Product analytics is supposed to give us answers, but half the time it just gives us more questions.

What is the biggest hurdle for your team?

1. Untangling instrumentation (when does this event actually fire?)

2. Decoding data (what does this property mean?)

3. Translating business needs into events (building a registration funnel from scratch)

Hey Product Hunt! Alex here.

A few months ago, while building our own AI features, we kept hitting a wall.

We're moving fast from simple RAG chat interfaces to autonomous, multi-turn AI agents.

Traditional analytics were built for static page views and button clicks. But when an agent can take 10 different paths to solve a single user intent, traditional funnels break down. You need to analyze:

• Looping behavior: Where do users get stuck re-prompting?

• Cost-per-conversion: Is a high-token agentic chain actually driving higher retention?

• Agent friction: Did a hallucination cause a rage click, or did latency cause the drop-off?

One thing we heard loud and clear from early engineering teams: "We don't want another telemetry agent bloat, and we can't send sensitive prompt data to a 3rd party cloud."

We took that feedback to heart when designing Kubit s architecture:

1. Plug into OTel: If you already send traces via OpenTelemetry or use a standard CDP, setup takes minutes. No complex proprietary SDK wrappers required.

2. Bring Your Own Warehouse (BYOW): With our Zero-Copy architecture, your data stays in your Snowflake, BigQuery, or Databricks instance. Kubit runs analytics on top of your data infrastructure, giving you full control over compliance, security, and data sovereignty.