In their recent MIT Sloan Management Review article, “Designing and Developing Analytics-Based Data Products,” authors Thomas H. Davenport and Stephan Kudyba note that a large variety of data products now enhance the consumer experience. LinkedIn’s “People You May Know” feature is one example. So, too, is Zillow’s Zestimate, which uses publicly accessible housing data to predict what price a homeowner might get for the sale of his or her house.
But as ubiquitous as this kind of incorporation of data and analytics into the consumer experience now feels, relatively little has been written about the process of developing these new generations of data products, say Davenport and Kudyba. They’re changing that.
To find out what leading companies are actually doing in the field to create, refine, and generate value from data products, Davenport and Kudyba interviewed data scientists, met with representatives from large companies that are exploring data- and analytics-based products and services (including State Street Corp., GE, Monsanto, the World Bank, Thomson Reuters, and Caterpillar), and interviewed managers at more than 40 companies that had some data product development activities underway.
The result is a model by Davenport and Kudyba of seven steps that companies go through in the development of data products. Taking their lead from a 1996 article by Marc H. Meyer and Michael H. Zack that outlined specific steps in designing and developing information products, Davenport and Kudyba have augmented and updated the Meyer-Zack model. “Data product development activities today are rarely undertaken in a traditional product development sequence that involves identifying the need, developing the product, and then taking it to market,” they write. “Rather, product development activities often take place in a continuous, iterative fashion, with the important activities conducted in parallel.” The sequence also includes a few new steps that they have identified.
On Dec. 1, 2016, Davenport, the President’s Distinguished Professor of Information Technology and Management at Babson College and a Fellow of the MIT Initiative on the Digital Economy, and Kudyba, an associate professor of Business Analytics and MIS at the Martin Tuchman School of Management at the New Jersey Institute of Technology, participated in a webinar expanding upon what they found and how other companies can get into the business of data-based products and services. The webinar was hosted by MIT Sloan Management Review and made possible with sponsorship support from Xively. It was moderated by Steven Paul, a contributing editor at MIT SMR, and highlighted on Twitter at the hashtag #MITSMRevent. Among the speakers’ key points:
Information in the “data economy” is “the new oil.” Davenport said that like oil, information “needs some work — it needs to be refined and distributed and consumed.” Many organizations will want to monetize their vast data assets like companies monetize oil, he said, “and the key way to do that is to create some data and analytics-based products and services.”
There are a broad number of ways for organizations to monetize data. Davenport and Kudyba listed six strategies: selling data directly to customers; selling “data products” with analytics to customers; embedding data and analytics into new or existing products and services; attracting online customers and employing an advertising-based model; providing access to platforms with various types of data; and improving internal business processes with data and analytics.
Much of today’s data economy is centered around using data to help customers make better decisions. Davenport said that where data and analytics have been used in the past primarily for internal company decision making, in the data economy, more organizations are asking how they can develop products and services based on data and analytics that are of value to their customers — taking an external perspective on these capabilities. “A number of organizations don’t yet have the maturity, I guess, to actually charge for these products and services,” said Davenport, “but it’s pretty easy to use them to catch eyeballs — to attract and keep attention. And since many online business models are based on advertising, that can lead to a monetary good as well.”
“The model for this idea of data products really comes from the online industry,” said Davenport. “Google, of course, has a whole host of data products, starting with Search and going into some advertising-oriented ones, and more narrow ones like Google Books and Google Maps and Google Scholar.” Other examples he listed include LinkedIn’s People You May Know and Jobs You May Like features; Netflix’s Cinematch; Zillow’s Zestimates; and Facebook’s People You May Know.
Data products for consumers have also moved off-line, with products such as Google’s Nest thermostat. Other off-line examples include the relationship between Fitbit and health insurance companies, which offer discounts for using the fitness tracker; Adidas’ smart soccer balls, which provide instant feedback on a kick’s power, spin, and trajectory; and Phillips’ smart toothbrush, which tracks how often and how long a user brushes and coaches users on spots that they’re missing.
Data products have moved into consumer services, such as the insurance company Progressive’s voluntary “Snapshot” program. The program, which is available in most states in the United States, tracks and reports a customer’s driving behavior through data such as a car’s speed, braking, time traveled, and distance. “About a third of Progressive’s new customers seem perfectly willing to have their insurance company spying on them in exchange for substantial discounts if you’re a good driver,” said Davenport — the program yielded about $2 billion in premiums in 2013. The sensors also can monitor car movements that could be related to maintenance issues, and could be able to identify, for instance, a car’s need for a front-end alignment. “You can start to imagine how, although it’s very early days for this, that the IoT and its various forms could be a fantastic way to start to develop some new products and services that you could offer to customers. We’re just seeing the beginnings of that with Progressive.”
Today’s manufacture of information products requires two new steps compared with development 20 years ago. Kudyba explained that the 1996 Meyer-Zack model for the manufacture of information products was composed of five information processing steps: acquisition, refinement, storage/retrieval, distribution, and presentation. Kudyba and Davenport’s research shows that the manufacturing process in today’s era of big data requires a new initial step of conceptualizing the project and a new concluding step of extracting market feedback. This last step is easier than ever before, because there are so many ways to keep in touch with the ultimate consumer. It’s also necessary, Kudyba said, to make sure the product adds actual value to the customer. The model is a closed loop, too: “It cycles,” Kudyba said. “It never really stops. Given the potential of the value of information products, this should be an ongoing process for many organizations in this big data era.”
Conceptual modeling, the first step of data product development, is essentially “the art of analytics.” Conceptual modeling entails an organization’s ability to perceive and facilitate a market need for information, Kudyba said. This has three requirements: knowing where new data resides that can produce a product of value; having a skill base to know whether the data can be processed and analyzed properly; and structuring available data sources to form a coherent product. “There are so many data sources out there, whether it’s structured or unstructured, streaming data, what have you,” said Kudyba. “You’ve got a science of analytics, and an art of analytics, and I like to focus on the art, because not too many people do these days. What I consider the art of analytics is thinking about all the data sources that are out there, and identifying variables that really add value — to models, to information, or to data products.”
In the conception stage, organizations should focus on a metric of interest to their market. “When you identify that metric of interest, then it simplifies the creation of the data product,” said Kudyba. “I say the same thing when you’re doing data mining: ‘think first, analyze later.’” Thinking first means figuring out what you want to better understand. If you can identify the hot metric and then source the right data for variables that impact that metric, “now you’ve got a really cool data product.”
Closing tips: How can organizations be most successful in the data economy? Davenport and Kudyba closed their presentation with quick tips for how to be successful in the data economy with the IoT. Their list: Get some good data. Make sure you have rights to it. Develop strong integration, cleaning, and analytics capabilities. Try them out on yourself first. Understand your customers’ decision needs and processes. Add some automation. Prevent security/privacy problems. Address pricing/bundling/terms issues. And handle any conflicts with your existing business. “Some companies said to us, ‘these data products are the pirate ship, they’re not the mother ship, and we don’t want them to cannibalize the mother ship,’” noted Davenport. That means being clear about how new data products and services might affect your existing business in ways that are both challenging and exhilarating.
IoT and Developing Analytics-Based Data Products
No comments:
Post a Comment