"We find quite often, however, that your historical data is incompatible with the schema that you want to use going forward."

Can you go into this more? What schema do you find companies are using at scale and how is this different from their initial setup?

Once you learn about Snowplow's canonical event model (https://github.com/snowplow/snowplow/wiki/Canonical-event-mo...) or read Keen's data modeling guide (https://keen.io/docs/event-data-modeling/event-data-intro/) you quickly realize all the stuff that you haven't been recording :)

Nice post - Why did you choose to build on Keen IO vs accessing a service like Mixpanel's API?

We considered building on Mixpanel (and still are open to it) — we were 100% certain that Keen IO had the API we needed, and we liked that their business model was in line with what we're doing.

Also, I was worried that since Mixpanel is somewhat of a competitor, it wouldn't be wise to use them as a foundation.