Doctoral Student

Clarence Lee

Clarence is a doctoral candidate in the Marketing Unit at Harvard Business School.  His research examines the drivers behind consumer adoption, usage, and purchase dynamics of digital goods, where he model consumer behavior using state-of-the-art Bayesian statistics and structural econometric techniques.  Digital products and platforms are increasingly present in almost all consumer interactions.  In such settings, understanding consumer choice and the dynamics of engagement and usage become critically important in order to acquire, serve and retain them as consumers.  

Before HBS, Clarence received undergraduate and graduate degrees in Electrical Engineering & Computer Science, both from MIT.  For his master’s thesis, he built a Bayesian Probabilistic Inference Engine to power User-Adaptive Morphing Websites.  Prior to pursuing graduate studies, he has conducted nanotechnology research at IBM and space system design at MIT Lincoln Laboratory.  In 2005, he cofounded the MIT Technology Fair, a MIT organization that showcases latest technologies from the private and public sectors.

Working Papers

  1. Designing Freemium: a Model of Consumer Usage, Upgrade, and Referral Dynamics

    Abstract. Over the past decade "freemium" (free + premium) has become the dominant business model among internet start-ups for its ability to acquire and monetize a large install-base with limited marketing resources. Freemium is a hybrid strategy where a firm offers both a perpetually free but limited version of their service, and a premium version with enhanced features that require a fee, and where firms regard the free product as a promotional tool. The model leads to several questions interesting to marketers, which we explore in our framework. How much value should the free product provide to consumers relative to the premium product, given the inherent cannibalization effect? What is the right referral bonus incentive to offer to customers? How does sharing influence customers' likelihood of upgrading to the premium product? We develop an empirical microfoundations-based framework to understand dynamics of consumer behavior of plan choice, usage, and referral in the Freemium setting and apply it to a novel panel data set from a leading Cloud storage service. Using Bayesian methodology, we estimate the structural model and perform counterfactual analysis. We find that the value of free consumers is approximately $22 per year, and that the existence of the referral program contributes to 65% of this value — signifying the importance of the referral program. In addition, we conduct profit maximization simulations, and we observe an asymmetry in the magnitude of the change in upgrade rates as we increase and decrease prices. Lastly, we explore simulations to maximize the average consumer referral rate by changing the referral incentives. Contrary to the belief that more is better, we find the existence of an optimal incentive point for referrals. Thus, we are able to characterize both the individual value of consumers to the firm as well as the network value of customers, providing a mechanism to capture the impact of consumer-to-consumer interactions.

    Keywords: Discrete-Continuous Choice Dynamic Structural Models; Bayesian Estimation; Word-of-Mouth; Digital Services; freemium; entrepreneurship; Business Model; Motivation and Incentives; Marketing Strategy; Internet; Consumer Behavior; Marketing Reference Programs; Business Startups;

    Citation:

    Lee, Clarence, Vineet Kumar, and Sunil Gupta. "Designing Freemium: a Model of Consumer Usage, Upgrade, and Referral Dynamics." Diss., Harvard Business School, 2013. (Job Market Paper.)
  2. Where do the Most Active Customers Originate and How Can Firms Keep Them Engaged?

    In this paper, we study how firms offering Web services can acquire and develop an active customer base. We focus on two basic questions. First, how does the method of customer acquisition affect the way customers use the service to meet their own needs and to interact with one another? Furthermore, how do firm-to-consumer communications affect the way customers use the service relative to customer-to-customer communications? Using data from a Web service start-up, we estimate a multivariate hierarchical Poisson hidden Markov model that captures the joint dynamics of customer engagement (personal and social usage) at the individual customer level. We segment the customers by the three most typical acquisition tools for Web start-ups: Word-of-Mouth (WOM), Mass-Invite (e.g. Techcrunch), and Search (e.g. Google). We find that customers who hear about the service through Search and Mass-Invite exhibit higher usage behavior as compared to customers from WOM, and that customer-to-customer communication is more effective than firm-to-customer communication at keeping customers engaged post-adoption. Even though WOM-acquired customers use the service less than customers from other adoption routes, WOM in the form of customer-to-customer sharing more robustly transitions customers into higher usage states. Hence, firms may be well-advised to encourage customers to share with each other post-adoption. Our work calls for a deeper understanding of the mechanisms that drive usage behavior, and for further exploration in the possibility of segmenting customers based on how they were acquired.

    Keywords: Customer Engagement; Adoption Routes; Hidden Markov Models; search; Word-of-Mouth; digital media; entrepreneurship; Customer Relationship Management; Search Technology; Mathematical Methods; Consumer Behavior; Entrepreneurship; Marketing Reference Programs; Web Services Industry;

    Citation:

    Lee, Clarence, E. Ofek, and Thomas Steenburgh. "Where do the Most Active Customers Originate and How Can Firms Keep Them Engaged?" Working Paper, 2013. (Revise and Resubmit at Management Science.)
  3. Viral Videos: The Dynamics of Online Video Advertising Campaigns

    Keywords: Online Advertising; Advertising Campaigns;

    Citation:

    Elberse, Anita, Clarence Lee, and Lingling Zhang. "Viral Videos: The Dynamics of Online Video Advertising Campaigns." 2013. (Invited for resubmission at Marketing Science.)

Other Publications and Materials

  1. User Adaptive Web Morphing: An Implementation of a Web-based Bayesian Inference Engine with Gittins Index: Master of Engineering Thesis

    Citation:

    Lee, Clarence. "User Adaptive Web Morphing: An Implementation of a Web-based Bayesian Inference Engine with Gittins Index: Master of Engineering Thesis." Massachusetts Institute of Technology (MIT), 2008.