Name:
Perry Beaumont, PhD
Title:
Lecturer
Country:
United States
Organization:
Columbia University, School of Professional Studies
Description:
The School of Professional Studies is one of the schools comprising Columbia University in the city of New York. It offers seventeen master's degrees, courses for advancement and graduate school preparation, and certificate programs.
Awards Categories:
Challenge:
For context, the applicable course as pertains to this Submission is Security Analysis, and it is a finance class taught at Columbia University within the School of Professional Studies. Security Analysis is also the name of a classic investment book written by Benjamin Graham and David L. Dodd, and Benjamin Graham was a professor of Warren Buffet when Buffet attended the Columbia Business School.
The class is very popular with students, and is taught each semester. The course is a post-baccalaureate offering, and many attendees are enrolled in a master’s degree program for business or public service. The class lasts 12 weeks, and students generally have a basic understanding of statistics and analytics.
A core element of the course involves building bridges for students between finance theory and practice, and the homework exercise involving Dataiku specifically relates to identifying important distinctions between the available attributes of a successful brick-and-mortar retail business and an online retail business. A helpful way of approaching this is to tap into a real-world dataset, as well as an online enterprise platform. Accordingly, Google Query was used to access actual (anonymized) eCommerce metrics from Google’s merchandise website, and Dataiku was selected for performing the analysis.
Solution:
Dataiku was immensely helpful, in different ways:
It was a great pleasure to collaborate with Dataiku personnel inclusive of Adela Deanova (concept development), Damien Jacquemart (programming contributions), and Josh Hewitt (classroom account setups), who each uniquely contributed to the success of the initiative.
- Product flexibility & extensiveness:
By virtue of Dataiku making its product available with a variety of venues, from a 14-day free trial to the leveraging of synergies with AWS, Azure, Oracle VM VirtualBox, and more, an array of learning opportunities are presented to help students appreciate the value of the Dataiku proposition.
The visual enhancement tools available within Dataiku per recipe, display of model results, and graphing possibilities - all combine to help make for a meaningful interactive learning experience.
Impact:
Generally speaking, the conversion rate for a brick-and-mortar retail store is about 20%; that is, about 20% of the persons who enter the store end up making a purchase. By contrast, the conversion rate for a person who visits an online retail store is closer to 2%.
Accordingly, with the appreciably smaller number of conversions online, yet with the ability to collect dozens of metrics related to a customer’s online experience (i.e., the customer’s device used to access site, length of time on pages, page path to checkout, and so forth), there is an opportunity to identify the factors that contribute to a greater likelihood of success in driving online sales. Even an insight that results in an additional 0.5% point in sales (from 2% to 2.5%) represents a 25% improvement in conversions (2.5%/2%-1=25%).
As a result of working through the Dataiku module, students were able to obtain a variety of invaluable insights. Not only were they able to better grasp the enormous amounts of data that can be generated by an eCommerce business, but were able to appreciate the tremendous power of Dataiku to generate meaningful analyses from especially large files.
Their analytical skills increased markedly, though perhaps even more impressive was the greater comfort level they exhibited with regards to drawing connections between the mathematical results and the practical implications. In the process, it became quite evident that students were becoming increasingly confident with developing a bilingual vocabulary to constructively evaluate both quantitative and qualitative dynamics of decision-making. The recommendations they made very much reflected a depth and breadth of understanding that went well beyond what would have been possible for them to achieve simply by reading a case study.
By way of one particular example, the univariate analysis tool within Dataiku very much provided a useful guide for students to evaluate the information content and value-add of each variable within the dataset, and opened up constructive conversations related to the true key performance indicators within an eCommerce context.
In brief, by virtue of digging into the data themselves, they were able to have a far richer learning experience, and one that will surely stay with them for a long time to come. By using actual Google Query data in combination with Dataiku, students were able to see for themselves what the customer experience looks like with well-defined data relationships, all while building bridges between textbook theory and real-world insights.
As an additional element of measuring success with this initiative, students who have taken this course routinely contact me to say that they are using Dataiku with other applications, both academic and in the business world. In brief, they are taking the basic skills developed in the classroom and are actively applying them in a variety of other contexts.
