Why Workforce Dashboards Show Different Numbers—and How to Reconcile Them

Two workforce dashboards can disagree because they count different people, use different dates, or apply different rules to the same records. A mismatch can also come from missing updates, duplicate rows, or an incorrect calculation. Before choosing a number, establish whether the reports are answering the same question.

Reconciliation means explaining the difference with evidence. It does not mean changing one total until it matches the other. This guide shows a practical method using fictional headcount reports, followed by a reconciliation record you can copy.

Write the question both reports should answer

Replace “What is our headcount?” with a definition that someone else can reproduce. For example:

How many distinct employees belong to the operations team at 5 p.m. on September 1, using the agreed local time zone, excluding contractors and people whose employment has not started?

Specify the treatment of leave, departures, multiple assignments, and transfers. A person with two jobs may count once in headcount and twice in an assignment report. Neither calculation is automatically wrong; the report labels and intended uses must make the distinction clear.

Workday's general headcount-reporting guide recommends defining the population, reporting date, and snapshot or historical scope. That is useful reporting practice, not a claim that every HR system uses the same rules. Workday: headcount reporting

Save the original reports before investigating

Keep dated copies of both reports, their filter settings, and their displayed totals. Record the source refresh time as well as the export time: downloading a report now does not prove its underlying data is current.

Use an authorized working copy containing only the fields needed for the comparison. For headcount, these might include a stable person identifier, worker type, team, assignment identifier, effective start and end dates, and record-update information. You do not need names or compensation to demonstrate the counting method.

If access restrictions hide part of the population, record that limitation. Ask the report owner for a permitted reconciliation view rather than treating unseen records as absent.

Compare definitions before comparing people

Check these fields side by side. They often reveal why two apparently identical totals differ.

The UK Government Data Quality Framework distinguishes completeness, uniqueness, consistency, and timeliness. Those are useful checks here: a current dataset can still contain duplicates, and two consistent totals can still omit the same people. The framework treats quality as fitness for the intended purpose. Government Data Quality Framework

Pay particular attention to dates. An effective date describes when a change applies; an update timestamp describes when it was entered or received. They can differ. A late-entered correction may alter a historical snapshot even though the reporting date stays the same.

Agree on boundary rules with the data owner. A field called “end date” might mean the last included day or the first excluded day. Do not guess which interpretation the source uses.

Match the underlying records

Apply the agreed definition to copies of both datasets, preserving the original report values. Use a stable identifier with a verified mapping between systems. Names alone are a poor basis for matching people who share a name or whose names have changed.

These filtered working copies, counted under the agreed definition, are the normalized reports. Separate their results into three groups:

• People included in both normalized reports.
• People included only in report A.
• People included only in report B.

For each difference, record the relevant identifier, the reason, and supporting evidence. Keep unresolved matches separate. Do not merge two records merely because they look similar, or discard a second assignment without first deciding whether the intended measure counts people or jobs.

If you only have aggregate totals, you can investigate definitions and refresh times, but you cannot establish which people explain the difference. Ask for an appropriate breakdown before declaring the reports reconciled.

Worked example: 120 and 105 both reconcile to 108

All numbers and identifiers in this example are fictional. The target is distinct operations employees at the stated 5 p.m. cutoff. The agreed definition includes employees on leave, excludes contractors, and excludes starts after that cutoff.

Report A displays 120 rows. Its source was refreshed at 5:30 p.m. and includes assignments, contractors, and future starts. Report B displays 105 distinct employees, but its source was last refreshed at 8 a.m.

Three employees started at 9 a.m.; their records were entered and became available to reporting at 10 a.m. They are absent from B's 8 a.m. extract and present in A's later one. Both normalized copies use the 5 p.m. effective cutoff, including when filtering A's newer data. No other relevant roster changes occurred between 5 and 5:30 p.m.

The data owner verifies the following differences against the underlying roster and effective dates. The categories in the bridge below do not overlap.

Report A's calculation is 120 − 6 − 4 − 2 = 108. Report B's is 105 + 3 = 108. Together, the explained adjustments account for the original difference of fifteen.

The extra assignments are legitimate assignment records, not necessarily bad source data. They are removed from the person count only because the target measure is distinct employees. Likewise, future starters may belong in a hiring pipeline report, even though they do not belong in this snapshot.

The three hires are included in B's normalized working copy only after verification. The original report remains 105, with its earlier refresh time recorded. Its owner can refresh or rebuild it using the agreed cutoff; an analyst should not silently overwrite its displayed total.

Matching totals are not enough

Suppose two reports each show 108, but one includes an employee the other omits and excludes someone the other includes. The totals match while the membership differs.

In this fictional example, the final identifier check confirms the same 108 people in both normalized reports, with no unresolved differences. That record-level check, together with the verified definition and roster, supports the reconciliation. Equal totals alone would not.

Decide whether to fix a report or explain a valid difference

If the reports serve different purposes, preserve those purposes and label them clearly. “Active assignments including contractors” and “distinct employees at the cutoff” should not be presented as interchangeable headcounts.

If the intended definition is the same, give each discrepancy an owner and a correction path. A missing update may require a refresh; an incorrect team mapping may require a source correction; a row count used instead of a person count may require a calculation change. Re-run the comparison after the fix and retain the explanation.

Do not average conflicting totals or select the number that looks more plausible. If the difference remains unresolved, report the two values with their definitions and the specific open issue. CIPD's people-analytics guidance recommends auditing data quality and gaps before analysis. CIPD: people analytics

Copy the reconciliation record

Use one record for each metric and reporting cutoff. Keep detailed person-level evidence in the appropriate restricted location; the summary can reference it without reproducing personal information.

The useful outcome is a number someone can reproduce and an explanation someone can inspect. Start with one disputed metric, preserve the originals, and work through the differences until each is explained or explicitly left open.