REAL ENGINE OUTPUT · FICTIONAL DATACompany site ↗
MODELINTEL / EVIDENCE TO ANSWER / REPRODUCIBLE LOCAL TEST

The top line barely moved.
The records tell another story.

RUNBIRD v2.5's actual Python analysis was run against the sample files below. This browser displays its resulting evidence, file fingerprints, and release gate. Click an investigation stage to see the complete chain.

Proof boundary: Actual local RUNBIRD engine output on fictional test data. This page does not run Python in your browser, connect to Power BI, certify business correctness or imply a customer outcome.
Baseline
230Record total
Candidate
236Record total
Net change
+6Misleading in isolation
Gross movement
114Across three changed records

01 / Input boundary

HASHED / READ ONLY

What would this mean for a buyer?

A leadership report is about to be published. The issue is not that one total moved by six units, but whether the underlying records, classification, and KPI definition have changed without an approved explanation.
Discuss a paid investigation ↗

RUNBIRD · Source-to-decision proof. Reproduce the output with the supplied input files and `evidence_to_answer.py` from RUNBIRD v2.5. Source hashes and outputs are preserved; run records do not establish real-world authorization or client outcomes.