FAQ
How is this different from pointing an LLM at all our feedback?
An LLM over a document pile summarizes sources. It cannot reason about relationships it was never given. Poth builds a model of your customers first, connecting signals to accounts, segments, features and behaviour, then reasons over that network. The difference shows on any question that spans sources, which is most of the questions worth asking.
How do I know a finding is real and not a plausible story?
Every finding carries its evidence, its segment and its confidence, and the competing explanations that were rejected are shown with theirs. Statistical testing is applied rather than reporting the first correlation. If a conclusion cannot be traced, treat it as a lead, not a decision. See Read the evidence.
Will this bury our customers in surveys?
No. Surveys are the last resort. A question is generated only when two hypotheses survive and existing data cannot separate them, and each question tests one specific explanation. Low volume is what keeps response rates alive. See Adaptive surveys.
How many sources do we need before this is useful?
One, to get a first answer. Start with calls or support tickets, the sources richest in unstructured signal. Reasoning gets stronger with each additional connection, and product analytics is what upgrades a complaint into a cause.
Can it tell us why customers churn, not just that they do?
That is the design goal. Churn reasons appear in call language and support tone before they appear in usage data. Poth connects those to behaviour and tests explanations against operational reality, so the answer arrives as a cause with a lever rather than a score.
What happens when Poth is wrong?
Disagreement usually points at a missing source rather than a broken conclusion. Connect the system holding the knowledge your team has and the reasoning updates. Because rejected hypotheses stay visible, you can also see exactly where the reasoning diverged from yours.
Do you train on our data?
No. See Data and privacy.
Who is behind Poth Labs?
A technical founding team from YC S26: Matthew Wong, cofounder and CEO, previously at Palantir and Stanford, and Mojmir Horvath, cofounder and CTO, previously at Tietoevry.
We will show how Poth attacks the root cause instead of summarizing the complaint.