Last night, I attended a Meetup hosted by The Hive, an incubator and think tank in the Big Data space. The presentation featured LinkedIn’s use of Big Data visualization tools to support their product development efforts.
The speaker, a Senior Data Scientist, shared the type of visualizations that LinkedIn uses to reveal more about its users and the effectiveness of LinkedIn to create the professional network it has envisioned. And because this company does deal with millions of data points everyday, the visualizations were beyond any type of scatter plot, distribution, or similar plots that are typically used in data analysis. The point of showing these visualizations was to illustrate how typical summary statistics – average, standard deviation, etc. – cannot reveal the whole story lying underneath the data. However, this is not what
really caught my attention.
The speakers then explained how certain visualizations are generated at set times during the day to track the daily performance of select LinkedIn features. These visualizations have become part of the regular workflow at LinkedIn. Data visualizations are not just used to uncover patterns to help solve specific problems. LinkedIn has truly achieved the status of being a data-driven organization.
The term “data-driven” has become so blasé, as if companies have NOT used data in the past to make decisions. Being truly data-driven was something that LinkedIn could achieve; it had the advantage of being a younger company, thus having a clean slate in setting up its workflows incorporating data analysis and visualization from the start.
For any company that wants to be “data-driven”, it is not enough to use the Big Data tools to solve specific problems one time. Being data-driven means that data analysis and/or visualization must be integrated into select business processes. It does not have to be to LinkedIn’s extreme, but incorporating some level of data monitoring – summary statistics, data plots, whatever can reveal anomalies quickly – will need to be present for a company to claim the “data-driven” moniker. And that is a lesson that the majority of companies still need to learn.