Knowledge graph

Why the shape matters#

In a table, a churn reason and a usage drop are two cells that happen to share an account id. In a graph, they are connected, and so is the feature both point at, and so is the segment those accounts belong to. Agents navigate those connections to find patterns that a per-source analysis cannot see.

This is the structural reason Poth answers questions that span sources. It is not a bigger summary. It is a different data model.

What lives in the graph#

• Customers and accounts, with their plan, tenure and segment
• Signals: calls, tickets, survey responses, chat, notes
• Product entities: features, flows, integrations
• Behavioural facts: what accounts did and when
• Hypotheses and their supporting or contradicting evidence

What agents do with it#

Agents traverse the graph to test explanations, which means a hypothesis about mid-market churn can pull in a call from six months ago, a support pattern and an activation curve without anyone joining those datasets by hand.

Next#

• Hypothesis engine shows what the traversal produces.
• What Poth is puts the graph in context.