Compare two reporting periods, identify material changes and return to the original evidence.
See what you will get
Profit grew. Did cash flow follow?
Synthetic teaching sample · not a real company / fund
Profit grew. Did cash flow follow?
Metric (RMB million)
Jan–Sep 2025
Jan–Sep 2026
Change
Revenue · §Income statement
1,000
1,120
+12.00%
Net profit · §Income statement
120
132
+10.00%
Operating cash flow · §Cash flow
100
92
−8.00%
Receivables · §Balance sheet
180
240
+33.33%
Cash flow / net profit
83.33%
69.70%
Weaker cash conversion
Synthetic teaching material: DEMO-CONSUMER, no ticker. Flow figures cover January 1–September 30; receivables are point-in-time balances at September 30. Management mentions delayed customer payments; aging and subsequent collections are missing, so the cause remains unresolved.
Keep the records in your workspace
Change table, source references and follow-up questions
report.mdevidence.csvfollow-up.md
Each run writes to a new reports/{run-id}/, preserving inputs and previous reports. Status distinguishes completed, needs-input and failed.
What to prepare
Company identity, two comparable reports and one question. Use readable PDFs or text with source locations.
Use the earnings review Skill to compare the two reports in inputs. Does operating cash flow support profit growth? Keep source locations and separate facts, management explanations, inference and missing evidence. Use only these files and write to a new report folder.
First ask your host to inspect the method version, inputs, output location and dependencies in read-only mode. Downloading, discovering a Skill and completing its sample are separate steps.
Its limits
Resolve currency, units and quarterly versus cumulative periods before calculating. One period cannot establish growth. Request readable text for scanned files. A management explanation is not verified causation.
No network retrieval, account connection or result upload by default. Model handling depends on your host and service settings; this is not a promise of fully offline processing.
Continue your work
At the next disclosure, continue from your own original reasons. Do not let AI invent a past thesis.
