SYNTHETIC DATA · Synthetic example. Fictional data. Northwind Ops Consulting, its projects, clients and staff are invented for demonstration.

Project hours reconciliation, August 2026

Northwind Ops Consulting (fictional) · Inputs: project_status.csv and time_entries.csv · Generated 2026-09-29 08:47

Five agreed metrics

24Open projects

15Completed projects

1019.25Total recorded hours

6.50Unmatched hours

1Unresolved-status records

Each figure was compared with a separately computed expected value (different script, different method) before this report was allowed to be written:

Metric	This report	Expected (independent)	Result
Open projects	24	24	match
Completed projects	15	15	match
Total recorded hours	1019.25	1019.25	match
Unmatched hours	6.50	6.50	match
Unresolved-status records	1	1	match

Reconciliation: every input row accounted for

Time entries (time_entries.csv)	Rows	Hours
Rows received	300	1020.75
Less: exact duplicates removed (rule R2, see change log)	1	1.50
Total recorded (after duplicate removal)	299	1019.25
Matched to a project	297	1012.75
Unmatched: project ID not in status export	1	4.00
Unmatched: project ID blank	1	2.50

Projects (project_status.csv)	Rows
Rows received	40
Open	24
Completed	15
Unresolved status (needs a decision)	1
Accounted for	40

Exception register: decisions needed

Nothing below was guessed or silently dropped. Each item stays open until the client decides.

ID	Type	Where	Record	Detail	Hours	Decision needed
EX-01	unmatched project id	time_entries.csv row 299	TE-00297	project_id P-1099 is not in project_status.csv	4.00	Is P-1099 a new project missing from the status export, or a mistyped ID?
EX-02	missing key	time_entries.csv row 300	TE-00298	project_id is blank (2026-08-20, Staff E)	2.50	Which project should this entry be charged to?
EX-03	conflicting status	project_status.csv row 13	P-1012	marked Complete on 2026-08-14 but 2 time entries (4.50 h) dated after it: TE-00295 2026-08-19, TE-00296 2026-08-26	4.50	Is P-1012 complete (late entries to be re-dated/billed) or reopened?

Changes applied by rule

Where	Record	Field	Before	After	Rule
project_status.csv row 6	P-1005	status	`open`	`Open`	R1: normalised
project_status.csv row 25	P-1024	status	`COMPLETE`	`Complete`	R1: normalised
time_entries.csv row 183	TE-00058	(whole row)	`TE-00058,P-1033,Staff B,2026-08-10,1.50,Analysis`	`(removed)`	R2: removed exact duplicate of row 59

Rules used

R1 Status text is trimmed and case-normalised to Open or Complete. Any other value (including blank) is unresolved.
R2 An exact duplicate time row (every column identical) is removed; the first occurrence is kept.
R3 A time row whose project_id is blank, or not present in the status export, is unmatched. Its hours stay in total recorded hours and are reported as unmatched hours.
R4 A project marked Complete with hours dated after its completion date, or with no completion date, is unresolved. It is not counted as open or completed until the client decides.
R5 A project_id that appears more than once in the status export is unresolved.

Input fingerprints (SHA-256, first 12): project_status.csv 80b4c4caf83a · time_entries.csv eb00bd406dd1. Row-level detail: reconciled_time_entries.csv, reconciled_projects.csv, change_log.csv, exceptions.csv. This is a demonstration on fictional data. It shows the checks performed; it is not evidence of results for any real business.