Problem
Operations and go-to-market teams at HubSpot wanted to act on CRM data but needed an analyst for every question: they had to know whom to ask and then wait, sometimes until the answer was no longer relevant. The data sits in several systems and has billions of rows. A dedicated team and a customer-facing product would need hiring, training and time.
Example from the article: operations admins asked "How complete is this CRM record and property?", "Is this CRM property outdated, and when was it last updated?", "Is this property redundant or used anywhere?" and "Who has access to this data?". They spent hours collecting answers and asking in Slack channels.
Solution in brief
An analyst built interactive tools on the existing stack: dbt models in Snowflake, new Looker explores, and dashboards with filters and parameters. Users enter what they need instead of waiting for a one-off report. The same tools work as a proof of concept for product and engineering.
Questions this case answers
- How to let non-technical teams answer recurring data questions themselves when every request currently goes through an analyst queue
- How to validate demand for a data product cheaply before product and engineering commit roadmap time to it
- How to let operations and marketing teams answer recurring data questions without waiting for an analyst?
- How to build an internal self-service data tool on a warehouse, dbt and a BI layer without engineering help?
Tools and methods
- dbt — scheduled transformations in the warehouse that produce documented tables other analysts can reuse
- Looker — explores and dashboards with filters and parameters for end users
- Slack — automated alert messages sent from Looker rules
Google or a one-time email code. No password.