Lessons from Shanghai (In and Out of Class)

Our cohort of 75+ students representing 30+ languages / countries from the NYU Stern School of Business Masters of Science in Business Analytics (MSBA) 2019 class descended upon Shanghai this past week. What an amazing trip; first time in Shanghai and mind blown! The mix of class lectures and experiencing Shanghai City was an unforgettable combination.

Best module yet for the MSBA 2019 class with 3 renowned professors, outside speakers, and team activities. Network and Platform Analytics with Professor Arun Sundararajan, author of The Sharing Economy covering network effects and rise of the sharing economy. Data Visualization with Professor Kristen Sosulski, author of Data Visualization Made Simple covering best practices of all forms of visualization and how we can use visualization to make our stories and messages more compelling. Decision Under Risk with Professor Ilan Lobel, who continued to wow us in our decision modeling journey with more advanced modeling techniques using solver, crystal ball, analytics solver, and good ole excel.

An MSBA international trip is not complete without the globetrotting Professor Anindya Ghose, author of TAP, a must read for anyone trying to understand mobile / digital economy. Professor Ghose is as much an expert on the digital economy as he is on rooftop bars world over from London to Shanghai :). And a special shout out to Ava Danville, Henry Xie and everyone else who made it all possible for us to have a such a productive and fun week without a hitch.

Before I get to the in-class lessons, I would like to take a moment to wonder about the city of Shanghai. Shanghai, what a contrast from New York City. A city with ~24 million, the level of efficacy in its infrastructure that supports such a large population is truly mind blowing. What’s even more jarring are the high rises that rivals that of Manhattan were mostly built within the last 25 years. I might be a bit off but there are about ~24 bridges and tunnels across the Huángpǔ river that connects the city. Meanwhile, the politicians in New York have been debating about building another new tunnel between New Jersey and New York for the last 25 years. The ride from the airport to Shanghai took about 7.30 minutes to cover 18 miles travelling at a top speed of 431 kph (278 mph). Last time I took a high speed train in the USA was back in 2004, the Acela from New York to Washington at the comparably leisurely speed of 150 mph and not much has changed since then. China on the other hand continues to blanket the country with high-speed rail covering 17,000 miles as of 2018 and expected to be 24,000 miles by 2025.

Another surprise for me was the level and maturity of digital commerce in Shanghai. Everyone from the street food vendor to the high-end retailers seem to be digitally enabled with just a phone, ready to scan the barcode in your phone to receive payments. As a hapless foreigner, I carried around cash everywhere since very few retailers take credit cards. Eventually, I figured out a way to link my credit card to my new WeChat account. Wechat is ubiquitous; it’s your digital life in an app and one stop shop for everything. Whether you want to chat, make a call, order anything, make a payment, and order a cab, WeChat has it. It’s your facebook+ whatsapp + amazon + credit card+ venmo+ grubhub + uber + laundromat app + location mapper. More than anything else, I was impressed with how WeChat bridges the online and offline world. Think about the amount of data WeChat has on its users; when paying at a store using its barcode or receiving physical delivery of something ordered online with a level of precision, breadth, and depth perhaps not even our mighty Google or Amazon can match. This is something pointed out in Kai-fu Lee’s new book AI Super-Power I read on my flight to Shanghai but seeing how pervasive and real it is was something else. As Kai Fu Lee argues and most of us in the digital analytics space agree, data is the new oil that will power the next generation of Artificial Intelligence / Machine Learning technologies that will elicit and automate human repetitive tasks. China’s technology company are sitting on pile of data with a level of breadth and depth that few other companies around the world can match.

On to the lessons learned in the classroom….

Network and Platform Analytics

My favorite part of the class was having speakers coming in from VIPKids; a global platform that matches foreign English teachers with children in China through its platform and Didi; the taxi app that out-competed and outperformed Uber out of China. To see how inherent Analytics was part of their technology platform was remarkable. VIPKids uses kids and teacher’s facial expression to measure sentiments from each lesson and then uses the measurements to provide better services. How about Didi using its mountain of data collected to predict and route its drivers to the optimal path to the seconds. Didi can measure impact to traffic flow if any of its drivers stop at any location to the seconds interval. Maybe Uber has that level of precision, but I am doubtful.

Aside from the speakers, Professor Sundararajan used the remaining time to cover cool network and platform concepts such as:

Triadic closure – If someone is friends with two people, there is high probability that those two people will also become friends over time forming a triadic closure.

Power Law – Within a network, a few select enjoy disproportionate power over others. It builds up over time even with slight bias. For example, in a network of web pages, a web page may start out with equal weighting but if one page is looked at more favorably over others, over time it will become disproportionately important over its peers with all other important pages pointing towards it. The idea of centrality is at play, a webpage is thought to be important if other important web pages think it’s important. Same applies to people within a network, a person is thought to be important if other important people think he/she is important.

Bridge vs Hub – A hub is a central point within a network cluster while a bridge is a person / component that connects disparate sets of network clusters. Based on research, it’s been found that a person that serves as bridge tends to enjoy greater success in their career over time than those that may serve as a Hub; this could be due to a greater level of information that flows through a bridge. This is especially true in the digital economy.

Platform Economies – Professor Sundararajan covered various platforms and traced their rise. This is a fairly recent phenomenon from Ebay to Amazon to AirBnB. What was interesting is that pretty much all platforms followed the same playbook. Create a base product > iterate to provide great user experience > build / acquire user base giving away for free or at discount > scale up > monetize when you are at a position of strength.

Describing platform economy Professor Sundararajan went back in history to reinforce what platforms represent; it’s a new form of trust. As humans, we started out trading between tribes in the early days within close confines of geography. As humans traveled and the concept of government / currency medium came about, trade expanded to far reaches (trade of silk / other goods for gold / currency). Government eventually rolled out standardized contracts and property rights laws to further instill trust and to formalize trade which had yet more of an expansive impact on trade. In the 19th and 20th century the trust factor further expanded to large corporations to expand trade throughout the world. Crowd based platforms represent a new form of trust system that serve as a medium. When you think about it, a corporation like AirBnB or BlaBlaCar doesn’t really own any properties or cars, yet both have been super successful in facilitating transactions. Based on research, total strangers on both AirBnB and BlaBlaCar enjoy more trust than people’s coworkers. It will be interesting to see if trust in traditional corporations erodes in the long term or this new form of platform companies coexist with them harmoniously.

Data Visualization

Professor Sosulski's method of teaching is the best. She puts a significant amount of time and effort into creating content and builds on each lesson iteratively to deliver more challenging content. Walking through hands on exercises really reinforces the best practices of dos and don’ts. It was embarrassing to learn that even with 10+ years in management consulting, I was committing some visual faux pas :). Aside from the technical knowledge to create some advanced Tableau and R visuals, Key takeaways from Data Vis:

Follow the Process of Data Visualization by asking the 4 simple questions: Who is the audience? --> What’s the data --> What’s the task? --> What’s the best visual display? 

• Use visual to supplement your story or message
• Less is more and declutter, declutter, declutter
• Tailor your content for the medium you are delivering in (i.e. an interactive visual display maybe different than a static handout)
• Know your audience and their level of sophistication (i.e. if you use a boxplot for a lay audience, walk them through what it is first)
• What sort of display to use with what sort of data (I.e. Line chart for time series, bar chart for categorical, boxplot for distribution, bubble / scatter for multivariate etc.)
• Pay attention to trend, transition, and trace 

Decision Under Risk

Simply put, Prof Ilan Lobel much like his predecessor Professor Jiawei Zhang (Module 1 - Decision model), does magic with numbers. Want to know how to allocate your marketing dollars optimally based on the click through rate? There is solver for that. How about allocating your employees while minimizing cost? There is solver for that. Or how to allocate your investments to maximize your return while minimizing risk? You get the idea. Decision under risk took it a step further by looking at optimal decision and the cost of choosing an option over another, or what’s the best optimal path based on the number of different decisions using tree plan. We covered various topics and software to guide the decision making process. Topics covered:

1. Linear Models
2. Simulation
3. Stochastic Optimization without Recourse
4. Stochastic Optimization with Recourse
5. Decision Trees
6. Dynamic Programming