Reconciling a Disputed Turnover and Engagement Dashboard
You are the people analytics lead for a 2,400-employee healthcare services company. The CHRO's monthly dashboard reports 18% annual turnover, 72% engagement, and 1.9% absenteeism. At an operating review, three regional VPs challenge the numbers: one says turnover is closer to 24% because contract nurses are included; another says the dashboard excludes 180 contract workers entirely; the third says engagement is meaningless because only 41% responded and the survey vendor changed. You find: voluntary separations 312, involuntary 96, average headcount 2,400; 11% of termination reasons blank; two business units calculate turnover as separations divided by beginning headcount, and one uses total headcount instead of average. First-year turnover is 34%, regretted turnover 9%, and high-performer turnover is 22% of exits. The CHRO wants a single 'source of truth' before the next board meeting in 30 days.
1. Reconcile the turnover figures. Which number should go to the board, and what data-quality issues must be fixed?
Read the worked answer
Using employees only and average headcount 2,400:
- Voluntary: 312/2,400 = 13.0%
- Involuntary: 96/2,400 = 4.0%
- Total: 408/2,400 = 17.0%
The 24% claim likely uses a lower beginning headcount or includes contractor separations. Including 180 contractors in the denominator without their separations understates employee turnover: 408/2,580 = 15.8%, which is not comparable. Contractors should be a separate contingent-workforce metric.
Data-quality issues: 11% missing termination reasons; inconsistent denominators; unclear whether regretted and high-performer exits are counted; survey response only 41%; vendor change broke the trend. The board number should be employee total turnover 17.0%, with voluntary 13.0% shown separately and a footnote on definition and excluded contractors. Do not present a single blended number. Flag first-year turnover 34% and high-performer exits as the real risk.
2. Design the metric governance and reporting fix for the next 30 days.
Read the worked answer
Create one data dictionary with: population (employees versus contractors), denominator (monthly average), separation types, effective date, and owner. Build calculations in one governed layer, not in each unit's spreadsheet. Add validation rules: termination reason required; separation date must fall in the period; headcount reconciles to payroll; duplicate IDs blocked. Reconcile monthly to payroll and finance.
Publish a dashboard with three tiers: board-level stable metrics, operating metrics by region, and diagnostic cohorts. Show trends and n-size; suppress small cells. For engagement, report response rate and mode effects; do not compare pre/post vendor change without a bridge study. Assign a metric owner and a change-control log. In 30 days, deliver a reconciled prior-year baseline and a one-page definition appendix; defer new metrics until the baseline is stable.
3. The VPs dispute the dashboard. How do you handle the operating review and what do you say about uncertainty?
Read the worked answer
Acknowledge the dispute as a definition problem, not a loyalty test. Bring a one-page reconciliation showing each VP's number, its denominator, and what it includes. Agree on the decision use first: if the question is how many employees are leaving, use employee total turnover; if it is how much labor capacity is being lost, show contractors separately.
Present uncertainty clearly: 41% survey response means the 72% engagement score has a wide nonresponse risk, not just sampling error, so label it provisional. For turnover, give the point estimate and a monthly range, and show cohort cuts for first-year, high-performer, and regretted exits. Commit to fixing the pipeline, not defending the old dashboard. Ask each VP to nominate a data steward. Follow up with a written metric charter so the next review debates actions, not arithmetic.