Data Analytics Evolution at LinkedIn – Key Takeaways

October 26, 2012
284 Views

At Teradata Partners Conference 2012 held this week near Washington D.C., Simon Zhang’s talk on “Data Sciences and Analytics Evolution @LinkedIn,” provided many useful insights for oraganizations wanting to expand into the space of decision making using Data Analytics built on Big Data ecosystems.

At Teradata Partners Conference 2012 held this week near Washington D.C., Simon Zhang’s talk on “Data Sciences and Analytics Evolution @LinkedIn,” provided many useful insights for oraganizations wanting to expand into the space of decision making using Data Analytics built on Big Data ecosystems.

  1. LinkedIn’s big data ecosystem contains eight basic functions working in a cyclic mode. The first function starts with understanding the company’s products indepth. Second, establishing tracking mechanisms to get the data about the product. Third, data management and data quality function focus on deploying good quality data across enterprise. Fourth, Adhoc analysis on the data provides first cut understanding of the data gathered. Fifth, business intelligence is used for standardized reporting. Sixth, deep analytics functions are used for extracting important patterns. Seventh, obtain insights to extract relevant knowledge from the patterns. Finally, the decision step derives the value utilizing the knowledge gained.
  2. These functional layers could evolve to be very diconnected loosing the sight on value generation. Therefore, when building these teams, formulate a team that works like one person; have a set of mixed skills cover the breadth and depth on all eight components of the model. The success is attributable to reducing or removing the boundaries within and across teams. When hiring people, they value skills to about 5%, IQ and EQ to about 15% and the passion to succeed to 80%.
  3. LinkedIn follows the “three second rule” to set the performance targets for the information delivery. LinkedIn believes speed matters when it comes to adaptation. Adaptation exponentially increases as the response time goes towards sub-seconds.
  4. The information provided to business should be specific and focused towards closing the deal. A lot of thinking and processing goes on before the final snippet of information is shared as the final action. For example, if there was a question on which companies need to be approached for specific product sales, behavioral data from about two million companies is analyzed to arrive at traget prospects. Then, the identity data is used to determine who within the selected companies should be approached. Finally, the social data of those individuals is analyzed to provide insight on how they need to be approached in order to close the deal. Thus, analytics at LinkedIn sets the focus on reflecting (meaning close to the truth) and not to merely predicting.
  5. The culture at LinkedIn is driven towards the final results by passing the charts or reports. Beautiful charts, dashboards and scorecards may look good, but are not enough if the focus has to be closing the deal.

In summary, LinkedIn seemed to be one of those companies that are heavily dependent on using big data integration coupled with analytics to provide insights for decisons.