Written by Tamr
“You can’t deduplicate six million records manually… hell will freeze over before you finish,” according to Turing Award Winner Dr. Michael Stonebraker.
Dr. Stonebraker, CTO and co-founder at Tamr, was recently a guest on the “Inside Analysis” radio show hosted by Eric Kavanagh where he and co-panelists Greg Michaelson of DataRobot and Bruno Kurtic of Sumo Logic discussed how machine learning is changing how enterprises do business.
Machine learning is “the only possible way to” solve the thorny problem of data unification at scale in large enterprises, according to Dr. Stonebraker. “All enterprises decompose into business units so they can get stuff done. Business units are independent or semi-independent and they build their own data repositories. There is huge value to integrating them downstream either for better customer experience, to save money, or for cross-selling.” He gives as an example of two Tamr customers, GE and Toyota Motors Europe.
According to Dr. Stonebraker, in large conglomerates, “the ideal is to have one procurement system… GE had 75”. He explains that this gets in the way of leveraging their scale in negotiating “favored nation status” with suppliers to get the best terms and conditions. By integrating 75 independently constructed supplier databases with a total 6 million records, Tamr enabled GE to save $80M a year through better supplier negotiations.
At Toyota Motor Europe, there were independent customer databases across each country in Europe. As a result, if you buy a Toyota in Spain and you move to Germany, they do not know who you are in Germany. So they are working with Tamr using machine learning to integrate 250 customer databases with about 30 million records written in 40 languages.
Dr. Stonebraker explains, “All Tamr projects are sold on ROI and generally speaking, our projects have been tried with some conventional technology and failed. So the way our projects come into existence is we convince the customer they will save money by using our technology. The only way that is going to happen is with incredible automation.”