Data catalogs are changing the nature of working with data (sponsored by Alation)

Deb Seys shares the results of a study that she oversaw at eBay in collaboration with the Kellogg School of Management at Northwestern University. Examining the work of 2,000 analysts and almost 80,000 queries, the study revealed that a data catalog can be used as a learning platform that increases analyst productivity and creates a more collaborative approach to discovery and innovation.

Deb Seys shares the results of a study that she oversaw at eBay in collaboration with the Kellogg School of Management at Northwestern University to study the future of work and how a data catalog can be used as a learning platform. The study started as an academic endeavor looking at the nature of tacit learning among software developers, data scientists, and data analysts. However, the findings are potentially more broadly applicable. To get started, eBay made available four years’ worth of logs of users interacting with a data catalog product. In that period of time, just over 2,000 analysts performed almost 80,000 queries, all of this work logged through a data catalog. As one of the earliest adopters of data cataloging technology at scale, eBay is one of the few enterprises in the world that could have provided enough observations to support statistically significant research. The log files were used to test the time that it took a brand new user to complete an inquiry on a new set of data in a database that was unfamiliar to them. A timer started from the moment the user started typing their question to the moment that they received a response that they were satisfied with. At the same time, the files were used for looking at whether giving guidance to that user in the form of example queries written by experts had an effect or not. Just like shopping on Amazon, this data catalog has the capability to make recommendations to users. But instead of “people who bought this also bought this,” the recommendation in a data catalog is “people who asked this business question also looked at this data.” Users who never saw recommendations were compared with users who saw recommendations from different classes of “expert” colleagues. The conclusion of the research was that this recommendation approach to learning was indeed as a more productive approach than just letting analysts loose to experiment on their own. But the study also revealed insights about the social aspect of learning, including the effects of a collaborative approach to discovery and innovation. This session is sponsored by Alation.