An HR dashboard in Power BI: what to show and what to leave out.
From the hand-built monthly report to the dashboard that refreshes itself. Data model, indicators and frequent mistakes.
The HR report is usually built by hand every month, takes two to three days of work, and reaches management after the period has already closed. That delay is not a detail: it turns a management instrument into a historical document.
Before opening Power BI you have to resolve the data model, and that is where the project is decided. It needs at least four tables: employees with their history of positions and salaries, movements with dated hires and exits, changes with absenteeism and vacations, and a calendar table. The last one is not optional: without a proper date dimension, any comparison between periods becomes fragile.
The most important design decision is how history is represented. An employee promoted in June must appear with the previous position in earlier months, not the current one. If the model only stores the present state, every analysis of turnover by position and salary evolution is distorted.
The indicators we include in the management dashboard are few and connected to the business: total personnel cost over revenue, voluntary turnover in critical roles, time to fill vacancies, salary positioning against the market, absenteeism, and succession-plan coverage. Six figures that fit on one screen.
What we leave out of the management dashboard, even if it exists in operational detail: number of résumés received, training sessions delivered, evaluations completed. They are useful to the department and change no decision at the top.
On refreshing, there are two paths. A direct connection to the source is ideal, but it requires a queryable database, and in many companies payroll data lives inside a closed system. The realistic alternative is an export file in a fixed format and location, generated automatically by the system and read by Power BI on a schedule. Less elegant, just as effective, and much faster to implement.
A precaution that is not technical: individual salary information must not be visible to everyone who opens the dashboard. Power BI allows rows to be restricted by the user viewing them, and that configuration must be defined in the design, not added after an incident.