Shanghai’s first Startup Weekend arrived in July 2011, drawing a few hundred eager participants to what would soon be headlined “Startup Hub on the Rise.”
It was the growing startup activity in this booming metropolis of 25 million people that whetted a new Fudan University graduate’s interest in innovation and technology. That interest led her to enroll in NC State’s Master of Global Innovation Management (MGIM) program and undertake groundbreaking Big Data project work for CIMS and Air Products & Chemicals, Inc.
Shenyi Zuo graduated from Fudan, one of China’s top universities, that July with a bachelor’s degree in digital advertising and marketing. She recalls that her statistics courses would later help her to learn the use of Big Data tools, while friends who were starting their own companies made her aware of “startups, innovations and cool technologies.”
Shenyi Zuo demonstrated how Natural Language Processing could be used to accurately translate Chinese into English.
During the next year, Shenyi interned in three Shanghai-based international media agencies as well as the Shanghai Futures Exchange. But early in 2012 she discovered a blog by an applicant to the MGIM program, a one-year dual (international) master’s from North Carolina State’s Jenkins Graduate School of Management in partnership with the IAE Graduate School of Management near Aix-en-Provence, France.
China to France to Raleigh, North Carolina
MGIM is focused on managing business innovation, and Shenyi was intrigued with how new technologies get transferred to the market. But it was the two-country feature that she found particularly appealing. “I was still very young and agreed with the friends and colleagues who said I needed to see the world,” she told IMR in a recent interview.
Shenyi was accepted for MGIM’s Fall 2012 semester in France after a Skype interview with the academic director, Anne-Marie Guérin, who was impressed by Shenyi’s top 5 ranking in her Fudan digital media major. She also found Shenyi “open-minded, curious, willing to learn from every experience, and very determined and focused.”
Guérin added that her applicant’s involvement in the student innovation project to develop a business model and value chain for the global digital security company Gemalto “contributed a lot to increase the quality of the group work and was deeply appreciated by Gemalto.”
Before long these personal characteristics would also win over Shenyi’s CIMS instructors in Raleigh. But that gets ahead of our story.
Another feature which interested Shenyi was the program's approach to teaching and exposing students to new thinking and experiences - embracing the philosophy that successful innovation management requires is a combination of knowledge and exposure to art and science. For example, in France, students learned through shorter classes and covered topics relevant to their cultural immersion experience, such as art, language and life in Provence. In the US, classes were spread throughout the semester and focused primarily on more traditional business topics. Both programs provided opportunities to engage with industry in company sponsored - student team projects, leaving the students with practical and invaluable experience and newly gained perspective. Consequently, Shenyi would advise any friends considering the same choice "to enjoy the unique experience of studying in each country.”
Using Big Data for Real
Shenyi’s U.S. studies commenced with the spring 2013 MGIM semester at NC State in Raleigh, where she became “one of our top products,” says Paul C. Mugge, CIMS executive director and innovation professor in the Poole College of Management.
That path to “top product” began in Frederick Renk’s Innovation Practicum where she and two other students worked on a project for Air Products to assess potential growth of the lithium-ion battery market. “Shenyi provided valuable information to the team using unstructured text analytics (Big Data),” says Renk, a CIMS Industrial Fellow who mentors the spring MGIM student practicum projects. “She learned the tool well and quickly realized how powerful and efficient it could be.”
Shenyi credits Prof. Michael Kowolenko’s MBA elective in data-driven decision making for her first exposure to using Big Data “for real.” She had heard about it at Fudan, of course, “but it was very new then and I didn’t know how to use it.”
Will This Work in Chinese?
However, it was an encounter at the spring 2013 CIMS Sponsors meeting that really made Shenyi’s instructors take notice. Discussing the battery project with Shenyi, Mark Listemann, an Air Products R&D executive, asked her, in effect, whether Natural Language Processing (NLP) would work in Chinese.
Air Products had been experimenting for some time with ways it might use Big Data (see “How Big Data Added Value for Air Products,” CIMS IMR May/June 2013). Listemann explains that his question was motivated by “the suspicion that much of the primary source material relating to companies would be available only in Chinese, particularly for local (non-multinational) companies, so it is critical that we are able to utilize Chinese language dictionaries and rules to identify relevant documents.
“There are also websites that can only be accessed from within China, so we needed to demonstrate that sufficient information is available from web crawls initiated outside of China.”
Shenyi had the answer for Listemann a week later. She showed him Chinese dictionaries she had created for the battery project and demonstrated how the IBM Content Analytics (ICA) software could crawl Chinese-language websites and identify various key sentences, like the following from a McKinsey report: “The price of a complete automotive lithium-ion battery pack could fall from $500 to $600 per kilowatt-hour (kWh) today to about $200 per kWh by 2020 and to about $160 per kWh by 2025.”
Shenyi explains that ICA and other software have difficulty translating Chinese into English accurately because they are so different. For example, Chinese words are characters, Chinese has no tenses and there are no spaces between Chinese words. “Consequently, I had to use my knowledge of Chinese grammar to solve the problem; for instance, I used keywords to insert the spaces correctly.”
The Data Modeler
With Shenyi having demonstrated what Renk calls “her unique combination of skills,” he and Listemann began designing a CIMS summer internship project for her. The project would be to identify and qualify potential new customers for Air Products in China’s auto parts supplier industry.
Timothy Michaelis, a CIMS research associate and MGIM 2012 grad, would be her supervisor but recalls being impressed by how hands-off he could be. “She was a quick study and not afraid of learning new software, which I see as a rarity.”
Michaelis says the MGIM degree in innovation management had given Shenyi “unique abilities to approach and analyze difficult business development and innovation problems.” Moreover, her familiarity with a variety of business strategy frameworks, which she gained in her MGIM courses, “have made her capable of identifying critical pieces of information needed to build a business development case. A fundamental understanding of strategy is critical in developing NLP models to answer strategic business questions.”
When I asked Michaelis about her specific contribution to the project, he explained that she had used the strategic management principles she had learned (Porter’s 5 forces, PEST, SWOT, etc.) to develop a decision model for Air Products relevant to China’s metals processing market.
“By combining her competency in strategy and NLP software she was able to deliver a unique solution to Air Products. By developing text algorithms (rules) she was able to pinpoint relevant information in the China metals market from the collected web data; for example, a list of locations of metals companies in China, their locations, specific processes/equipment they used, and any industry mergers/partnerships mentioned. She provided a visual map of this information.”
Michaelis noted the importance of Shenyi’s access to an Air Products marketing manager for metals processing in China who could make sure her model was correct and returning good information. “In essence, Shenyi was a data modeler who could effectively communicate with the China subject matter expert.”
Back Home, Shenyi Reflects
Shenyi’s internship with CIMS has ended and she is back home with her Master of Innovation Management degree to seek a position in one of China’s burgeoning industries. “I even want to start my own international business someday,” she told IMR, adding that she was confident that somewhere she would be using Big Data.
Meanwhile, she looks back on her CIMS experience as “very good and powerful for me.” The hardest part? “Finding a way to use the ICA tool in the Chinese language.” But she acknowledges the help she received at weekly meetings with the marketing and computer-modeling people on the Air Products Shanghai team.
Shenyi considers her project still in the pilot stage and requiring more testing. Air Products’ Listemann finds the results to date encouraging, observing that “the ultimate output will also be much more user-friendly for our local teams if it has not been translated into English, since this process is prone to error.”
Asking the Right Question
Fred Renk praises Listemann for asking the interesting question (Will this work in Chinese?) and challenging his students to demonstrate it. “It is very important that companies envision the possibilities for this capability and come up with real issues/questions to address. Absent strong company interactions, we are only left to guess what would be useful and valuable. Mark and Air Products should get a lot of credit for doing this part well.”
CIMS director Paul Mugge echoes Renk by observing that many companies CIMS interacts with don’t really know how to frame strategic questions. “Some companies can’t even think of the question. Our mantra at CIMS is that Big Data is not about the technology but about asking the right question.” (See Fig. 1.)
Getting Robust Data
Mugge finds two other aspects of Shenyi’s project remarkable. First, it demonstrates his “still unproven” hypothesis that Air Products got much richer knowledge out of the Chinese language than it would have from a straight English translation. Shenyi agrees, noting that she tested Google translations of Chinese dictionaries to English and back from English to Chinese. “Both were bad,” she says. “Direct translations from Chinese language proved more accurate.”
“What this means for global companies,” says Mugge, “is that their market intelligence functions should include multilingual people. In America we sit back and assume everyone’s going to speak English but Shenyi ended up with a few million web pages all in native language—a tremendous increase in the amount of knowledge they had available.
“Not only don’t many companies have multilingual people in their market intelligence functions but they don’t even have the function!
What they call market intelligence is people reading English-language reports. They’re not even taking advantage of the new capabilities computers are making possible.”
Mugge hopes to undertake more projects like Shenyi’s, which he considers “a tremendous first step.” Stating that he would like to run comparisons of what’s found from English searching of Chinese websites versus Chinese searching of Chinese websites, “I think there would be a night-and-day difference.”
Getting Robust Data Fast
Mugge’s second accolade goes to the remarkably short time it took Shenyi to complete her project—three months from receiving the project brief in June to presenting the results in August. “That is unheard of,” Mugge exclaims, “even when aided by supercomputers.”
What was Shenyi’s secret? “I had no secret, but hard work and help from the Air Products China team in learning the China market quickly and the web sites to search. They also helped me improve the dictionaries.”
But if there was any secret it was the eight-step decision model that CIMS has developed to help companies use Big Data analytics to make truly data-based strategic business decisions (see Fig. 2). “Shenyi followed it to a tee,” Mugge says, “and it is what we teach our master’s students.”
The CIMS Big Data process was designed so that the student teams could complete a sophisticated innovation project in one four-month semester. They are doing this repeatedly, says Mugge, “so the process we are teaching is not only paying off in robustness of data but in speed.”
This is important, Mugge continues, because research by Stephen Markham, CIMS research associate and Poole College professor, has shown that over the past 20 years all phases of innovation except the initial discovery phase have speeded up. That initial phase takes as long as it did 20 years ago!
Betting on Big Innovation
The implication of this, Mugge suggests, is that too many big companies are not “decision ready”; even when presented with huge amounts of data contradicting a current strategy or spotting new market opportunities, for example, they are deferring decisions and actions.
“At CIMS we are trying to use this process to motivate companies to act faster and with more confidence in the face of opportunities to make the big bets—the bets that can lead to what we call Big Innovation. Too many companies today are stuck in the incrementalism that leads to the commoditization of their products and services, and eventually to shrinking profit margins and even their demise. I believe Big Innovation is more likely when companies have more facts earlier.”
That, ultimately, is the promise of Big Data, promise a little closer to being realized because a young Chinese student discovered the MGIM program and chose to see the world.
If you would like to read this article in Mandarin, please click hereMichael F. Wolff Editor, IMR firstname.lastname@example.org