegel is the founder of the Predictive Analytics World Tradeshow.

Predicting cashflow is certainly common enough, but what about predicting the costs ASSOCIATED with cashflow? What are the specific financial impacts of moving from check to ACH in terms of speed and costs? Should you offer a discount to encourage preferred payment methods? And, if so, how MUCH of a discount can you justify? Also, what are the costs of matching invoices and fixing invoice typos? There clearly IS a cost, but have you ever calculated precisely what it IS? One person who HAS is Rodney Gardner, head of global receivables at Bank of America Merrill Lynch. Rodney, how many companies are blissfully UNAWARE of these costs?

When a consumer walks into a grocery store … buys some apples … and pays with a Visa credit card, …. Who owns that data? The grocery store? The shopper’s bank? The retailer’s bank? Visa? The processor in the middle? The shopper herself? All of the above? Payment companies today can crunch that purchase data a million different ways, but they struggle with who has the right to leverage it? Then there are regulatory issues. Payment rules such as PCI put restrictions on security handling of the data and when it can unencrypted so it can be crunched. And new European rules such as GDPR impacts companies who touch Europe in any way. GDPR is cracking down on data retention.

When Human Resources executives discuss analytics, topics turn to employee evaluations and various indicators of happiness levels. But for Roy Altman, the manager of HR Analytics for the Memorial Sloan Kettering Cancer Center, he wanted to understand how employees work together. He saw an almost total absence of metrics about how employees mesh together. All of the analytics explored employees as individuals, never as part of a group. Altman, who oversees a wide range of personnel analytics for the New York metro cancer treatment facilities, thinks there has to be a better way. Roy, what are the practical HR analytics obstacles to measuring how people perform in a group?

One of the oldest concepts in retail marketing is single view of the customer. That simply means recognizing the shopper as she moves from desktop to mobile device to calling a call center to visiting a physical store. But shoppers share desktop devices with family members and many use multiple mobile devices throughout the day. So how accurately and reliably can those shoppers indeed be tracked? Into this effort goes analytics and probabilistic identification. Jeff Rosenfeld, the VP of analytics at retailer Neiman Marcus, said his chain is doing quite a bit of consistent identification online and mobile. As for call centers and in-store, not so much. Jeff, with the ability to track mobile devices in-store—especially if they ride on your store’s Wi-Fi—so well-established, why aren’t you using that?  

All day long, mobile devices are being bombarded with shopping offers. Marketing analytics routinely explores which offers are likely to be accepted and which will be ignored. But does it make a difference if that sales pitch arrives when shoppers are sit

Outsourcing sales and customer support to a third-party is a common tactic for growing companies, but it can be quite unnerving. Are your customers and prospects being properly taken care of? Are your renewal and acquisition rates lower than they would be if your internal people handled it all? In one sense, that’s a classic what-if query. The problem is it’s almost impossible to know what customers and prospects would have done differently with different people. Infusionsoft, a Goldman Sachs backed company, tried an intriguing way to make those projections. They dubbed it a Customer Health Score and it looks at the number of logins, contacts added, tags created, appointments sought and more than a dozen other variables. With us today to better understand this process is Hal Halladay, the chief people officer at InfusionSoft. Hal, how does exploring those metrics tell you what the customer or prospect would have done differently with different people?  

The dictionary defines optimal as best, as in “the optimal approach is the best approach.” But when it comes to analytics and making the right decision, Jeff Camm argues that optimal is very often far from optimal. Camm is the associate dean of business analytics at the Wake Forest School of Business. Camm points to misplaced confidence in algorithms. In his research, he found that the best analytics recommendation for most enterprises is almost NEVER the choice that is mathematically ideal. Jeff, in a world of almost pure science, how can that be?  

Interview with Barry Devlin, recognized as the Father of the Data Warehouse. In this podcast we discuss the role of data preparation, data cleansing, and the boring mundane stuff and why the boring stuff matters. A lot. In just about every business analytics project, there are the fun parts and the drudgery parts. Most data scientists want to plunge in with the algorithms and the complex modeling, while rushing through mundane production tasks, such as data cleanliness and data governance. In short, they want to jump to the exploratory areas while blowing past production issues. But, like almost everything else in business, there’s a cost for avoiding the boring details. Barry Devlin, author and consultant with his 9Sight Consulting firm, has been leading in data warehouse work since 1985. Devlin argues that these corner-cutters are increasing how long projects take to get done and their financial cost.

Incompatible databases bedevil every company’s analytics efforts and healthcare companies are no exception. The $60 billion Kaiser Permanente managed care company is exploring whether it’s found a way around those incompatible datasets courtesy of an approach developed at the University of Oxford. It’s dubbed the Resource Description Framework Oxford and nicknamed RDFOX. Alan Abilla is the chief of medical informatics and innovations for the Convergent Medical Terminology team within Kaiser Permanente

Nowhere is crunching numbers more revered than in a sales team, with all manner of compensation tied to deals closed and revenue booked. But, bizarrely, sales analytics today often tracks the wrong elements and thereby fuels wrong conclusions. Kevin Styers, the director of sales operations at ShopKeep, has spent years figuring out old sales

How efficiently is your time used in meetings? If you needed to make decisions on the number of conference rooms you need make on an analysis of scheduled meetings, how accurate would it be? For architectural firms, this is crucial. Are rooms being created that aren’t needed? Think about calendar listings. When the company shuts down a project that held team meetings every Wednesday morning, does anyone bother to delete that meeting? Dal Adamson is a product manager at Teem and he has been trying to fix the meeting management nightmare so companies what they do and don’t. Exactly how much time do we waste in meeting that don't help the company? Listen in to hear the answer.

In the hospital business, few financial arrangements are as critical—and fraught with complexities and red tape—as insurance payments. Other than deep discount on the rates, the biggest headache is payment speed. Yes, say insurance cashflow to any hospital CFO and watch the involuntary shudder. In Philadelphia, executives at Thomas Jefferson University and Jefferson Health, found a way to use analytics to accelerate chemotherapy insurance payment authorizations from 22 days to five days. And this reduction in insurance payment latency also increased the speed of patient care - all through the use of data analytics. Learn how they did it.

When former Microsoft CEO Steve Ballmer rolled out his extensive government database in April, it marked the first time much of that data had been easily available to the public. And for USAFacts, the word "easy" is key. The $10 million effort attacked some 30 years of data from 70 municipal, state and federal agencies. The goal was to make the mountains of information easily accessible by the lay public. In many respects, that's the essence of data analytics. It's about helping users spot the patterns and deviations from a massive amount of numbers. That requires understanding the user—and the questions the user is likely to want to ask—as well as the data. Turning back to Balmer's USAFacts effort, did it in fact make those terabytes of data easily accessible to the public? Or did it needlessly limit that access to a handful of categories that its developers thought up? To figure this out, we're turning to Dave Johansen …. VP of engineering at Numetric.

Imagine two restaurant chains (one pizza and the other ice cream) that recently had to fully redesign their data warehouse. One of them had more than 30,000 codes for coupons—and there were really only about 21 different coupons. Instead of changing the coding for existing coupons, they kept creating new ones, which were worded SLIGHTLY different. In the end, it made effectiveness and analysis worthless and it cost them thousands of person hours. This is a case where business intelligence wasn’t intelligent at all.

There’s an old-school analytics thought that the more data that is analyzed, the more accurate will be the results: the answers, the conclusions, the recommendations. But there’s also the signal to noise ratio issue, which holds that the more irrelevant data that is being examined, the more difficult and time-consuming it is to find those answers. And that brings us to the Petabyte Problem. Today’s companies are collecting exponentially more data than a few years ago. Even worse, much of that data is happening in hidden corners of the company, such as on mobile devices and in the cloud. End result? It can makes analytics far more challenging. Indeed, it can make knowledge management a nightmare. Put simply, companies no longer know what they know—and are therefore condemned to repeatedly and expensively solving the same problems over and over. Michael Stevens is the chief operating officer at AccessData. Michael joins us today to try and figure out a way around the Petabyte Problem.

A leaked document showed that Facebook targeted audience members that felt "overwhelmed", "stressed", "anxious", "nervous", "stupid", "silly", "useless", and "failure" with relevant ads that fit their emotional profile. We talk with Ron Guymon, Chief Data Scientist at Numetric about the technical aspects of data science like this (no moral discussions or judgment, just technical).