Cohorts AI uses a proprietary algorithm to build a custom model that segments our customers based on actual product telemetry. The model draws on several billion of our own telemetry data points and applies unsupervised learning to construct these segments. I appreciate that the underlying data feels trustworthy and that the results usually align with ARR, even though ARR isn’t included in the model. By clarifying which customers are ideal versus “shelfware,” and by showing what usage (or lack of usage) puts them in those categories, we’re able to make more informed decisions about how to approach and connect with .. and market to - our customers in a far more meaningful and intentional way than without Cohorts AI. The QuadSci team is exceptionally responsive to requests and feedback from our team. Review collected by and hosted on G2.com.

When a business is changing quickly—building, acquiring, or divesting products—the model understandably needs to be rebuilt. This doesn’t happen often, but until a rebuild occurs, the existing model can become stale. Once the model is in place, our latest telemetry is incorporated on a monthly cadence (although it's possible this can be changed, we are in discussions). As a result, our cohorts are only updated 12 times per year, which makes it hard to see the impact of tactical, customer-facing changes quickly. For example, it’s difficult to track cohort shifts for newly onboarded customers, where we expect usage to start at zero and then improve rapidly over the course of weeks rather than months. Review collected by and hosted on G2.com.