How to Validate a Safety Dashboard for the Monthly Review in 8 Steps

A safety dashboard is not a report to admire. It is a decision tool, and if it cannot change the monthly review, it is decoration. Across 25+ years in multinational EHS, Andreza Araujo has seen the same failure repeat: the chart looks clean while the work still depends on one person remembering what the chart never checked. In Muito Além do Zero, lagging indicators only look backward.

If your team already knows the difference between TRIR, LTIFR, DART, and SIF rate, this guide shows how to test whether the dashboard is ready for the review that follows. For the metric map itself, the board decision article keeps the vocabulary aligned before the first slide opens.

What you need before starting

Before the review, collect four things. You need one decision question, one owner for each metric, one recent deviation to test against, and one source list that explains the definitions and denominators. Without those four items, the dashboard becomes a pile of numbers that can impress a room without helping it decide.

• One decision the review must support.
• One owner for each metric.
• One recent deviation, near miss, or repeat exception.
• One source list for definitions, denominators, and time windows.

In Diagnóstico de Cultura de Segurança, Andreza Araujo treats measurement as a live test of the system, not an archive of past activity. That is the right mindset here, because the dashboard should prove that the organization still understands the work it claims to manage.

Step 1: Decide the decision the dashboard must support

Start with the decision, not with the data file. If the meeting cannot answer whether to stop work, keep investing, change an owner, or close a control gap, the dashboard is too wide. The verification test is simple: read every metric title and ask what decision it supports. A common error is to build a dashboard around what data exists instead of what leadership needs to decide.

James Reason's latent failure logic fits this first step, because a poor dashboard is often the visible layer of a deeper planning error. The review looks weak when the organization collects numbers first and asks questions later.

Step 2: Separate lagging, leading, and precursor signals

Separate the metric families before anything reaches the slide deck. Lagging numbers show what already happened. Leading signals show whether the system is doing the preparatory work. Precursor signals show whether a known failure is getting closer. In Muito Além do Zero, Andreza Araujo treats this distinction as basic discipline, because one green chart can hide three different kinds of weakness.

• Lagging signals show consequence.
• Leading signals show current control effort.
• Precursor signals show rising exposure.

If the team wants the broader vocabulary before the review, this indicator guide keeps the families separate. A dashboard becomes more honest when each family is visible on its own line.

Step 3: Verify definitions and denominators

A number without a definition only creates arguments after the review, and a denominator without a common rule makes sites look better or worse for the wrong reason. Check who defined the metric, which population it covers, and what date window it uses. If one site counts contractor hours one way and another site counts them another way, the dashboard is not comparing like with like.

In A Ilusão da Conformidade, Andreza Araujo warns that a clean surface can hide a broken process. The same pattern appears here when the slide is neat but the source logic is messy. If the number cannot be explained in one sentence, it is not ready for leadership review.

Step 4: Trace each metric to one field action

Every metric needs a field action attached to it. If the number rises or falls, who changes what on Monday? A dashboard becomes real when the metric has a named owner, a due date, and a control that changes because the number changed. A common error is to let the review end with a calm promise to monitor the situation while nobody owns the next move.

That is where James Reason still helps, because the visible number often fails after a hidden process has already drifted. The review should expose that drift before it becomes routine.

Step 5: Compare the dashboard with one recent deviation

Pick one near miss, one repeat exception, or one temporary deviation and ask which line on the dashboard would have warned you earlier. If nothing on the chart would have moved before the event, the chart is telling a story after the fact. A dashboard that only reacts after the event is useful for reporting, but not for control.

Across 250+ projects, Andreza Araujo has seen leaders trust numbers faster than field evidence, which is why the dashboard must be tested against a real case, not against theory alone. If the current review cycle keeps missing the same kind of deviation, the dashboard is not seeing the work.

Step 6: Remove vanity metrics and duplicate counts

If two numbers describe the same thing, keep the one that changes action fastest. If a number exists only because it is easy to collect, drop it. The goal is not a full spreadsheet. The goal is a dashboard that reveals what leaders need to fix. A common error is to stack percentages from different sources and pretend they can be read as one line.

If a green number cannot explain itself, it should not stay on the slide. The dashboard is supposed to reduce uncertainty, not decorate it.

Step 7: Match the dashboard depth to the audience

The board needs a small number of strategic signals, the plant leader needs control loss patterns, and the supervisor needs an operational view that can be acted on in the shift. If everyone sees everything, nobody sees what matters. If the same slide is used for every audience, the review becomes ritual instead of control.

For the audience split, the article on safety dashboard failures is a useful companion, because it shows how a single view can hide different decision needs. The right level of detail is the one that makes the next action obvious.

Step 8: Close the loop after the review

A dashboard that updates but never changes the owner, the deadline, or the follow-up conversation is just a prettier archive. The verification test is simple: after the meeting, check whether one metric was retired, one owner was named, or one control changed. In Diagnóstico de Cultura de Segurança, measurement is never an end in itself, because evidence only matters when it changes behavior.

If the review ends and nothing is different in the field, the dashboard has already failed. The last step is not a new chart. It is proof that the chart changed the work.

What to verify before the dashboard goes live

• The review has one decision question.
• Every metric has a definition, a denominator, and an owner.
• Lagging, leading, and precursor signals are separated.
• One recent deviation was used as a test case.
• The meeting ends with action ownership, not notes.
• The slide works for the audience in the room.

When those checks are in place, the dashboard becomes part of control, not a monthly ritual. That is the standard Andreza Araujo keeps pushing across her books and advisory work, because leaders do not need more numbers. They need fewer numbers that say more.

For teams ready to tighten the review, start with the book store and then move to Andreza Araujo for a culture diagnostic, because the fastest way to improve metrics is to stop rewarding numbers that cannot explain themselves.