We're happy to announce we've finished and released the first version of Impact History Feeds for organizations in our software product today, now available to all partners and customers  🎉

While the initial user interface is relatively simple, what's import "under the hood" is the underlying event data logging pipeline (or stream) that allows us to aggregate and segment individual actions into histories, analytics reporting, as well as inform our machine learning models and outputs.

At a data level (to give you one example), this is similar to how companies like Amazon, Google, Spotify, and Netflix use machine learning and "artificial intelligence" (AI) in their products.

Using matrix factorization on our event data, we can solve for vectors with different algorithms (ex: gradient ascent or descent) to find similarity between different database entities as a cosign of vectors, estimate nearest neighbors, and make predictive recommendations within our product.

At a structural level, this is how most AI works, and it starts with gathering, preparing, and structuring your data in ways your models can access and learn from.

Prepare data = ✅