How a small HVAC company found $1.3M they didn't know they were losing.
Built the entire data foundation from scratch - real-time dashboards, custom integration platform, and operational intelligence that connected every part of the business.
ServiceTitan had the data. The business still could not see what mattered.
A growing HVAC company running on ServiceTitan had all the data — jobs, invoices, technician performance, memberships — but it was trapped inside a system that only showed surface-level reports.
Decisions were made on gut feeling because no one had time to dig through the numbers or stitch together what was really happening across sales, dispatch, invoicing, and memberships.
Revenue was growing, but nobody could clearly explain where it was coming from, where it was leaking, or which technicians and workflows were actually driving performance.
ServiceTitan reports were too shallow for real business decisions
No visibility into unsold pipeline or close rates
Technician performance was invisible beyond basic job counts
Revenue was growing but nobody could tell where or why
A complete intelligence stack for the whole business.
The work covered the warehouse, dashboards, and the custom tooling needed to get ServiceTitan data into usable form.
Real-time dashboards in Metabase
Technician Scorecard, Revenue Command Center, Good-Better-Best Analysis, and Membership Performance all update from live warehouse data instead of manual spreadsheet pulls.
A full ServiceTitan data pipeline
We built ServiceTitan -> Fivetran -> BigQuery infrastructure with Bronze, Silver, and Gold layers so the business had clean data it could trust.
A custom ServiceTitan MCP server
AI tools can query jobs, invoices, customers, dispatch, technicians, and more directly through a controlled interface tied to the business's real data.
A ServiceTitan CLI for instant access
Power users can pull jobs, revenue, estimates, memberships, and technician data directly from the terminal without clicking through the web UI.
Warehouse first, then dashboards and AI-ready access on top.
The stack was designed to make ServiceTitan data accurate, queryable, and usable across both human dashboards and machine interfaces.
The pipeline runs ServiceTitan -> Fivetran -> Google BigQuery -> Metabase. Bronze holds raw immutable source data, Silver handles cleansing and deduplication, and Gold exposes business-ready marts optimized for real reporting patterns.
On top of the warehouse, we built two open-source tools: a Model Context Protocol server that lets AI assistants query ServiceTitan data directly, and a CLI with 46+ commands for instant access to jobs, revenue, technicians, estimates, memberships, and more.
Revenue accuracy was cross-verified against ServiceTitan's own dashboard. Numbers match, which made the outputs trustworthy enough for daily operating decisions.
ServiceTitan -> Fivetran -> BigQuery -> Metabase pipeline
Bronze/Silver/Gold data architecture
MCP server: 467 tools, 15 domains, published on npm as @rowvyn/servicetitan-mcp
CLI: 46+ commands, oclif framework, 201 tests passing
Dashboard-matched revenue via ServiceTitan's Reporting API
Hidden revenue and operating reality became visible.
The business moved from gut-feel reporting to a system that surfaces pipeline, close rates, technician performance, and revenue with source-verified accuracy.
The dashboards surfaced $1.3M-2.1M in unsold pipeline that was sitting in ServiceTitan the entire time, invisible. A 46% close rate was made visible for the first time, giving the sales team a concrete number to improve against.
The custom integration platform connected dispatch, invoicing, customer history, and technician performance into a single system the business fully controls. Two open-source tools came out of this work and are now available to every ServiceTitan customer.
Hidden revenue and operating reality became visible.
The business moved from gut-feel reporting to a system that surfaces pipeline, close rates, technician performance, and revenue with source-verified accuracy.
More from this engagement.
Start Your Conversation
If your business software is full of data you still cannot act on, let's fix that.