Two Headcounts, Two Turnover Rates, One Board Presentation

The request for 'one number' is the trap. Headcount is not one number; it is whichever definition the question requires. Deliver a defined metric with a reconciliation bridge, not a compromise figure.

Next 10 days

1. Define each metric before counting: filled FTE (budget positions with an incumbent, excluding contractors), active employees (on payroll as of a stated date), and headcount (all persons with an employment record). Pick the definition that matches the Board's question - capacity and budget almost always mean filled FTE.
2. Build the bridge: 2,412 (HRIS) minus contractors miscoded, minus unpaid-leave records where the definition excludes them, plus/minus timing and effective-dating differences = 2,340 (payroll). The bridge itself is the deliverable.
3. Fix turnover for the presentation: state the denominator and the period. If 140 records lack separation dates, the HR number is an undercount - separations are invisible. Do not present it as the accurate one.

What to present

• One primary metric with its definition printed on the slide.
• A short, candid data-quality note and the remediation timeline. Boards tolerate bad data disclosed; they do not tolerate bad data discovered later by a reporter.

The root cause is architectural, not analytical: three sources merged without a system of record, a data dictionary, or ownership. Controls, in order of impact:

1. Designate a single system of record for each data domain - position and FTE from the budget/position control system, person and employment status from HRIS, pay from payroll. Reconcile rather than duplicate.
2. Publish a data dictionary. Define active, separated, on leave, contractor, and filled vs. authorized, with the responsible steward for each field.
3. Enforce effective dating. Point-in-time reporting is impossible without it; this is why the two systems disagree.
4. Mandatory field validation at the point of entry - separation date required to close a record; contractor flag driven by position type, not manual entry.
5. Quarterly reconciliation between payroll, HRIS, and position control, with a signed variance explanation and a threshold for escalation.
6. Change control - no report goes to the Board or an external body without a steward sign-off on definitions.

Assign ownership to named individuals, not to 'HR.' A governance charter without a name attached is a document, not a control.

Small denominators produce unstable rates, and a watch list built on them will generate false alarms and, worse, false confidence.

• Recompute. Four of 22 is 18% for the quarter, which annualizes to a number that will look catastrophic and mean very little. Use a rolling 12-month rate and report the raw count alongside it.
• Compare sensibly. Benchmark against divisions of similar size and job family, or use a three-year average for that division rather than the countywide rate.
• Classify the separations. Retirement, resignation, termination, transfer, and death are different phenomena. A division with two retirements is a succession issue; a division with four resignations in eight months is a management issue.
• Read the qualitative record. Exit interview themes, grievance volume, leave usage, and internal transfer requests out of the division are more diagnostic at this scale than the rate.

Action: keep the division on a watch list with a review trigger rather than a performance flag - e.g., three or more non-retirement resignations in a rolling 12 months triggers a structured conversation with the division head. Be explicit with leadership that a small-unit rate is a signal to investigate, not a finding.