EHS Data Analyst in 60 Days: What to Do When the Dashboard Starts Hiding Risk

An EHS data analyst is often hired to clean reports, but the real job is to decide which numbers deserve to interrupt work, which numbers belong in a trend view, and which numbers should trigger a supervisor conversation before a weak pattern becomes normal. Across 25+ years leading EHS at multinationals, Andreza Araújo has observed that dashboards hide risk when leaders reward neatness before they reward ownership. As Andreza Araújo argues in The Illusion of Compliance, a clean report can coexist with a weak operating system. In Muito Além do Zero, the English gloss of Far Beyond Zero, the point is even sharper. Zero on a slide is not the same thing as control in the field.

An EHS data analyst is the person who turns raw safety data into decision-ready signals for operations, HR, and leadership. The first 60 days matter because the analyst will either build trust in the numbers or quietly help the organization confuse activity with risk control.

Key Takeaways

• Build source ownership before building a dashboard, because a beautiful chart fed by vague data only hides more risk.
• Treat metric freshness as a control issue, not a reporting habit, since stale inputs create late decisions.
• Separate administrative completeness from field proof, because a closed record does not mean the exposure changed.
• Use response rules before leaders ask for one number, so every metric has an owner, a threshold, and a next action.
• Keep TRIR, LTIFR, DART, recurrence, and verification in their own lanes, because each answers a different question.

What an EHS data analyst needs to understand before starting

The first thing to understand is that safety data is not neutral. Every field in a system reflects a choice about who reports, who classifies, who approves, and how long the organization is willing to wait before acting. That is why OSHA recordkeeping definitions, BLS fatal injury tables, and ISO 45001 matter even when the analyst never quotes them in a board pack. They set the frame for what counts, what does not count, and what deserves attention before the next shift starts.

James Reason's latent-failure lens is useful here, because data quality problems often look technical while their causes sit in ownership, incentives, and work design. Across more than 250 cultural transformation projects supported by Andreza Araújo, the same pattern repeats: if the organization wants the dashboard to tell the truth, it has to decide that truth is worth the inconvenience.

The analyst is not the owner of risk, but the analyst is often the first person who sees that the same incident system, action tracker, and attendance report are being blended as if they were all the same thing. That blending creates a dashboard that looks complete while still hiding the question leaders actually need to answer.

Days 1 to 7: map sources, owners, and denominators

The first week should be spent on inventory, not on color. List every source that feeds the dashboard, name the owner of each source, and write down the denominator that makes the number meaningful. If a metric has no owner, no refresh cadence, or no denominator, it is not ready for leadership use, because nobody can tell when it went stale or whether it is measuring the same thing twice.

This is the stage where the analyst should separate systems by function. An incident log is not an action tracker. A complaint channel is not an absence report. A verification checklist is not a training record. When those sources are merged too early, the organization gets a single view that is easy to read and hard to trust.

If this table feels basic, that is the point. Basic structure is what keeps a dashboard from becoming an expensive collage of unrelated numbers.

Days 8 to 15: separate signal from administrative noise

By the second week, the analyst should test whether the numbers are telling the same story at different speeds or several different stories with one label attached. Metric freshness is the first test. A recent entry can still describe an old condition, and a late correction can make the current month look worse than the one that actually created the risk.

That problem becomes visible when the dashboard mixes TRIR, LTIFR, DART, action aging, and verification pass rate as if they were equivalent. They are not. TRIR tells the analyst what entered the record. LTIFR tells how often lost time appeared. DART tells how often cases crossed a specific threshold. Action aging tells whether decisions are stalling. Verification pass rate tells whether the control worked after implementation.

As Andreza Araújo argues in Safety Culture: From Theory to Practice, repeated decisions are what make culture visible. For an EHS analyst, that means the dashboard should expose repeated delays, repeated exceptions, and repeated control failures, not just monthly totals that make the slide look orderly.

Days 16 to 30: define response rules before the board asks for one number

The third week is the point where the analyst should stop asking, "What should we measure?" and start asking, "What should happen when the number moves?" A metric without a response rule is decoration. The value comes from the decision it triggers, because leaders do not need more color. They need a clear next move.

That is why the analyst should define thresholds with operations, HR, and EHS before the monthly review. If critical action aging crosses a line, who owns escalation? If the same recurrence appears twice, who decides whether the issue needs redesign instead of retraining? If verification fails, who confirms whether the control failed because of design, maintenance, supervision, or behavior? Those questions should be answered before the board asks for a single clean headline.

This is also where ISO 45001 becomes practical. The standard asks the organization to plan, operate, evaluate, and improve. A response rule turns that language into an operating rhythm, which is what keeps the dashboard from becoming a passive reporting exercise.

Days 31 to 45: test freshness, aging, and ownership

Once the basic logic is set, the analyst should look for time lag. Fresh data is useful only when it reaches the people who can still change the work. If a near miss arrives after the crew has already moved on, the number may be accurate and still useless. James Reason's point about latent failures fits well here, because the delay is often not a data problem alone. It is a sign that ownership moved too slowly or that the route to escalation is too weak.

The best test is simple. Pick one issue that mattered last month and trace it from first signal to final decision. Who saw it first? How long did it take to classify it? Who got it next? When did it reach the person who could change the plan? If the path is unclear, the analyst is not only measuring latency. The analyst is measuring how long the organization can ignore the warning before someone calls it a warning.

Across more than 250 cultural transformation projects, Andreza Araújo has seen that leaders trust data more when the analyst can explain the path from field signal to board action in plain language. That explanation is often more valuable than another chart, because it shows where truth gets delayed or softened.

Days 46 to 60: build a one-page board pack that changes decisions

By the final two weeks, the analyst should build a board pack that fits on one page and forces decisions. One page is enough when it contains three things: the exposure, the trend, and the decision owner. The board should be able to see, at a glance, which risk is rising, which control is weakening, and what the organization will do before the next cycle.

A good board pack does not start with TRIR because TRIR is a lagging signal and often too broad to guide action on its own. If serious exposure is rising while the recordable rate stays flat, the board needs to see the gap, not the comfort. That is why BLS fatal injury tables, control verification results, recurrence patterns, and aging actions deserve a separate lane from routine counts. They keep fatal exposure from getting lost inside an average.

A board pack should also show what changed since the last review. If nothing changed, the analyst should say so plainly. If the same issue has been explained three times and not fixed, the pack should show that the problem is not data quality alone. It is decision quality.

Common traps

The first trap is chasing completeness before usefulness. Teams often spend weeks cleaning every field and still fail to answer the leadership question. A 90 percent complete dashboard that triggers action is better than a perfect dashboard nobody reads.

The second trap is averaging away the hotspots. A site average can hide a department, shift, or contractor group that is carrying most of the exposure. Andreza Araújo has seen this repeatedly in multinational work, because averages make the system look calmer than the floor.

The third trap is treating zero as proof. Muito Além do Zero is useful precisely because it warns against that reflex. Zero can mean strong controls, but it can also mean silence, delay, or a classification habit that has become too convenient.

The fourth trap is assuming the analyst can fix ownership gaps alone. Data work can reveal weak ownership, but it cannot replace it. If the line owner, HR lead, or operations manager does not act on the signal, the analyst has only built a more elegant way to document inaction.

Resources to deepen

• Safety Culture Diagnosis: Learn how to do your own, for a practical way to test whether the organization is seeing real patterns or only polished reports.
• Safety Culture: From Theory to Practice, for the link between repeated decisions and visible culture.
• The Illusion of Compliance, for the gap between formal controls and lived exposure.
• Muito Além do Zero, the English gloss of Far Beyond Zero, for understanding why zero is not enough as a headline metric.
• Headline Podcast, for leadership conversations that connect metrics to real decisions.

For practitioners ready to apply this end to end, Safety Culture Diagnosis offers the step-by-step playbook.

FAQ

What should an EHS data analyst fix first?

Fix source ownership first. If nobody owns the incident log, action tracker, verification record, and refresh cadence, the dashboard will keep mixing different realities into one misleading summary.

Should the board see TRIR first?

Not by itself. TRIR can sit on the board pack, but it should share space with control verification, recurrence, action aging, and serious exposure signals, because those measures change decisions sooner.

How do I know a metric is stale?

If the number is recent but the work has already moved on, it is stale for decision purposes. Freshness is not only about date. It is about whether the signal still matches the condition leaders can change.

What if leaders want one number only?

Give them one headline, but make it a decision headline, not a vanity headline. The best single number is the one that triggers a named action, not the one that merely looks clean.

Which book should a new analyst read first?

Read Safety Culture Diagnosis first if you need a structured way to audit the system. Read The Illusion of Compliance next if the dashboard looks better than the work. Read Muito Além do Zero when the organization keeps treating zero as the endpoint.

Recommendation

Start with one dashboard, one business unit, and one response rule set. If the board cannot explain what action each number triggers, the dashboard is not yet a decision tool. If the data cannot name an owner, a cadence, and a threshold, the data is not yet mature enough to carry executive weight.

An EHS data analyst earns influence when the numbers become hard to ignore and easy to use. That is the real job. The slide deck is only the proof that the job is being done well.

For the next step, keep Headline Podcast in the loop and use the next review to test whether the board is reading a dashboard or making a decision.