Internal dashboards that agree with the BI tool
If an internal dashboard disagrees with the official BI tool, people go back to the BI tool. I built a scheduled sync of BI and CRM data into Postgres and checked it until the totals matched, plus the SQL reports the team reviews performance with.
Why a comparison group matters, with made-up numbers
Measured effect: +6 points. The comparison group moved +12 too, so that part is taken out. This is the number worth acting on.
Fictional numbers. The method is the one I use: difference-in-differences against a comparison group
Getting the numbers right
Totals are rebuilt from the underlying data model rather than read off a finished view, and weighted the same way the BI tool weights them, so the two agree.
Reports that started as scripts I ran by hand became scheduled jobs once their definitions settled.
Decisions
- Store only what a page reads, so queries stay fast.
- One definition of each period, reused everywhere.
- Exclusions are listed on the page, never applied silently.
Before and after
| Before | After | |
|---|---|---|
| Internal totals against the BI tool | Not compared | Match |
| Refresh | Manual exports | Scheduled |
This describes internal work at qlub. Names, internal rules and business data are left out on purpose. If you want the detail behind any of it, email mikaeel2013@gmail.com and I will walk you through it.