The question “what is the turnover formula?” It seems to ask for a single account, but this is precisely the point that causes confusion. Today different definitions coexist. SHRM describes the turnover rate as separations in the period divided by the average number of employees, while recent Brazilian pages also call turnover a formula that takes the average between admissions and dismissals before dividing by the headcount. Both calculations can be mathematically correct and yet answer different management questions.
To put together a series that remains comparable a year from now, the best way is to choose the question first and only then the formula. If the intention is to measure loss of people, the numerator should be layoffs. If the intention is to measure the volume of movement between entrances and exits, admissions also come into play. If the intention is to understand the expansion of the picture, the admission rate and the net variation need to appear separately. Mixing these questions into a single percentage is what makes seemingly similar dashboards tell incompatible stories.
The question comes before the formula
A good definition of turnover needs to make four elements explicit: which event is included in the numerator, which population forms the denominator, what is the observation period and what is the organizational scope. Without this, “8% turnover” does not tell you whether 8% of people left, whether there was strong hiring accompanied by departures, or whether the rate was calculated on the initial headcount instead of the average headcount.
If the question is how many people left the organization
Use the churn rate as the primary measurement. The calculation is: dismissals in the period / average headcount for the period x 100. This format approximates the reading adopted by references such as SHRM and is the most direct when the question is retention, workforce stability or intensity of departures.
The numerator must contain the shutdowns that belong to the defined scope. The denominator must represent the average size of the population that was exposed to the risk of leaving during the
same range. This coherence between numerator, denominator, period and scope is more important than choosing a specific acronym for the indicator.
If the question is how much the frame moved between inputs and outputs
You can use a turnover index: [(admissions + dismissals) / 2] / average headcount x 100. This formula appears frequently in Brazilian HR content and summarizes the average volume of arrivals and departures in relation to the size of the staff. It can be useful for measuring the operational load of recruitment, integration and dismissal, but it should not be presented as if it were the same thing as the rate of people who left.
The decisive detail is the meaning. If a company is growing a lot, admissions increase this rate even when few people are leaving the organization. In a frame reduction, the average between inputs and outputs can make the movement appear smaller than the off rate. Therefore, the name of the KPI needs to reveal what it measures.
If the question is how much are we hiring
Use a separate admissions rate: period admissions / period average headcount x 100. The United States Bureau of Labor Statistics makes exactly this conceptual separation in JOLTS, publishing hiring and separation rates as separate measures. This logic is useful for the corporate dashboard because high hiring does not, in itself, mean high turnover; can mean expansion.
If the question is why are people leaving
Separate shutdowns by type before turning everything into a single rate. A minimum reading usually distinguishes voluntary and involuntary departures. JOLTS, for example, separates voluntary departure requests from employer-initiated layoffs and terminations, in addition to maintaining a category of other separations. For internal management, the taxonomy may be different, but it should remain stable over time.
A unique example shows why the formulas diverge
Imagine a company that starts the month with 100 people, hires 30, registers 10 dismissals and ends with 120. Suppose that, of the 10 dismissals, 6 are voluntary and 4 involuntary. Using the simple average between start and end, the average headcount is 110.
Based on the shutdown rate, the result is 10 / 110 x 100 = 9.1%. Using the movement formula that uses the average between admissions and dismissals, the result is [(30 + 10) / 2] / 110 x 100 = 18.2%. Based on the admissions rate, it is 30 / 110 x 100 = 27.3%. If someone divides just the dismissals by the initial headcount, it comes to 10.0%. None of these percentages arise from a calculator error; they differ because numerators and denominators have changed.
The interpretation also changes. The 9.1% responds to how much the average staff lost in layoffs that month. The 18.2% summarizes the average intensity of entries and exits. The 27.3% shows the strength of admissions. The net gain was 20 people. Just putting “turnover: 18.2%” on the dashboard erases the most important information: the business was expanding and most of the movement came from hiring.
Average headcount: the denominator that changes everything
The average between initial and final headcount is a simple and useful approximation when the situation varies relatively continuously. The formula is: (headcount at the beginning + headcount at the end) / 2. It is easy to reproduce and, if the rule remains the same, helps to build a consistent historical series.
The problem appears when the change is sudden. If a major hire happens near the last day of the month, the average between the two points may suggest that that larger picture existed for half the period, although this did not happen. In scenarios like this, a daily or weekly average headcount better represents the population actually present throughout the interval. The same goes for restructuring concentrated on one date.
The operational rule can be simple: use the best denominator that your data can stably reproduce. If the system stores daily headcount, prefer the daily average. If you only keep monthly closings, document the chosen approach. What cannot happen is changing the method in the middle of the series and comparing the results as if nothing had changed.
Expansion, reduction and layoffs need to appear in context
Turnover should not be read in isolation from headcount movement. A company may have many admissions and few departures because it is opening units. Another may have few admissions and many dismissals due to a planned reduction. Both scenarios can produce relevant movement indices, but require opposite interpretations.
In a layoff, the involuntary rate may increase due to a structural decision and not due to a retention problem. This does not mean removing these events from the total data. It means showing the total and, alongside it, the voluntary and involuntary cut-off. If the question is talent retention, the voluntary indicator gains diagnostic weight. If the question is workforce stability, total separations remain important.
Voluntary and involuntary turnover: use the same denominator
To compare turnover components, keep the denominator the same. The voluntary rate can be calculated as voluntary dismissals / average headcount x 100. The involuntary rate follows the same logic, only changing the numerator. Thus, the two installments can be read together with the total rate without creating artificial differences caused by different bases.
It is also worth maintaining a category of “others” when retirement, death, end of contract, transfer between establishments or specific situations do not neatly fit into voluntary or involuntary. Forcing each output into one of the two boxes may improve the appearance of the graph but worsen the quality of the data.
Internal transfers cannot change meaning depending on the screen
At the company level, a transfer between departments is not normally an exit from the organization and, therefore, should not be included in corporate turnover. At the level of an area, however, the same transfer can be relevant as an internal departure if the question is mobility or stability of that team. The point is to register the event as a transfer and decide whether to include it according to the scope, instead of treating it as a definitive shutdown in all analyses.
The same rule applies to promotions and job changes. They may affect indicators of internal mobility, succession or filling vacancies, but should not inflate the company's dismissal rate.
Monthly, year-to-date and 12-month are not the same series
A monthly rate responds to what happened that month. A year-to-date sums events since January and recalculates the rate on an average basis from the same interval. A 12-month rolling window always uses the most recent 12 months and is useful for reducing seasonality. These three numbers can coexist as long as they have different names.
Avoid adding monthly percentages to create an annual turnover. The denominators change from month to month, and the simple sum loses this information. For a 12-month termination rate, add the terminations for the interval and divide by the average headcount for the same interval. For a weighted monthly average, add the monthly separations and divide by the sum of the monthly headcounts, logic similar to that used by the BLS in its annual average separation rates. The label needs to make it clear which of these readings is being displayed.
It is also not safe to automatically multiply a single month's rate by 12 and call the result annual turnover. This extrapolation would only make sense under a strong assumption of stability, which is precisely what many businesses do not have.
Small areas require rate and side-by-side counting
In small departments, one or two departures can cause large percentage fluctuations. Therefore, the rate must be accompanied by the absolute count of dismissals and the average headcount. A window of three, six or 12 months can help you see trends without hiding what happened in each month.
When comparing areas, avoid a pure ranking of percentages. Teams with few members are more volatile, seasonal areas have different dynamics and growing units start from different bases. The indicator is most useful when it answers “where is the output concentrated?” than when you try to declare a winner and a loser without context.
How to create a mini data dictionary for the dashboard
The dictionary doesn't need to be sophisticated, but it does need to prevent the definition from changing silently. For each KPI, record the name of the indicator, the business question, the formula, the numerator, the denominator, the period, the scope, the rule for including link types and the effective date of the definition.
At the event level, record at least one stable link identifier, event date, type of event, grouped reason, organizational unit of origin and destination, type of link, indication of internal transfer and indication of inclusion in the headcount. If there is rehiring, define whether it creates a new analytical link or continues the previous one. If temporary workers and third parties are outside the headcount, their arrivals and departures should also not be included in the numerator of that KPI.
It is also important to separate headcount from FTE. Headcount counts people; FTE translates to full-time equivalent workload. Mixing people terminations in the numerator with FTE in the denominator changes the unit of the rate. If there is a need for an FTE-based indicator, it should be named and documented as another KPI.
A simple protocol for not changing the formula in the middle of the story
Before publishing any number, write down the question it answers. Then fix the numerator and denominator, define the calendar, choose the scope, and document the event rules. Then, calculate at least the total dismissal rate, the voluntary rate, the involuntary rate, the admission rate and the net change in headcount. If the organization also wants to maintain the movement rate based on the average between entries and exits, treat it as an additional metric, not as a silent substitute for the exit rate.
When a definition needs to change, preserve the previous version, record the date of the change and, if possible, recalculate the history with the new rule. Without this, an apparent improvement or worsening may just be a change in methodology.
So which formula to use?
To the most common question - “which portion of the frame is leaving?” - the clearest choice is terminations in period / period average headcount x 100. To understand hiring dynamics, calculate admissions separately. To measure the combined volume of inputs and outputs, use the movement formula with the average of the two events, but give it a name that does not confuse it with the churn rate.
The objective is not to choose a universal formula and discard the others. It is to prevent percentages with different meanings from receiving the same name. A reliable dashboard makes the question visible, keeps the denominator coherent, separates voluntary from involuntary, treats frame expansion and reduction as context, and maintains the same definition throughout the series. When this happens, turnover stops being a decorative number and becomes a management measure that can actually be compared.
