03 · Data Quality + Root Cause
When the numbers don't match.
Tracing an ~$150K difference between executive sales reporting and analytics reporting back to how transaction edge cases were being handled.
Signal
Reports diverged
Investigated
PQV + orders
Root causes
Returns + exchanges
Variance
~$150K
Executive Flash
Order population
Returns
Exchanges
Exclusions
The question
Why did the executive BOD Flash and the analytics view disagree when both were intended to represent the same sales performance?
The investigation
I decomposed the populations and compared order logic, PQV definitions, returns, exchanges, cancellations, test populations, and warehouse exclusions rather than assuming the difference was a simple SQL error.
The finding
The investigation traced the material variance to exchange-related handling and differences in how edge cases were represented across the reporting logic.
The lesson
Reconciliation is not just checking whether two numbers match. It is understanding whether the two systems are answering the same business question, using the same population and definitions.
Confidentiality note
The underlying company data is confidential; this public case study describes the analytical method and uses only generalized figures.
← Back to all work