Two years ago we set up Open-Source Snowplow at CarGurus to fulfill a need for self-managed client-side instrumentation. Since that time it has become incredibly impactful for the entire company and has scaled significantly beyond what was originally envisioned. The following is an overview of why we set up the system, our experience with it, what we have learned, and where we see it continuing to go.
This post is quite long, so before getting too far into the details....
In a world where the importance of data is steadily increasing yet the cost of computing power is steadily decreasing, there are fewer and fewer excuses to not have control of your own data. To explore that point I instrumented this site as inexpensively as I possibly could, without sacrificing reliability or functionality. I have full control of all data that is generated, the instrumentation is highly customizable, the output is simple to use, and I don't have to be available at all hours to keep it working.
The components utilized include Cloudfront, S3, Lambda, Athena, and the Snowplow Javascript tracker.
And It costs less than $1 per month.