M360 Compass · Management insights
HR metrics: which numbers support better decisions?

HR metrics become useful when they change a decision. A small business can fill a dashboard with numbers while leaving its most pressing questions unanswered: which role needs clarification, which hire needs attention and which development conversation should happen next? Start with these decisions, then choose the measures that can inform them.
The eight measures below include working definitions and practical limitations. The proposed method supports a short monthly review. It does not rank employees or assume your software already calculates every formula. You can begin with two reliable measures and expand once their usefulness is clear. A smaller dashboard with accountable owners is easier to maintain than a large collection nobody uses.
Connect HR metrics to a specific decision
For each measure, complete this sentence: “If this number changes, we will decide whether to…” If you cannot identify an action, set the measure aside temporarily. A longer hiring cycle might prompt a review of interview scheduling; it does not automatically justify spending more on advertising.
Statistics Canada’s quality guidelines describe dimensions including relevance, accuracy, timeliness, interpretability and coherence. Apply a simple lesson to your own dashboard: pair each number with a stable definition, a known reporting period and a visible limitation.
Create a short measure card containing the management question, formula, population, source, reporting frequency, owner and possible decision. Also identify who needs access. A team of three cannot support the same reporting breakdowns as a workforce of three hundred. Avoid unnecessary detail that exposes an individual situation when an operational summary would answer the question.
Four measures for recruitment and stability
1. Time to fill a position. Count days from approval to open the role until acceptance of the offer. Use the median for positions closed during the reporting period and display the number of cases. Compare similar roles. Possible decision: remove delays at an internal approval or scheduling step.
2. Offer acceptance. Divide accepted offers by offers that received a response, then multiply by one hundred. Keep outstanding offers separate. Possible decision: check whether role expectations were explained early enough. A declined offer still needs context; the ratio does not tell you why the person declined.
3. Voluntary departures. Divide voluntary departures during the period by average headcount for that same period. State how you calculate average headcount and label the reporting period. Possible decision: investigate recurring explanations with the appropriate managers. A turnover rate alone cannot establish the cause of any departure.
4. Presence after initial onboarding. Within a hiring cohort, count employees still present at a chosen milestone, such as three months. Include only people who have reached that milestone. Possible decision: review orientation or role expectations. Show the underlying counts, particularly when only a few people were hired.
Four measures for execution and development
5. Clarity of critical responsibilities. Count priority responsibilities with a documented owner and decision-maker, then divide by the total reviewed. Define what “documented” means before counting. Possible decision: resolve unclear ownership before scheduling more meetings to compensate for it.
6. Individual check-ins held. Divide completed one-on-one meetings by those actually scheduled for the period. A postponed meeting still requires a replacement date; a cancellation is not a completed conversation. Possible decision: protect management time. This measure tracks a practice, not the quality of the discussion.
7. Development actions practised. Count development-plan actions that included a real practice opportunity and compare them with actions due. Possible decision: simplify an overambitious plan or arrange access to a suitable assignment. A checked box without an example of practice is weak evidence of progress.
8. Commitments completed by the agreed date. Divide completed commitments by commitments due in the period. Retain original dates and explain changes. Possible decision: reconsider capacity or dependencies. Repeatedly moving deadlines to improve the number hides the scheduling problem rather than resolving it.
Fictional example: turn eight numbers into two decisions
Fictional example. Boreal Workshop is an invented twenty-eight-person business in Sherbrooke. All figures and outcomes in this example are illustrative.
Three recently filled roles took eighteen, thirty and forty-two days. Their median is thirty days. With only three cases, the owner avoids claiming a general trend. Instead, the team reviews the stages in each hiring process. Two applications waited a long time for the final interview because nobody had reserved time in the interviewers’ calendars.
The first decision is to reserve interview windows before opening the next position. The operations manager owns the change and will verify that it happens during the next recruitment. The team continues measuring time to fill, while first checking whether the process itself has changed.
The second decision concerns individual check-ins: nine of twelve scheduled meetings took place. Three urgent requests displaced the others. The manager books replacement meetings and removes a recurring group meeting that no longer serves a clear purpose. At the next review, the team will examine both whether check-ins occurred and whether participants found them useful.
No financial improvement is inferred from this small example. The measures help select and verify actions. They do not, by themselves, establish the effect of those actions on the company’s results or explain every change in employee experience.
Run a monthly review that produces follow-through
Begin the review by checking data quality. Mark a missing source as “unavailable” rather than replacing it with zero. Compare periods calculated in the same way. Annotate an acquisition, a seasonal closure or a change in the definition so people do not mistake a reporting change for an operational trend.
Choose no more than two actions for the next cycle. Give each an owner, a date and a specific piece of expected evidence. Open the following review with those commitments. If a measure generates no decision for several cycles, reconsider whether it deserves a place on the dashboard.
M360 Compass connects roles, expectations and decisions through RoleAlign. Performance and development connects feedback with development priorities and individual follow-up. These confirmed capabilities can support the conversations described here. They do not mean that all eight proposed formulas are automatically calculated by the platform.
Bring one management decision and a description of your available sources to an M360 Compass demonstration. This helps you check how the platform fits your workflow before deciding which reports would be useful. Keep responsibility for definitions with someone who understands the underlying work.
For a related management tool, explore calculating employee turnover. Use it to clarify a separate question, while keeping this method focused on the situation described here.
Frequently asked questions about HR metrics
How many measures should we start with?
Start with two or three measures connected to real decisions. Expand only when definitions, sources and ownership are stable. A list of eight possible measures is a menu, not a requirement to collect everything immediately.
How should a small team interpret a percentage?
Display the number of cases alongside the percentage and review several comparable periods. A single departure can change a small team’s result substantially. Avoid assigning a cause or claiming a lasting trend without additional evidence.
Should every measure have a universal target?
No. Establish your starting point and the problem you want to address. Roles, seasonality and calculation methods affect comparisons. An external benchmark can prompt a question, but it cannot replace an understanding of your operating context.
What if the data is incomplete?
Identify what is missing, which period is affected and who will complete the source. Use reliable information for a limited decision. Keep a correction history so future comparisons do not silently mix numbers built using different methods.
