Managing Experimentation in a Continuously Deployed Environment
The first day on the job, an Etsy developer is expected to push a live change to our web site. And thats just a warmup. Last year Etsys 150+ developers deployed nearly ten thousand changes. Over the same period we launched over 300 experiments. How do we keep track of whats being launched when any developer can deploy new code at any time? How do we communicate these changes across the company and to our larger community? And how do we run controlled A/B experiments in the midst of this continual change?