The choice between open vs closed is not binary.

There are different degrees of access and the decision should always fall downstream from the job-to-be-done:

- A public API for low-stakes speed
- A private-cloud closed model for quality with control
- A self-hosted open model when the data cannot leave at all

My team at @SocialCapital mapped this out in our most recent Deep Dive on Open vs Closed Models.

Below was pulled from the deck.