SYNTHETIC DIAGNOSTIC CASES. Not customer data or a RowSpeak run.
These 11 records are separate troubleshooting exercises, NOT an additive ledger.
U1 and U2 describe the same September revenue: 125.4 thousand USD = 125400 USD.
U3 has unknown units. Hold it; do not infer a factor from the numeric value.
P1 and P2 describe the same rate: 0.12 fraction = 12 percent. Do not add rates.
P3 lacks a unit. Hold it; 12 could be a count, percent or ratio.
C1 and C2 are distinct currency subtotals. Keep USD 1000 and EUR 900 separate.
D1 and D2 describe the same T900 on different date bases, not two sales.
D3 has an ambiguous date: March 4 vs April 3. Hold pending locale and basis.
There is intentionally NO combined grand total across these cases.
INPUT: input-cases.csv. ANSWER KEY: expected-decisions.csv.
Stress test: change U1 to 126.4; normalized value becomes 126400, so the
comparison with unchanged U2 has a genuine 1000 USD difference.
Date and currency policies require source-owner approval in real work.
Verified using Python Decimal; these are illustrative review decisions.
