How to Run a Leading-Indicator Quality Audit in 30 Days

A dashboard can show steady green cells while the operation still misses the signals that matter. The problem is rarely that leaders lack numbers. It is that the numbers they trust were never tested for field meaning, ownership and action. A leading indicator only helps when it proves a condition in the work, not when it only records a task.

The thesis of this guide is simple. A leading-indicator audit should measure indicator quality before the organization measures indicator volume. More observations, more audits or more closed actions can still hide drift if the metric does not point to a real control. Across more than 250 cultural transformation projects, Andreza Araujo has seen that pattern repeat. In Safety Culture: From Theory to Practice, repeated decisions define culture, which is why the audit itself becomes a culture test.

This article is for EHS managers, site leaders and dashboard owners who need a practical way to separate useful signals from KPI theater. James Reason's work on latent failures fits here because weak conditions usually sit behind calm charts, not behind noisy ones. A good audit is the fastest way to make those weak conditions visible.

What you need before starting

Before the first review, gather the current dashboard, the top indicators that reach leadership meetings, the source systems, the latest month of raw data and the names of the people who own each measure. If your site does not yet have stable definitions, start with How to Build a Safety Metric Dictionary in 30 Days because a quality audit can only judge what the organization has already defined.

Choose one site, one business unit or one contractor portfolio. Do not start with the whole company, because a 30-day audit needs a narrow scope if it is going to uncover real defects instead of collecting general opinions. If your team is still learning how to read evidence, Safety Culture Diagnosis: Learn how to do your own gives the diagnostic discipline this work needs.

The other thing you need is time with the field. A leading-indicator audit is not a spreadsheet exercise. It is a disciplined way to test whether the number still matches the work as it is actually done.

Step 1: Choose five indicators that already influence decisions

Start with the indicators that already shape monthly reviews, contractor ranking, bonus discussions or board conversations. If a metric does not influence a decision, it can wait. The audit should first examine the numbers that already carry authority, because those are the ones most likely to hide weak assumptions.

This is also where the article on Metric Review Cadence: Monthly Safety Steps becomes useful. A clean review rhythm matters only after the dashboard has a manageable list of indicators. If the site starts with too many measures, the team will spend the 30 days arguing about volume instead of quality.

The verification question is simple. Could a site leader name these five indicators without looking at the slide? If not, the number may be present, but it is not yet part of the operating language.

Step 2: Write the exact claim each indicator is supposed to prove

Every indicator should answer one question. Does it prove field coverage, control health, response speed or decision discipline? If one number tries to prove all four, it will prove none of them well. That is the own thesis behind this audit: quality matters before volume, because a number without a clear claim becomes a count with a safety label.

A training completion rate, for example, can prove attendance. It cannot prove that the task is now safer. A closed action can prove administrative closure. It cannot prove that the barrier works under production pressure. Andreza Araujo has seen this confusion repeatedly in large-scale transformation work, and it is one reason Safety Culture: From Theory to Practice treats repeated decisions as the real evidence of culture.

If the metric is tied to incentives, bonuses or public ranking, write that down too. The article on Safety KPIs vs Bonuses vs Control Checks shows how reward pressure can distort what a number is allowed to mean.

Step 3: Check the numerator, denominator and time window

Once the claim is clear, test the arithmetic. A leading indicator that uses the wrong denominator can look improved while exposure did not change at all. Hours worked, task counts, line hours, area counts and contractor shifts each tell a different story, so the audit must verify whether the denominator fits the risk the indicator claims to describe.

The easiest check is to recalculate the last three months using one definition and one time window. If the trend changes when the window changes, the metric may be sensitive to accounting choices rather than to field conditions. That sensitivity is not a minor technical issue. It is a sign that the dashboard is rewarding formatting more than control.

Use this step to compare the current metric with the broader measurement structure in How to Build a Safety Metric Dictionary in 30 Days. When the dictionary and the audit disagree, the audit should not blame the field first. It should question the measurement rule.

Step 4: Test the evidence trail on ten live records

Pick ten live records across the five indicators and follow each one from dashboard row to source document to field proof. A completed form is not proof by itself. A photo is not proof by itself. A signed closeout is not proof by itself. The record only matters when it still points to a real control in the worksite.

This step is where many teams find the gap between reporting and reality. If two reviewers reach different answers on the same record, the definition is still personal, not operational. That is a direct warning sign, because a leading indicator must be repeatable if it is going to guide leadership action.

The article on New EHS Data Analyst in 90 Days: What to Fix Before Metrics Mislead Leaders is useful here, because analysts often inherit exactly this problem. They are asked to trust numbers that nobody has traced back to the field.

Step 5: Define the action rule before you review the color

A red, amber or green status means nothing if nobody knows what should happen next. Write the action rule before the review starts. If the indicator turns red, who acts, by when and with what field check? If the team cannot answer that in one sentence, the metric is decorative rather than governing.

This is where dashboard theater often begins. A team celebrates a green trend and never asks what the color is supposed to change. That is why the review must connect the number to an owner, a response time and a verification method. Without those three elements, the color is only a slide effect.

When the action rule is weak, the safest move is often to remove the indicator from leadership review until the rule is repaired. The article on Zero-Accident Targets: 5 Distortions in Safety Metrics is a good warning here, because a metric can look precise while still pushing the wrong behavior.

Step 6: Compare the indicator with exposure and work changes

A useful leading indicator should move for a reason that the site can explain. If the work changed, the number may move. If the number moved and the work did not, the metric may be reflecting reporting noise, not control improvement. That is why this step checks the trend against production mix, overtime, contractor presence, seasonal heat, maintenance windows and changeover load.

Do not read the indicator in isolation. A site can produce more observations during a shutdown and less during normal production, even though the real exposure is higher in the second period. The number is not lying. It is simply incomplete when the work context is missing.

James Reason's latent failure logic is useful again because weak conditions often build slowly. A clean trend does not remove the conditions that create the next event. It only means the current report has not yet learned how to describe them.

Step 7: Turn the monthly review into a weekly decision rhythm

Monthly review is often too slow for a weak signal that should trigger same-week action. For the 30-day audit, create a weekly rhythm with the same five indicators, the same owner names and one short field question for each signal. That way the review stops being a long meeting with slides and becomes a decision loop with evidence.

The article on Metric Review Cadence: Monthly Safety Steps matters again here, because cadence is not a calendar habit. It is a control design. A rhythm that only presents the trend but never changes the work is just a recurring report.

At the end of each week, ask whether the signal changed the work, the supervision or the escalation path. If the answer is no for three consecutive reviews, the indicator is probably too weak, too vague or too detached from real control to stay in the executive pack.

Step 8: Decide which indicators stay, change or die

After 30 days, every indicator should fall into one of three buckets. Keep it if the field proof is strong, the owner is clear and the action rule works. Change it if the claim is useful but the definition or denominator is weak. Kill it if it only rewards motion or produces comfort without control.

This is the part many teams avoid, because removing a familiar metric feels risky. In practice, keeping a weak indicator is more dangerous. It trains leaders to trust the wrong signal and gives the site another routine to perform without changing the hazard.

Andreza Araujo's broader body of work makes the same point in a different form. In Safety Culture: From Theory to Practice, culture is what repeats. If the metric repeatably rewards appearance instead of evidence, the dashboard is teaching the wrong habit.

Final checklist

• The site has a short list of five indicators that already influence decisions.
• Each indicator has one claim, one owner, one denominator and one time window.
• Ten live records were traced from the dashboard to the field.
• Every color or threshold has an action rule with an owner and a deadline.
• The weekly review compares the signal with exposure and work changes.
• Each indicator is now classified as keep, change or kill.
• The dashboard is judged by control evidence, not by activity counts alone.

If the audit exposes weak indicators, do not rush to add more. Repair the ones that still matter and remove the ones that only create noise. A stronger dashboard is usually smaller, clearer and harder to game.

If you want the broader method behind this work, start with Safety Culture Diagnosis: Learn how to do your own and then return to Safety Culture: From Theory to Practice. The point is not prettier charts. The point is a measurement system that forces the operation to show real control before leaders call it progress.

Learn more about Andreza Araujo's work

FAQ

What is a leading-indicator quality audit?
It is a structured review that tests whether a leading indicator still proves a real field condition, has a clear owner and can trigger a specific action. The audit is about signal quality, not metric volume.

How is this different from a dashboard review?
A dashboard review often asks whether the numbers went up or down. A quality audit asks whether the numbers still mean what leaders think they mean and whether the field can prove the story behind them.

Who should own the audit?
EHS usually coordinates it, but operations must own the action rules and the field proof. If the audit lives only inside EHS, the number may be well reported and still poorly managed.

What if the site has no clean definitions yet?
Then the first step is to build the dictionary before you audit the quality. The article on How to Build a Safety Metric Dictionary in 30 Days gives the right foundation.

How often should the audit run?
Run the full audit in a 30-day cycle, then keep a weekly light review for the indicators that still matter. If the signal is weak or volatile, shorten the review rhythm until the work stabilizes.