Full Record of Cloud Study Camp（II）

In recent years, data has played an increasingly important role in various decisions. In our daily routine, all kinds of online services such as search engines, online encyclopedias and even social networks are continuously producing rich datasets. Even offline activities can reflect people’s intentions and preferences. Then what can we do with these datasets? How can we combine them with advanced technologies like Machine Learning and Data Science algorithms to improve consumer experience and business performance? On the lecture on August 9th, we were excited to have the second section of our “Cloud Study Camp”: Utilizing Online Data to Improve Business Performance, we were honored to have invited three guests from academia and industry, Dr. Shachar Reichman, Oded Poncz and Assaf Binstock, to explain the mystery of big data for us.

The first guest to share is Dr. Shachar Reichman. Dr. Shachar Reichman is an assistant professor at Tel Aviv University School of Management and a research affiliate at the MIT Sloan School of Management. Dr. Shachar Reichman focuses on utilizing big data to improve businesses’ performance and consumers’ experience, also analyzing Networks-of-Networks to make better decisions. To start his speech, he first gave an example of how Kawhi Leonard, a famous NBA star, makes an offensive or passing decision while holding the ball near the top of the arc. This example helped us initially form the concept of data driven decision making. Different from traditional decision-making schema, which can be summarized as ‘Hippo’, Highest Paid Person’s Opinion, data driven decision making relies more on big data. So, what is big data exactly?

According to professor’s argument, big data has four features that can be summarized by four V’s, Volume, Velocity, Variety and Veracity.

◆ Volume: Data at Rest, terabytes to exabytes of existing data to process;

◆ Velocity: Data in Motion, streaming data, milliseconds to seconds to respond;

◆ Variety: Data in Many Forms, structured, unstructured, text and multimedia;

◆Veracity: Data in Doubt, uncertainty due to inconsistency, incompleteness and ambiguities…

Data can reveal a lot. Data tells us our preference between distance and price, which can be different at different time in a day, and this can help us build a better insight into how decisions take place. Even our gaming data, how we react to different accidents and behave in various situations, can influence future employers’ decisions. They can judge by our game performance and infer whether we are beneficial or harmful to their company. Google map records our driving behaviors and unexpected turnings, and these can be essential to insurance companies.

Consumers do a lot of online activities nowadays, consequently, a remarkable size of data are produced. Over time, big data is becoming a new type of corporate asset that will cut across business units and function much as a powerful brand does, representing a key basis for competence, according to McKinsey Quarterly. To illustrate big data’s astonishing power for competence, Dr. Shachar Reichman gave us a detailed account of the results he obtained in his collaborative research, Consumer Location Dynamics and Gas Station Choice. In traditional ways, we may choose a gas station location basing on the place of residence and store location. Currently, about 80% of the U.S. population own a smart phone and 94% of 500+-employee companies collect/store location data. The millions of dots on the map trace highways, side streets and bike trails — each one following the path of an anonymous cellphone user, which is scary but true. Collecting consumers’ data and tracking their location dynamics, we can build models to improve businesses’ performance as well as consumers’ experience.

Next, an industry expert from Ubimo, Oded Poncz, showed us how their team use big data to make optimal decisions. Oded Poncz founded Labpixies which was acquired by Google in 2010. And in 2013, he founded Ubimo which was acquired by Quotient. Today he is the CTO of Ubimo and VP of Engineering in Quotient. Oded is passionate about large scale technology, science and nature.

What is Ubimo?

Ubimo is a SaaS location intelligence company that empowers businesses to understand and act in real-world behaviors, bridging the knowledge gap between the digital and physical worlds. They aim at connecting the physical and digital worlds with location intelligence.

How does Ubimo work?

Ubimo uses advanced AI technologies to combine behavioral data over 150 milion monthly active users with the most comprehensive geographic index in the US. Ubimo connects proprietary location and movement data of 150+ million daily active users across 12+ million US venues to first- and third-party online data, and in this way, they can help understand the entire customer journey. Besides, Ubimo empowers organizations to unlock people’s real-world movement patterns and online behavior through customer analysis by asking and answering important business questions. Finally, Ubimo transforms location data into insights, distilling big data into actionable next steps. These technologies help Ubimo to plan strategically, execute accurately and attribute successfully.

Then Mr. Oded Poncz introduced a platform called Ubimo Polaris and demonstrated how it made sense on the website. This platform is good at precise targeting. We can reach consumers at any scale with Polaris Realtime DSP through geotargeting, venue level targeting, decision-moments targeting and mobile retargeting. Also, Mr. Oded Poncz introduced a concept of “audience”. We can optimize our outreach by defining and creating custom and granular datasets based on real-time behavior and combining them with first- and third-party data.

Ubimo is transforming a lot. Transforming mobile display made Ubimo one of the top 5 mobile programmatic platforms worldwide; transforming Out-of-Home made OOH an essential part of any omnichannel strategy, and transforming shopper marketing helped increase ROI by 29%.

Programmatic Digital-Out-of-Home Platform is another powerful product of Ubimo. Consumer spend 70% of their day outside, surrounded by Out-of-Home (OOH) advertising. The ability to access digital OOH inventory programmatically, via real-time bidding, opens up a new world of possibilities to effectively reach target audiences. Ubimo provides a Programmatic DOOH solution designed and built for the OOH industry, and Ubimo has the unparalleled targeting capabilities.

After a short break, Dr. Shachar Reichman continued to share the topic of using predictive analytics to reduce uncertainty in enterprise risk management. The credit rating of firms has a significant role in the financial systems. This rating measures the risk of the firm and the probability of default or delayed payment of bonds and other debt instruments, and is one of the key paraments in investment, financial and operational decision making. Traditional forecasting of credit risk has relied on statistics from government agencies, annual reports, and financial statements. Statistics are, however, often published with significant delay, which limits their usefulness for prediction. These are disadvantages of traditional ways of estimating credit risk. In this part of lecture, Dr. Shachar Reichman mainly talked about how credit rating is made, key parameters in investment and how to reduce information asymmetries.

In traditional ways, we may make credit rating basing on some financial factors like net income, current ratio, quick ratio, etc. Then we can build analytical models using these data, also we may even apply machine learning algorithms on them as well. As a result, we can receive some indicators like Z-Score and O-Score below. This is right but not enough. There are various kinds of factors that can influence the financial stability of a firm, thus forming the gap of credit rating. Models built above do not account for alternative information sources like news, online activity and social media, besides, each firm is analyzed separately, which is not reasonable. 

To make credit rating more accurate and reduce uncertainty in enterprise risk management, we must take a lot more into account. Online activities can affect risk seriously. In February 22, 2018, Kylie Jenner tweeted about Snapchat and its stock lost 1.3 billion dollars in value. Then comes the question, how can we take these online activities like Wikipedia visits, social media and Google trends into consideration? We can use Pageviews Analysis to help us. This website counts the number and frequency of each hot word appearing on the page, and these data can be used into modeling. Online activities really reveal a lot in predictions like flu trends this year, real estate market and unemployment.

Finally, Assaf Binstock, CEO and Co-Founder at BeeEye, showed how their product used AI algorithms to reduce risk and assess a company’s credit rate. This year, financial sector is facing increased credit risk while unrealizing full business potential. 2020 Pandemic has created more pressure, credit demand and risk to tackle. At the same time, most modelling methods are in a cumbersome way, not utilizing full data potential. We are facing a lot of difficulties from technology and resources. Then EyeOnRisk Platform, a credit-risk focused, smarter, faster model management platform came into being. This platform combines internal and external data sources and helps reduce our default rates and lower false rejections by managing our models effectively.

This session of Cloud Study Camp help us get a deeper insight into Big Data and learn more powerful approaches to make decisions based on data and combine data with real-world applications. Data is productivity!

// Reflection

DS&BA 何友

MF 殷铭

MF 彭思凯

MF 夏济舟

Review：