Example deliverable

What a pilot report looks like.

Synthetic example. This is not a customer result, live-trading result, or finding against a named product.

A report separates what was observed, how it was judged, what changed, and what remains unknown. It never turns a seeded fault into a customer defect.

Executive result

Path	Synthetic U.S. equity paper-order input
Window	20 expected sessions
Primary source	Synthetic source A
Reference source	Synthetic source B
Independence	Assumed for demonstration only
Evaluated observations	5
Decisions	2 permit, 2 block, 1 escalate
Production actions changed	0

Decision summary

ID	Evidence	Decision	Reason
SYN-001	PASS	PERMIT	Twenty sessions, monotonic dates, internally consistent OHLC, and closing-price deviation below 10 basis points.
SYN-002	FAIL	BLOCK	The primary series returned 12 of 20 expected sessions.
SYN-003	FAIL	BLOCK	The closing price on session 14 differed from the reference by 2.00%, above the fixed 0.10% threshold.
SYN-004	INDETERMINATE	ESCALATE	Both sources were readable, but their upstream lineage was unknown.
SYN-005	PASS	PERMIT	One extra response field did not change the evaluated values.

Example discrepancy

decision_id       SYN-003
instrument        SYNTHETIC-XYZ
session           2026-09-14
primary_close     102.00
reference_close   100.00
deviation         2.00%
threshold         0.10%
evidence_status   FAIL
observe_decision  would BLOCK

The customer technical owner classifies the result as a confirmed issue, expected difference, false positive, or unresolved. A confirmed discrepancy becomes a deterministic regression fixture with the same instrument, date, values, and threshold.

What this report would not claim

It would not say the primary source lied.
It would not call two sources independent without lineage evidence.
It would not treat a receipt signature as proof that the market fact was true.
It would not call a seeded fault a customer defect.
It would not extrapolate five observations into a reliability score.

The final recommendation

Redundant on the measured path: no confirmed discrepancy and no new control information.
Continue Observe Mode: useful signal exists, but false positives, lineage, or operational limits need more evidence.
Test bounded enforcement: confirmed discrepancies exist, false-positive behavior is acceptable, and the customer chooses one low-risk path for a separate test.

See the pilot scope

Read the method →