Renato Martins, Groupon, Inc.
Data Science is a “hip” term these days. Most people automatically associate Data Science with financial and feature performance applications. It turns out that Data Science provides a set of very powerful tools that can be used to assess software quality, especially in the cloud/online world. Most cloud companies that supply online applications use the canary system to assess release quality before fully deploying a new build. Most cloud companies that supply online applications also provide a great wealth of data that can be used to assess the quality of the release candidate before it is fully deployed. A green set of test case results is not enough to make the deployment call. Many release engineering teams find themselves starring at graphs trying to assess if the new build is meeting the quality criteria. Sometimes what they think they see is not what it really is. In this talk we will go over statistical methods to use to aid in making quality decisions and boost confidence on go/no-go deployment decisions. We will explain the different methods we’re currently using and the process in which we selected the methods and what we learned in the process.