Most employee engagement dashboards have the same problem: they're full of metrics that are easy to collect rather than metrics that are useful to act on. Survey scores, attendance, training completion — all measurable, all quarterly or annual, all telling you about a state of affairs that has already settled.
This is a list of twelve engagement KPIs worth tracking, each with its formula and — the part usually left out — the data source it actually requires. That last column is where most measurement plans quietly die: the KPI is fine, but nobody can produce the number without a two-week manual exercise, so it gets tracked once and abandoned.
Outcome KPIs (Lagging)
These tell you what already happened. They're the ones finance cares about and the ones that show up too late to prevent anything.
1. Voluntary turnover rate
Formula: (Voluntary departures ÷ Average headcount) × 100, annualized.
Source: HRIS. Available to nearly everyone.
Why it matters: The single most consequential engagement outcome, and the one with a dollar figure attached. Track voluntary separately from total — involuntary departures are a management decision, not an engagement signal, and blending them hides the trend. Full calculation walkthrough here.
2. Regrettable attrition rate
Formula: (Departures you'd have preferred to keep ÷ Total voluntary departures) × 100.
Source: HRIS plus a manager judgment field.
Why it matters: A 12% turnover rate where none of it was regrettable is a healthy company. A 7% rate that's entirely your top performers is an emergency. The headline rate can't tell these apart. Requires discipline — the flag must be set at departure, not reconstructed later.
3. First-year attrition
Formula: (Departures with tenure < 12 months ÷ Hires in the same period) × 100.
Source: HRIS.
Why it matters: Isolates onboarding and hiring-accuracy problems from general engagement problems. High first-year attrition alongside low overall attrition means you're recruiting or onboarding badly, not managing badly — a completely different fix. See the first 90 days.
4. Cost of turnover
Formula: Departures × Average salary × Replacement cost multiplier.
Source: HRIS + payroll.
Why it matters: Converts everything above into the only unit executives reliably act on. Standard multipliers: 33% of salary at the low end (Work Institute), 50% typical (SHRM), 150% for senior or specialized roles (Gallup). Worked example: 100 employees × $65k average × 15% turnover × 50% = $487,500 a year. Run it on your own numbers.
5. Internal mobility rate
Formula: (Internal moves and promotions ÷ Average headcount) × 100.
Source: HRIS.
Why it matters: A direct read on whether "career growth" is a real thing here or a slide in the onboarding deck. Lack of growth is one of the most cited reasons for voluntary exit; this is the metric that catches it before the exit interview does.
Sentiment KPIs (Periodic)
What people say when asked. Cheap, comparable, and limited by cadence.
6. eNPS
Formula: (% rating 9–10) − (% rating 0–6).
Source: Quarterly one-question survey.
Why it matters: The standard summary number, and genuinely useful as a trend. Report the three-way promoter/passive/detractor split alongside it — the same score describes very different companies. Benchmarks and pitfalls here.
7. Engagement survey participation rate
Formula: (Responses ÷ Invited) × 100.
Source: Survey tool.
Why it matters: An underrated meta-metric. Falling participation is itself a disengagement signal, and it invalidates everything else on the survey. Below roughly 60%, treat the results as directional at best. If it drops sharply, the most likely cause is that people didn't see anything change last time.
8. Manager effectiveness score
Formula: Mean of 3–5 manager-specific survey items, reported by team.
Source: Engagement survey, segmented.
Why it matters: Gallup attributes roughly 70% of the variance in team engagement to the manager. If you track one segmented metric, this is the one — a company-wide engagement average is mostly a weighted average of manager quality wearing a disguise. Requires team sizes large enough to preserve anonymity (8+).
Behavioral KPIs (Leading)
What people actually do, observed continuously rather than sampled quarterly. These are the ones that let you intervene while intervention is still possible — and the ones most engagement dashboards are missing entirely.
9. Recognition frequency
Formula: Recognition events ÷ Employees ÷ Month.
Source: Recognition platform.
Why it matters: A voluntary behavior nobody is required to perform, which is precisely what makes it a good signal. Employees who don't feel adequately recognized are about twice as likely to say they'll quit within a year (Gallup / Workhuman). Track the trend per team; a team whose recognition volume halves over six weeks is telling you something a quarterly survey won't for another two months.
10. Recognition reach (participation breadth)
Formula: (Employees who received recognition at least once ÷ Total employees) × 100, per month.
Source: Recognition platform.
Why it matters: Volume without reach is concentration — a program where the same eight extroverts collect everything is generating activity and inequality simultaneously. Reach is the metric that catches the people going invisible, who are disproportionately the ones who leave. More on diagnosing and fixing that skew.
11. Cross-team interaction rate
Formula: (Recognition or collaboration events crossing a team boundary ÷ Total events) × 100.
Source: Recognition platform or collaboration tooling.
Why it matters: A quantitative read on siloing, which is otherwise measured entirely by vibes. A declining cross-team rate usually precedes the coordination failures leadership notices two quarters later. See breaking down silos with recognition data.
12. Individual activity drop-off
Formula: Count of employees whose recognition given/received fell below their own 90-day baseline for 3+ consecutive weeks.
Source: Recognition platform.
Why it matters: The closest thing to a genuine early-warning metric on this list. It's measured against each person's own baseline rather than a company average, which matters — quiet people aren't disengaged, but quiet people who used to be loud are worth a conversation. The mechanics, including how to avoid using it as surveillance.
Putting It on One Dashboard
| Layer | KPIs | Cadence | Answers |
|---|---|---|---|
| Behavioral | 9–12 | Continuous | Who is disengaging right now? |
| Sentiment | 6–8 | Quarterly | How do people feel about it? |
| Outcome | 1–5 | Monthly / annual | What did it cost? |
The common failure is building only the bottom two rows. That produces a dashboard that is accurate, well-presented, and incapable of preventing anything — every number on it describes a situation that finished forming weeks ago.
Four Rules for Not Wasting the Effort
- Pick six, not twelve. A dashboard nobody reads has zero value regardless of rigor. Two from each layer beats all twelve tracked badly.
- Segment or don't bother. Company-wide averages hide the exact variation you need to find. A +25 eNPS overall can be +45 in engineering and -10 in support, and the average tells you to do nothing.
- Never tie these to individual performance reviews. The moment recognition frequency becomes a manager's scorecard number, you get recognition inflation and the metric stops measuring anything. This is the fastest way to destroy a behavioral KPI.
- Write down the intervention before you track the metric. If you can't name what you'd do when the number moves, you've built a report, not a KPI.
Where the Data Comes From
KPIs 1–5 come from your HRIS and are available to essentially everyone. KPIs 6–8 need a survey tool, which most companies over 50 people have. KPIs 9–12 need a recognition platform that exposes its data — and that's the gap in most stacks.
Our bias, stated plainly: we build Propsly, a Slack-native recognition tool, and its analytics produce KPIs 9–12 directly, so we would obviously highlight the layer we happen to sell. The argument we'd make without the product is just that the behavioral layer exists and most dashboards skip it. Propsly is free for unlimited users with the recognition mechanics and leaderboards; the analytics that produce reach, cross-team, and drop-off metrics are on the $50/month flat Pro tier. Plenty of teams start with a shared #wins channel and count by hand — that works too, and it's how you find out whether the metric is worth automating.
The Short Version
Engagement measurement fails in a predictable way: the metrics are accurate, the cadence is too slow, and by the time a number moves, the person it was describing has already resigned.
Keep the outcome metrics — they're how you make the business case. Keep eNPS — it's cheap and comparable. But put at least one continuously-updating behavioral metric next to them, segment everything by team, and decide in advance what you'll do when each one moves. That's the difference between measuring engagement and managing it.