Case studiesLeadership reporting
B2B software platform, Toronto
One set of numbers, read by the people who own them.
The leadership team ran reporting from a spreadsheet stitched together out of platform, CRM and Xero exports. Now a governed warehouse and a custom dashboard match their own finance model to the cent, and show each person only what their access allows.
total revenue against their own finance model, in all four closed FY26 quarters, now built from a governed warehouse.
- Data and BI
- Applied AI
- RevOps
- Company
- B2B software platform, Toronto
- Systems
- CRM, Xero, their product platform
- Work
- Data and BI, Applied AI, RevOps
- Engagement
- Data foundation, since July 2026
In their words.
Paraphrased from working sessions.
- Leadership, on the old reporting cycle
It’s super manual, and it takes us forever.
Exports from the platform, the CRM and Xero went into Excel, part of the revenue waterfall was keyed by hand, and teams got different values for the same metric.
What we found.
- 3×
Two different measures, funnel conversion and win rate, were reported under one name. They differed by roughly three times.
- 57
source tables were already tagged by sensitivity. Their access model was good, so we adopted it as it was.
- 13 of 33
plan metrics were waiting on decisions, not engineering. The hard part was definitions, not arithmetic.
What we built.
Quick wins first, then the bigger pieces, all in their own accounts.
A governed warehouse
A dbt model on BigQuery, fed by Fivetran from the CRM, the product platform, Xero, Jira, Google Analytics and Planhat, with identity resolved across systems.
- BigQuery
- dbt
- Fivetran
Access that follows the viewer
People sign in with their own Google accounts and every query runs as them, so BigQuery’s own permissions are the only gate. A CI check blocks any build that would expose a protected column.
- Google sign-in
- CI checks
- Live audit
A dashboard and a definitions register
A custom leadership dashboard on their own hosting, where every number traces back to its source, plus a register that gives every metric an owner.
- Next.js
- Metric owners
- Drill-through
What changed.
- To the cent
total revenue against their own finance model, in all four closed FY26 quarters.
- 36 → 4
The metric review went from a 36-sheet package to four questions.
- 0
protected-data failures in the live audit, across 122 tables.
The dashboard runs on their own hosting, and every number on it traces back to its source.
The stack
The stack, all in their accounts
- BigQuery
- dbt
- Fivetran
- CRM
- Xero
- Next.js
- Google Workspace
Bring the report nobody trusts.
On the first call we’ll trace one number back to its sources with you and show you where it splits.