The most visible data-quality failure is two reports quoting different numbers for the same thing. A reader who spots it begins to doubt every figure in both documents, and the reporting function loses authority. Reconciliation is the deliberate practice that prevents it – and it is more about a defined basis than about arithmetic.
This guide covers how to reconcile numbers across reports: why they diverge, how to reconcile deliberately, and how to record the reconciliation so it is not re-litigated. It is the companion to Data Quality for Reporting.
When two systems disagree, averaging them is the worst thing you can do.
Why numbers diverge
Divergence almost always traces to one of four causes.
- Different sources. The finance system and the CRM count revenue on different bases.
- Different definitions. “Active customer” means two things in two teams.
- Different periods. One report uses the calendar month, another the 4-4-5 close.
- Different treatments. One includes a restatement; the other does not.
Nearly every “the numbers are wrong” dispute is really one of these four. Naming the cause is most of the reconciliation.
The reconciliation method
When two numbers disagree, reconcile deliberately rather than averaging.
- Identify the difference – amount, direction and likely cause.
- Trace both numbers to their sources and definitions.
- Decide the authoritative basis for this report, and document why.
- Adjust or footnote the other, rather than blending them.
- Record the reconciliation, so next cycle applies the same rule.
The point is not which number is “right” in the abstract, but that the report uses a defined basis consistently and can explain any difference on request.
A worked reconciliation
Suppose the finance system shows revenue of $4,200,000 for the month, while the CRM shows $4,260,000.
- Difference: $60,000, with the CRM higher.
- Trace: finance is invoiced revenue; the CRM is booked revenue, including a $60,000 deal signed but not yet invoiced.
- Authoritative basis: finance, because the report is on invoiced revenue.
- Footnote: note the booked-versus-invoiced difference if it is material to the reader.
- Record: so the next cycle applies the same rule.
The reconciliation took minutes and settled a question that would otherwise recur every month.
Reconciling to the prior cycle
Reconciliation is not only across reports; it is also across time.
- Compare to the prior period for each headline figure.
- Explain any change, rather than letting a number drift silently.
- Distinguish a genuine change from a restatement.
- Carry the explanation forward, so a returning reader is not surprised.
A recurring report that reconciles across time catches drift that a single-cycle check would miss.
Recording the reconciliation
A recorded reconciliation is what stops the same dispute recurring.
- Keep a short log – figure, sources, difference, basis chosen.
- Link it to the metric dictionary, so the definition is captured where the metric lives.
- Review it when a definition changes, so the reconciliation stays valid.
- Reference it when a reader asks why two reports differ.
Without the record, reconciliation is repeated by memory, and the same arguments return each cycle.
Defining the authoritative basis
The reconciliation is only as good as the basis you choose.
- Name the basis – invoiced versus booked, gross versus net, calendar versus close.
- Record it in the metric dictionary, so the definition travels with the metric.
- Apply it consistently across every report and every cycle.
- Revisit it deliberately if the business changes, rather than by accident.
A basis chosen once and recorded is a rule; a basis chosen each time is a source of dispute.
Reconciling across time
Reconciliation across reports is half the job; reconciliation across time is the other half.
- Compare each headline figure to the prior period, and explain any change.
- Distinguish a genuine change from a restatement of an earlier figure.
- Carry the explanation forward, so a returning reader is not surprised.
- Flag a restatement explicitly, and reconcile it to the report that used the old figure.
Drift across cycles is quieter than a mismatch between reports, and it is the kind of error a reader notices only when it becomes large.
When the difference is legitimate
Not every difference is an error. Some are legitimate and should be explained, not eliminated.
- Different periods – one report is month-to-date, another is full month.
- Different entities – one includes a subsidiary the other excludes.
- Different bases – one is gross, one is net.
- Different timing – one reflects booked, one reflects invoiced revenue.
Where a difference is legitimate, the fix is not to change the number but to state the basis clearly enough that the reader does not mistake it for a discrepancy.
A minimal reconciliation routine
Enough to settle most disputes.
- Pick the figure that differs, and name each report’s basis for it.
- Identify whether the difference is period, entity, basis or timing.
- Where it is an error, trace it to the source and correct it.
- Where it is legitimate, state the basis so the reader is not misled.
- Record the reconciliation, so the next cycle starts from a known position.
The routine is short. Its value is that it happens every cycle, not only when a reader complains.
Common mistakes
- Averaging the two numbers. It hides the cause and produces a figure neither source supports.
- Fixing the report, not the disagreement. The divergence returns next cycle.
- No documented basis. The same reconciliation is re-litigated every month.
- Ignoring period differences. Calendar versus close cycles cause frequent, avoidable conflicts.
- Reconciling only across reports. Drift across cycles goes unnoticed.
Frequently asked questions
Why do numbers differ between reports?
Almost always because of a different source, a different definition, a different period, or a different treatment such as a restatement. Naming the cause resolves most disputes.
Should you average two conflicting numbers?
No. Averaging hides the cause and produces a figure neither source supports. Identify the authoritative basis and footnote the difference instead.
How do you decide which number is authoritative?
By which basis the report is on – invoiced versus booked revenue, for example – and by the metric dictionary. Document the choice so it is consistent across cycles.
How do you stop a reconciliation being re-litigated every month?
Record it: the figure, the sources, the difference and the basis chosen. Link it to the metric dictionary so the rule travels with the metric.
Is reconciliation the same as a data-quality check?
It is one of the checks. Reconciliation across reports and cycles is a subset of broader data-quality checks covering accuracy, completeness, timeliness, consistency, validity and uniqueness as well.
What is the authoritative basis?
The basis the report is on – invoiced versus booked revenue, gross versus net, calendar versus close – chosen deliberately, recorded in the metric dictionary, and applied consistently.
How do you reconcile across time as well as across reports?
Compare each headline figure to the prior period, distinguish a genuine change from a restatement, carry the explanation forward, and flag restatements explicitly.
What if the difference between reports is legitimate?
Then explain it rather than eliminate it – different periods, entities, bases or timing. State the basis clearly so the reader does not mistake it for a discrepancy.
How do you decide which report is right?
By the named basis, not by which report you happen to prefer. The basis is agreed in the metric dictionary, and both reports should conform to it or state their deviation.
How often should you reconcile?
At least every cycle, against the prior period and any cross-report figure. A reconciliation left to the year-end is a surprise waiting to happen.
What causes the same number to differ across reports?
Different periods, entities, bases or timing – and, less often, a genuine error. Naming the basis each report uses resolves most apparent discrepancies without changing a figure.
Should both reports change, or only one?
If the basis is the problem, one or both reports should state the basis clearly. If one report used the wrong figure, the error is corrected at its source, and the metric dictionary should have prevented it.
What if a difference cannot be explained?
Escalate it before the report is issued. An unexplained difference is either an error or an undocumented basis, and either way it should be resolved rather than presented to a reader.
Can software reconcile reports automatically?
It can compare figures and flag differences. It cannot judge whether a difference is a legitimate basis or an error – that remains a human call, informed by the dictionary.
Next step
Reconcile deliberately, never average, and record the basis so the same dispute does not return. See Building a KPI & Metric Dictionary for where definitions live, and Data-Quality Checks for Report Production for where reconciliation fits.
Sources
- Harvard Business Review (Nagle, Redman and Sammon, 2017): the prevalence of critical errors in new data records.
- Data-quality practice: reconciliation across sources, definitions and periods as a standard reporting control.
Good-practice claims are cited from their sources; no statistic in this article is invented.