A dashboard full of charts can create the impression that we know an organization very well. It can also make weak data feel more convincing, make us give too much weight to a percentage because it is presented well, or turn an average into a conclusion. HR Analytics becomes useful when it helps us notice something we could not see before and, above all, when it makes us ask one more question.
Data comes before KPIs
If a dashboard says the average age is 39 but birth dates are filled in for only half the workforce, the problem is not the chart. It is the base we are using to read it. The same applies to salary, hire date, manager, department or contract type.
A good HR dashboard should therefore show data completeness as well as metrics. It is less spectacular than a chart, but often more useful. Before discussing a result, ask: how many people am I actually observing? Which records are missing? Is the missing data random, or concentrated in a particular group?
That is why PDF HR places data-completeness indicators next to the metrics generated from the Employee Directory. They do not grade the database; they remind us that mathematical precision is not the same as information quality.
- Check the population before the result.
- Know which fields actually feed the metric.
- If a lot of data is missing, say so before interpreting the chart.
- Do not confuse mathematical precision with information quality.
Headcount: every snapshot needs a date
Saying “we have 120 people” sounds simple, but even the most basic figure changes meaning depending on the moment being observed. Do we mean today, year-end, or everyone who passed through the organization during a period?
When reading headcount, separate the current snapshot from historical movement. An employee with a recorded termination date is not equivalent to someone currently active; filtering a year means asking who belonged to the organization during that period, not simply who appears in today’s directory.
The useful question is therefore not only “how many people do we have?”, but “how many people did we have, when, and within which population?”. That small clarification prevents many poor interpretations.
Exits and turnover: call the metric what it is
Turnover is one of those words that can sound more precise than it is. Different companies use different formulas: exits over average headcount, exits over opening headcount, or separate voluntary and involuntary rates. Change the denominator and the number changes too.
That is why PDF HR describes the general snapshot as the “terminated share”: terminated records as a share of the population being viewed. The annual series instead shows a simple ratio between exits recorded during the year and the workforce base reconstructed from the Directory for that same year. It is readable, but it is not a universal definition of turnover.
Comparisons make sense only when the formula stays consistent. Before saying that turnover is “high”, ask who is leaving, from which areas, with what tenure, in what period and compared with which population. The number starts the conversation; it rarely finishes it.
- Always define numerator and denominator.
- Use the same formula when comparing periods.
- Before comparing organizations, check that they measure the same thing.
- Where possible, separate causes instead of stopping at the total.
Hires and net growth show movement, not quality
Ten hires and two exits produce net growth of eight people. That is correct, but it does not tell us whether the organization is growing well. The company may be staffing ahead of a project, replacing missing capabilities or simply adding people without changing team capacity.
Net growth is useful because it makes direction visible. It becomes more interesting when read by department, location, role or level. If all growth sits in one function, that may open a strategic question; if hires and exits are both high but cancel each other out, the net figure may hide a great deal of movement.
A balance close to zero can therefore describe stability or a revolving door. The difference is not in the KPI. It is in the story behind it.
Span of control: an average is not a verdict
Knowing the average number of direct reports per manager can be useful. It can surface fragmented structures, managers with very few reports or areas where one person coordinates a very large group. The problem starts when that average becomes a universal rule.
A team of highly autonomous professionals does not need the same level of coordination as a newly formed group. Standardized work differs from project work; geography, seniority and the manager’s role all change the context.
In PDF HR, span of control is derived from the direct reporting relationships stored in the Employee Directory and from matching manager names to employees. An incomplete hierarchy therefore produces an incomplete metric. Before judging the value, check that the structure in the data is the structure that actually exists.
Average salary can hide what you are looking for
Average salary is easy to read and therefore easy to over-interpret. If a department contains junior and senior roles, managers and specialists, the average may describe the calculation perfectly and the real people poorly.
Before comparing departments, understand the composition: how many levels exist, how many salary fields are complete, whether roles are genuinely comparable and whether a small number of very high or low values is moving the result. An average is a starting point, not a compensation benchmark.
The same principle applies to any pay comparison: internal data can help identify areas worth investigating, but it does not replace a proper compensation analysis, contractual context or a benchmark built on genuinely comparable roles.
Age, gender and composition: describe before judging
Demographic metrics can help describe a workforce. They can surface concentrations by age range, legal sex distribution or differences across functions. They do not explain on their own why that composition exists, and they do not justify conclusions about individuals.
It is worth being precise about the data: PDF HR analytics report legal sex, taken from the Directory field or derived from the Italian tax code. It is not the gender identity of the people involved and should not be read as such.
A percentage becomes useful when it is connected to a legitimate organizational question and handled with appropriate care. It becomes much less useful when it turns into a label. In HR, statistical description must always remain separate from individual assessment, with personal data handled according to the rules that apply in the relevant context.
The principle is the same: observe a pattern, check the data quality, look for context and only then decide whether there is something worth investigating.
The best dashboard takes you out of the dashboard
If all we know after looking at HR Analytics is that one number is red and another is green, we have probably learned very little. A useful dashboard should leave us with a short list of better questions.
For example: why are exits in this department concentrated in the last twelve months? Why do some employees have no manager assigned? Is the growth of this location consistent with the work it will need to support? Is average salary different because roles differ, or is there genuinely something to investigate?
This is where data meets HR work. Not to replace conversations with managers and people, but to make those conversations less random. Numbers can help us decide where to look; meaning appears when we bring them back into the reality of the organization.
- What question does this figure make me ask?
- What information is missing before I interpret it?
- Can I segment it to understand where the pattern comes from?
- Am I comparing genuinely similar periods and populations?
- What concrete check would I make tomorrow with managers or detailed data?
Checklist finale
- The population being observed is clear
- I know which fields feed each important metric
- I have checked the completeness of relevant data
- Numerator, denominator and time period are explicit
- I am not comparing different formulas under the same name
- I have segmented the data when an average hides different groups
- I have separated a pattern or correlation from a causal explanation
- The data leads to a concrete check, not an automatic judgment
HR metrics are descriptive and decision-support tools. They do not replace professional judgment, specialist compensation analysis, or the legal and privacy checks required by the relevant context.