Two business reports can be individually correct and still lead to conflicting conclusions because they may measure different things. They can use different cut-off dates, definitions, entity mappings, filters or source refresh times. Comparing the totals without comparing those rules creates false disagreement.
The answer is not to choose the report that looks more authoritative. It is to establish what each number means and whether the measures are comparable for the decision being made.
Four reasons reports diverge
| Cause | Example | Diagnostic question |
|---|---|---|
| Definition | One report uses invoiced revenue; another uses recognised revenue | Are the business definitions identical? |
| Time | One report uses invoice date; another uses posting date | Are the cut-off rules aligned? |
| Identity | Duplicate customer IDs split or combine transactions | Are the entities mapped consistently? |
| Freshness | One source refreshed overnight; another refreshed at noon | Are the sources complete for the same point in time? |
| Filters | Returns or cancelled orders are excluded in one report | Are the filters visible and equivalent? |
| Calculation | Currency conversion or rounding differs | Are calculation rules and reference rates aligned? |
A discrepancy should be treated as a diagnostic signal. It may reveal a data defect, but it may also reveal that the organisation has never agreed on one business definition.
A disciplined discrepancy investigation
- Start Two figures disagree
- Confirm the decision and metric definition
- Align date range and source freshness
- Compare filters and calculation rules
- Check entity mapping and duplicate handling
- Trace totals to source records
- Comparable and reconciled?
- Yes Explain the variance.
- No Document the limitation and assign an owner.
Comparing row-level records before aligning the definition produces work without resolving the real disagreement.
The sequence matters. Comparing row-level records before aligning the definition can produce a great deal of work without resolving the actual disagreement.
Create a metric contract
A practical metric contract records the meaning of a business measure in terms people can test.
| Field | Example of what to document |
|---|---|
| Name | Net supplier spend |
| Business purpose | Procurement category analysis |
| Owner | Named finance or procurement role |
| Calculation | Defined transaction value less specified credits |
| Date basis | Invoice posting date |
| Currency | Approved reporting currency and conversion rule |
| Exclusions | Cancelled transactions, duplicate records and specified taxes |
| Source coverage | Systems and periods included |
| Version | Effective date and change history |
| Known limitations | Delayed credits or incomplete supplier mappings |
This is more than a glossary entry. It connects the business meaning to the calculation, source data and governance process.
Why this matters for AI-driven analysis
A natural-language interface cannot resolve a disagreement that the organisation has left undefined. If “revenue” has three legitimate meanings, the system should ask which one the user intends or apply a clearly governed definition, not silently choose.
Qwry.AI’s platform direction includes business context and query-backed answers. Those capabilities are most useful when definitions, source coverage and permissions are explicit. See the platform overview for the broader approach.
Next step: Choose one metric that causes recurring debate. Compare the definitions, filters, time bases, source freshness and entity rules behind each report before attempting to reconcile the numbers.
FAQ
Frequently asked questions.
The answer depends on the business measure and the system authorised to own it. "Source of truth" should be assigned by data domain and metric, not assumed for every question.
No. Data quality concerns the condition and suitability of data. Metric consistency also requires agreed definitions, filters, time bases and calculation logic.
It can help standardise definitions and calculations, but it cannot by itself resolve missing source data, stale feeds, incorrect identity matches or unclear business ownership.
Record the owner, effective date, reason and version. Decide whether historical figures should be restated or remain calculated under the earlier definition.