(founder of mixpanel here)

I'm really excited about this release - sampling has been necessary to support some of our largest customers for many years, but has a bunch of issues, as Vijay points out in this post.

I love the fact that moving sampling to query time lets us beat the tradeoff - we get something like 90% of the speedup of ingestion time sampling with none of the downsides, and the only incremental cost is storage which is cheap. It's a big win.

For an analytics engine, this seems impressive. I wonder how it compares to memsql’s speed of querying.

Also Mixpanel’s pricing page seems to employ some dark patterns. How much does it really cost? So complicated.

Why not simply charge based on data and compute used. like GCP, AWS, & Azure billing?