I'm Head of Growth at a startup building an AI agent that presents your slides and attends meetings for you. Our best growth channel is the product itself: every time the agent presents, everyone in the meeting sees it working, and some of them sign up. Measuring that loop was hard. Presentations and viewers lived in our product database, signups in GA4 and HubSpot, revenue in Stripe, and paid campaigns in Google and LinkedIn.
The Databox team set everything up for us. They connected every source, defined our growth metrics once in the semantic layer (presentations delivered, meeting viewers, viewer-to-signup rate, activation, paid conversion), and added our ICP, plans and growth targets as context. We were tracking our viral loop properly within weeks, with no data engineer.
Now I ask Genie "which customer segments create the most new signups from people who watched their presentations" and get an answer with reasoning in seconds. A Routine sends a growth pulse every Monday in Slack with what moved and why, including how the viral loop is trending. A cohort Skill shows which signup sources turn into paying teams, which shaped where we put our budget. I use Databox MCP in Claude to plan experiments, like testing a new end screen after each presentation, based on real loop numbers. Board updates are Artifacts that Genie drafts from the same metrics my team uses.
For a product that grows through its own users, Databox finally made the loop measurable.
Thanks for the detailed review, Ulykbek. Glad the MCP server is helping you query live data straight from your dev environment, that's exactly the workflow we built it for.
On the JSON formatting friction: fair feedback, unconventional payloads shouldn't need manual cleanup before ingestion. We're looking at ways to handle that more natively.
On refresh intervals: sync frequency does scale with plan tier, and some of that comes down to rate limits on the data provider's side, not just us. That said, we hear you on real-time verification during testing, and it's worth us looking at how sandboxed/dev setups are handled specifically.
Thanks again for the detailed feedback, it helps us prioritize the right things.