How to Build a Safety Metrics Dictionary in 30 Days

A safety metrics dictionary is the document that defines exactly what each safety indicator means, how it is calculated, who owns it, what evidence supports it, and which decision it should trigger. Without that dictionary, two sites can report the same metric name while measuring different realities.

The danger is not cosmetic. When a board reviews TRIR, corrective action closure, near-miss quality, SIF exposure hours, stop-work activity, or critical control verification without common definitions, the discussion becomes a contest of labels. One plant may count contractor hours differently. Another may close actions when documents are uploaded. A third may treat every observation as a leading indicator, even when it reveals nothing about fatal risk.

Headline Podcast often returns to a simple leadership test: can the organization tell the truth early enough to act? A metrics dictionary supports that test because it turns safety data from a reporting habit into a decision system. Across 25+ years of executive EHS work and more than 250 cultural transformation projects connected with Andreza Araujo's practice, one pattern appears often. Leaders do not only need more data. They need fewer ambiguous definitions.

Key Takeaways

• A safety metrics dictionary prevents different sites from using the same indicator name for different calculations.
• The strongest dictionary connects each metric to a decision, not only to a dashboard field.
• Definitions should include formula, denominator, owner, evidence rule, review frequency and misuse warning.
• Lagging indicators still matter, although they should not be allowed to hide fatal-risk exposure.
• The dictionary closes only after operations, EHS, HR, contractors and executives test it against real records.

What you need before starting

Before starting, collect the last three months of dashboard exports, site scorecards, board reports, incident classifications, corrective action registers, contractor hour records, near-miss logs, critical control verification records and any written KPI definitions already in use. The point is not to admire the dashboard. The point is to discover where the same word means different things.

You also need a small working group with authority. Include EHS, operations, HR if people data appears in the dashboard, one finance or business analyst who understands denominators, one frontline supervisor, and a senior sponsor who can settle disputed definitions. If the group has no executive sponsor, the dictionary may become another technical file that nobody follows when reporting pressure rises.

Set one rule on day one. A metric enters the dictionary only if the team can name the decision it supports. If nobody can answer what a leader should do when the number moves, the metric may be interesting, but it is not yet a management indicator.

Step 1: List every safety metric currently reported

Start by listing every safety metric that appears in formal or informal reporting. Include corporate dashboards, site boards, contractor reviews, monthly business reviews, audit reports, ESG packs, injury logs, culture scorecards and supervisor huddles.

Do not merge similar names too early. TRIR, total recordable rate, recordable frequency, OSHA recordable rate and all injury frequency rate may look close, although each may carry a different denominator, legal frame or audience expectation. Keep the raw language first so the team can see how much variation already exists.

The first output is a plain table with metric name, source report, site or function, current formula if known, reporting frequency and apparent owner. The common error is asking people what the official definition is. Ask instead what file they actually use at month-end, because the lived definition often sits inside the spreadsheet.

Step 2: Separate legal records from management indicators

Separate metrics that serve legal or regulatory recordkeeping from metrics that guide management decisions. OSHA recordkeeping terms, RIDDOR reporting, MSHA Part 50 reporting or local statutory categories must follow their own rules. A company can add management indicators, but it should not quietly rewrite legal categories to make performance look cleaner.

This distinction protects the organization from two opposite errors. One error is treating every legal record as a leadership dashboard. The other is treating a management indicator as if it can override a statutory record. Both errors damage trust because people begin to suspect that the number exists to defend the company rather than describe risk.

Use the Headline guide on safety metric denominators when the team debates hours worked, headcount, exposure hours or task volume. Denominator discipline is where many good dashboards become misleading.

Step 3: Define the decision each metric should trigger

For each metric, write one sentence beginning with the decision owner. For example, the operations director reviews SIF exposure hours to decide which fatal-risk controls need field verification next month. The plant manager reviews corrective action aging to decide where leadership must remove a closure obstacle. The EHS manager reviews near-miss quality to decide whether reporting is producing usable learning or only volume.

This step forces a different conversation. A metric without a decision is often kept because it is familiar, easy to pull, or politically safe. A metric with a decision has a job. It tells someone where to look, what to challenge, what to fund, what to stop, or what to verify.

As Andreza Araujo argues in A Ilusao da Conformidade, often explained in English as The Illusion of Compliance, formal evidence can look reassuring while operational truth remains weak. A decision sentence reduces that gap because the indicator has to prove how it changes leadership behavior.

Step 4: Write the calculation and denominator rules

Write the formula in a way a new analyst could reproduce without asking the previous owner. Include numerator, denominator, multiplier if any, inclusion rules, exclusion rules, treatment of contractors, treatment of temporary workers, reporting period, late-entry rule and source system.

Denominators deserve special attention because they decide what the metric is really comparing. A rate based on hours worked answers a different question from a rate based on tasks, shifts, exposures, assets, work orders or critical lifts. If the dictionary does not say which denominator is used, leaders may compare sites that do not share the same exposure base.

The misuse warning belongs beside the formula. TRIR should not be used as proof that fatal risk is controlled. Action closure percentage should not be used as proof that controls changed in the field. Observation volume should not be used as proof that workers are engaged. These warnings make the dictionary harder to misuse under pressure.

Step 5: Add evidence rules for each metric

A metric definition is incomplete until it states what evidence qualifies. Corrective action closure may require field verification, photo evidence, engineering sign-off, worker confirmation or operating procedure revision depending on the action type. Without this rule, one site closes risk while another closes paperwork.

Near-miss quality needs the same discipline. A useful near miss usually contains the task, exposure, failed or weak control, potential severity, immediate protection and follow-up decision. A vague entry that says unsafe condition corrected may add volume to the dashboard without giving leaders a usable signal.

The Headline article on corrective action closure as proof is a useful companion here because it explains why closure should be tested in the work, not only in the action tracker.

Step 6: Assign one owner and one challenger

Every metric needs one owner who maintains the definition and one challenger who tests whether the metric still tells the truth. The owner may sit in EHS, operations, HR, maintenance, security or contractor management depending on the indicator. The challenger should come from a different function so the definition is not protected by the same group that reports it.

This pair prevents quiet drift. A metric drifts when a site changes data entry rules, a contractor changes hour capture, an action tracker adds a closure status, or a corporate report changes its period cut-off. Nobody may intend to distort the number, although the number stops meaning what leaders think it means.

Make definition changes visible. The dictionary should include version date, reason for change, approving role and the first reporting period affected. When leaders compare performance across time, they need to know whether the risk changed or the definition changed.

Step 7: Test the dictionary against real records

Test the draft dictionary with real records before publishing it. Select 10 records for each high-value metric and ask two people to classify or calculate them independently using only the written definition. If they reach different answers, the definition is not ready.

This is where hidden ambiguity becomes visible. The word contractor may exclude visitors at one site and include embedded maintenance at another. A corrective action may be closed when the owner signs off in one function and when the control works in another. A SIF exposure may be counted by task in one plant and by hour in another.

Use the Headline comparison of control health, TRIR and SIF exposure board metrics when executives need a clearer distinction between injury outcomes and fatal-risk control signals.

Step 8: Publish the dictionary with review governance

Publish the dictionary as a controlled management document, not as a hidden spreadsheet tab. Put it where dashboard owners, site leaders, supervisors and analysts can find it. Include a short change log and a named route for questions or disputed classifications.

Set a quarterly review for the first year. New work, new contractors, acquisitions, technology changes, regulatory changes and dashboard redesign can all make definitions stale. The review should ask which metrics were misunderstood, which decisions improved, which definitions were challenged, and which indicators should be retired.

Close the first 30 days by asking each senior leader to choose one decision they will now make differently because the metric is clearer. If nobody changes a decision, the dictionary may be technically accurate while still failing its management purpose.

30-day implementation timetable

Conclusion

A safety metrics dictionary will not make a weak safety system strong by itself. It will, however, make weak definitions harder to hide. When every metric has a decision, owner, formula, denominator, evidence rule and misuse warning, leaders can challenge the dashboard with more precision.

The practical test is simple. If two sites can use the dictionary and reach the same answer from the same record, the organization has moved from metric vocabulary toward metric discipline. That discipline matters because unclear numbers do not only confuse reports. They delay the decisions that protect people.