Making Data Work Harder for You— Podcast Transcript

[Music prelude]

Fred SCHMALZ: Hello, and welcome to Insight in Person, Kellogg Insight’s monthly podcast, produced by the Kellogg School of Management at Northwestern University. I’m your host, Fred Schmalz.

This month we take a look at how business leaders can incorporate data analytics into their companies—from why managers need to be involved in the data analytics process to how big data can be used in product development. In the first half of the podcast, we hear about how a working knowledge of data science can help business leaders get ahead and manage with confidence.

In the second half of the podcast, we look at how data analytics can tell us who’s buying our products—and whether they’re the kind of customer that might be a harbinger of failure for those products.

So stay with us.

[Music interlude]

ACT 1: Florian Zettelmeyer on why a working knowledge of data analytics is necessary for business leaders

Florian ZETTELMEYER: One of the really fascinating developments in our world is that it has become easier to measure things.

Fred SCHMALZ: That’s Florian Zettelmeyer, a professor of marketing at the Kellogg School and faculty director of the program on data analytics at Kellogg.

ZETTELMEYER: And because it’s easier to measure things and it’s easier to visualize things, it is also easier to present such information directly to managers and directly to decision makers.

SCHMALZ: Zettelmeyer says that in today’s data-driven world, managers need to think of data analytics as part of their repertoire, rather than something that falls outside of their domain.

After all, leaders are the ones who will have to make strategic decisions based on the data, so they should be the gatekeepers when it comes to deciding how data can add value to their business.

ZETTELMEYER: They now need to have a level of awareness of what you can and what you cannot learn on the basis of the data that you observe—which, frankly, was not a skill that has traditionally been trained for a managerial class but was a skill that was the domain of the data scientist or the analytics people.

SCHMALZ: For the data analytics process to be effective, Zettelmeyer says, managers need to be involved every step of the way. In order to do this, they need to have what he calls a “working knowledge” of data science.

So what exactly is a working knowledge of data science? Zettelmeyer says the necessary skills are not learned in an advanced econometrics class. In fact, they’re not really technical skills at all—they’re thinking skills.

ZETTELMEYER: You don’t really need to be great at math or at stats or computer science to obtain a working knowledge. I’m not saying that it is super easy to get it, but it is really a way of disciplined thinking that needs some practice but that fundamentally does not need a lot of technical background. Therefore, it is fundamentally acquirable by people who have succeeded in the business world, precisely for the fact that they are good analytical thinkers.

SCHMALZ: These thinking skills—and this discipline—can save managers from blindly trusting data scientists by assuming that just because they have the numbers, the analysis was performed in a reasonable way.

To do analytics right, you need more than data—you also need to draw upon an in-depth knowledge of your business, which is exactly the kind of domain expertise good leaders bring to the table.

ZETTELMEYER: This is something that often data scientists are not so good at finding. But it’s something that managers are excellent at finding. It really speaks to this idea that you get both very good data scientists as well as very good business people involved in analytics because that’s how you make progress and avoid mistakes.

SCHMALZ: Creating a culture where analytics add value to a business goes well beyond just putting together the right analytics team. Managing analytics across the organization is key.

ZETTELMEYER: Many of the core barriers to making analytics work have actually nothing to do with analytical problems per se. They are not about picking the right algorithm; they’re often not even about picking the right data.

What they are about is that, when you do analytics, you run into organizational and incentive barriers that prevent you from actually doing the analytics or that prevent you from acting on the analytics. For example, imagine that you want to get a 360 degree view on the customer, which is a big mantra in analytics. Well, the only way you are going to do that is if you can get all the business units to actually cooperate to provide data so that you do get a 360 degree view of the customer.

SCHMALZ: In other words, if a company’s business units aren’t in the habit of working collaboratively, using data science can quickly become a source of organizational tension. Performing analytics, like many aspects of managing an organization, is fundamentally political.

ZETTELMEYER: There are winners and there are losers from implementing analytics. There are people who have different agendas. Before long, if you are a manger, you will find yourself in a situation where you have team A reporting back to you with result A backed up by data science team A.

You have team B reporting back to you with result B backed up by data science team B. Guess what? Result A and result B are not going to be the same. The problem is that you can’t really go out there and say, “Let me get my data scientist in order to somehow figure out which of these two teams is right.” Because there are already ten data scientists in the room and they all disagree with each other.

SCHMALZ: Data analytics cannot be treated as a separate part of one’s business—it has to be done with organizational challenges in mind, and it has to be incorporated into the business plan itself. Too often, Zettelmeyer says, managers decide to collect data without knowing exactly how they will use it, and this leads to problems down the road.

ZETTELMEYER: You have to think about the generation of data as a strategic imperative. You can’t just go out in a company and somehow hope that the data that is available to you and the data that gets incidentally created in the course if business is the kind of data that’s going to lead to breakthroughs or the kind of data from which you can learn something about your business.

Analytics has to fundamentally start with business problems. I can’t tell you how often I’ve heard the sentence by an executive who comes and wants to talk to me and says, “You know, we have this enormous amount of data. We are overwhelmed by it. We don’t know what to do with it. But we are sure there’s something incredibly valuable in there.”

SCHMALZ: Whether such data can add value depends on what kinds of questions are asked, how the data are generated, and how the leadership makes decisions based on its analysis. When analytics goes bad, it usually does so because the insights are nonscientific, however much they appear to offer.

For Zettelmeyer, the way to do analytics well is to start with the problem and incorporate the analytics into the business plan itself. It also means empowering everyone in the organization to use their analytic skills to interpret and question the data

ZETTELMEYER: If we want big data and analytics to succeed, everybody needs to feel that they have this right to question. There has to be a culture where you can’t get away with thinking as opposed to knowing.

SCHMALZ: For Managers, having a working knowledge of data science has three main benefits. First, it helps you judge what good data look like. Second, it helps you identify precisely where analytics can help the business. And third, it helps you lead with confidence. Increasingly, having this knowledge will become the norm for all good business leaders.

ZETTELMEYER: Can you imagine a CFO going to the CEO and saying, “You know what? I don’t really know how to read a balanced sheet, but I have somebody on my team who is really good at it.” Okay, we would laugh that person out of the room today. Now, I hate to tell you but I know a whole bunch of people in marketing and other areas at the C-suite who without blinking an eye would go to the CEO and say, “You know what? This analytics stuff is complicated. I don’t have a full grasp on it, but I have assembled a crackerjack analytics team that is going to push us to the next level.”

I think this is an answer that is no longer acceptable.

[Music interlude]

ACT 2: Eric Anderson on Harbingers of Failure

Fred SCHMALZ: No matter how often you go to the supermarket, you are likely to see new products—from breakfast cereals to tasty yogurt flavors to household cleansers—competing for shelf space with established brands.

Yet new products tend to fail at very high rates.

Which begs the question, how can so many products make it through the development process and then flop once they land in stores?

In managing the introduction of new products, it turns out that predicting their success or failure doesn’t just depend upon whether lead users buy those products, but which lead users buy them.

Professor Eric Anderson, chair of the marketing department at the Kellogg School, wanted to know more about why products failed and how to identify those products early in the process.

In a recent study, Anderson and his coauthors found that the presence of certain buyers of a product may indicate that the product is likely to fail. And they found that these “harbingers of failure” tend to buy a lot of these failed products.

Here’s Eric Anderson.

Eric ANDERSON: What spawned out of this was a fairly creative idea about how to think about success and failure of new products. We started thinking about consumers who might systematically buy either winners or losers. Products that are winners are ones that survive on the shelf for a long time, things that you see all the time, like Hellman’s mayonnaise. Losers are products that you don’t know that well because they disappear very quickly—products like watermelon Oreos that are on the shelf for a short period of time and then quickly disappear.

SCHMALZ: When researching winning and losing products, Anderson found that there are indeed customers who have a systematic preference for products that tend to fail.

ANDERSON: It turns out that the more you buy of these failed products, the more likely you are to buy another failed product. That really reinforced the notion that these customers were behaving systematically.

We can talk about customers buying a soda product and systematically buying failed soda. Then we can show that this helps us predict in other categories, like household cleaners.

SCHMALZ: So who are these customers? Are they people far out on the taste spectrum, whose grocery carts are loaded down with raspberry beer and kumquat custard? Are they young people with adventurous, if curious, dietary habits? Millennials with little of the brand loyalty of their predecessors? Are they customers gravitating toward the cheapest products on the shelf?

The answer to this question surprised Anderson and his coauthors.

ANDERSON: What we’ve discovered thus far is in the packaged-goods world, these customers tend to be fairly educated, fairly wealthy, and have large family sizes.

SCHMALZ: What’s curious is that one would assume affluent people with larger families would tend to spend more at the supermarket, so intuitively, if customers buy more of something, that’s a good thing, right? Not necessarily.

ANDERSON: The twist is that the more you sell to customers who are these harbingers of failure, the more likely you are to fail. That’s what is the surprise. The more you sell, the more likely you are to fail when those sales are going to customers who buy Diet Crystal Pepsi.

That’s what people find interesting and shocking, that you can actually predict new product failure. That insight is what opens up all sorts of new possibilities for future research.

SCHMALZ: From a data-analytics standpoint, the consumer goods industry is an excellent market to test for harbingers, because, through supermarkets’ customer-loyalty programs, granular information is collected on who buys what, which helps determine success or failure on a micro level. But as we heard earlier, having more information is not automatically going to lead to insights.

ANDERSON: A misconception is that what you need to create value is lots of data or perhaps tremendous horsepower with respect to analytics and coming up with the next best algorithm for processing the data. Here neither of those actually leads to value.

What drives value is what I’ll think of as creative insight. It’s the spark of an idea that thinks about the world in a completely different way and doesn’t take the status quo as given. It pauses for a second and says, hey, can we think about things differently. Here, the question we asked, where we spun everything upside down, is, look, the new product-development process always takes positive feedback as a measure of success.

Each one of these measures tells me, yes, keep going, keep going. What we point out is, here’s a measure of success. You sell more to these harbingers of failure, and it’s completely negatively correlated with actual market performance. That’s what twists things and throws it upside down.

Much of the value is locked into what kinds of questions you ask. It’s very difficult to get new insights without new questions.

SCHMALZ: Information about the presence and purchasing habits of harbingers of failure can help companies improve the new product development process to give those products that do appear a greater chance of success. The key is to develop metrics that allow firms to classify product-testing respondents into harbingers and non-harbingers as early in the process as possible.

ANDERSON: By understanding that difference early on, you can start to make the appropriate investments. You can align your investment in the product and how you’re positioning it with your retail partners to say, this is serving a different purpose. This product might be a niche product and is serving a different goal rather than being a mass-market product.

SCHMALZ: The reason these products fail is not because they are inherently bad ideas. It’s perhaps that they were incorrectly positioned as mass-market products when they may have been niche products. Data analytics can help determine that product positioning, both in the supermarket and wherever people make purchases.

ANDERSON: If we have that kind of data, I think we can start to explore, does this extend into home electronics? Does it extend into appliances? We think it does. I think my conjecture is that it will extend into other industries, but hopefully it will lead to new insights and new discoveries along the way as we start to explore other categories.

[Music interlude]

SCHMALZ: This program was produced by Drew Calvert, Fred Schmalz, Kate Proto, and Michael Spikes. Special thanks to Kellogg School of Management faculty Florian Zettelmeyer and Eric Anderson.

You can stream or download our monthly podcast from iTunes, or from our website, where you can read more about Florian Zettelmeyer’s and Eric Anderson’s research and get tips on data analytics for managers. Visit us at insight.kellogg.northwestern.edu. We will be back next month with another Insight In Person podcast.