Safety Dashboard: What to Include for Sites, Executives and the Board

Decide who reads the dashboard and what they will decide with it before choosing a single chart. Then give every indicator a written definition, a source system, a refresh cadence and an owner. A dashboard without those four things is a picture of numbers nobody can defend.

Most safety dashboards start as a request from leadership for "the numbers", and the quickest numbers to produce are injury counts. That is how a dashboard ends up made almost entirely of lagging indicators, which tell a board what already went wrong and very little about whether prevention is working. This guide covers what to include for each audience, how to calculate the two rates most dashboards carry, which metrics reward the wrong behaviour, and how to describe a layout before anyone builds it.

Start with the decision each viewer makes

One dashboard for every audience fails each of them. A site manager needs to know which inspection is overdue this week. A board needs to know whether the organisation's most serious risks are under control and whether that is improving. The underlying data is largely shared by all three views, and the questions each one answers differ.

Site or operations view. Answers what needs doing now: overdue inspections, open corrective actions past their due date, hazards reported this week with no interim control, and incidents awaiting investigation. Reviewed weekly or daily. Detail goes down to the individual record.

Executive view. Answers where attention and budget should go: trends across sites, the rate of action closure, recurring hazard categories, and the injury rates over a rolling twelve months. Reviewed monthly. Each figure should allow a drill-down to the site responsible.

Board view. Answers whether risk is under control and whether the safety programme is working: a small number of indicators, the status of the highest-severity risks, serious incidents with their investigation status, and a short written commentary. Reviewed quarterly, with fewer numbers and more written explanation than the other views.

A practical test for each tile: if it changed by twenty percent next month, who would do what? If nobody can answer, the tile belongs on a different view or nowhere.

Put leading and lagging indicators on the same page

OSHA describes leading indicators as proactive and preventive measures that can show how effective safety and health activities are, and lagging indicators as measures of events that occurred in the past. Its guidance pairs them: leading indicators drive change, lagging indicators measure effectiveness. A dashboard that shows only one type gives half the picture.

The practical rule is pairing. Every activity count sits next to a measure of quality or outcome that would move if the activity were being done badly. Inspections completed on schedule sits next to findings per inspection. Actions closed by due date sits next to actions reopened or re-dated. Our guide to leading vs lagging safety indicators covers the definitions, the families of leading indicators and why a proactive measure is not automatically predictive.

The indicator table

The table below is a starting set. It is a recommended operating structure; choose the rows that match hazards in your own risk register, and drop the rest. The refresh column shows the minimum cadence for the view where the indicator is most used.

For each row, write a one-paragraph definition card before the indicator appears on any screen: the exact definition, the formula, the source system, what counts and what is excluded, the owner, the refresh cadence, and what happens when the number crosses a threshold. The card is what lets a new EHS coordinator reproduce the number six months later.

Calculate TRIR and DART from the source

Both rates use the same formula, published by the Bureau of Labor Statistics: the number of injuries and illnesses multiplied by 200,000, divided by the employee hours worked. BLS explains that 200,000 hours represents 100 employees working 40 hours per week, 50 weeks per year, which is what makes rates comparable between organisations of different sizes.

The difference is the count. For the total recordable case rate, BLS counts column H (cases with days away from work), column I (cases with job transfer or restriction) and column J (other recordable cases) of the OSHA 300A summary. For the DART rate, it counts columns H and I only. Hours worked should exclude non-work time, even if paid, such as vacation, sick leave and holidays.

BLS gives a worked example. A company with 7 recordable cases and 400,000 hours worked has a total recordable case rate of 3.5. Three of those cases involved days away, transfer or restriction, so its DART rate is 1.5.

Two dashboard practices follow from this. First, calculate the rates from the log and the payroll hours directly, and show the period they cover. A rate copied from last quarter's slide by hand is where errors start. Second, at a single site with a few hundred workers, one case moves the rate noticeably, so a monthly rate on its own invites overreaction. Show a rolling twelve-month rate with the case count next to it. If the 300 log itself is the bottleneck, our guide to automating OSHA 300 log reporting covers how to keep it current.

Avoid indicators that reward underreporting

The fastest way to improve an injury rate on a dashboard is to make reporting harder. OSHA recordkeeping rules address this directly. Employers must have a reasonable procedure for employees to report work-related injuries and illnesses promptly and accurately, and a procedure is not reasonable if it would deter or discourage a reasonable employee from accurately reporting. Employers must not discharge or discriminate against an employee for reporting (1904.35(b)(1)).

OSHA's 2018 interpretation memorandum clarifies that safety incentive programs, including rate-based ones, are permissible as long as they are not implemented in a manner that discourages reporting. For a dashboard, that has a practical meaning. Any tile that ties an injury count to a reward, a ranking or a manager's objectives needs a counterweight on the same screen.

• Show reporting activity next to injury rates. A falling injury rate with a falling near-miss and hazard-report rate deserves a question before it deserves praise.
• Avoid league tables of sites ranked by injury rate alone. Rank on the leading indicators the site controls, such as action closure and inspection completion.
• Avoid "days without an injury" as a headline. It resets on a report, which makes the report the thing people avoid.
• Track the lag between an event and its report. A growing gap is an early sign that reporting is being discouraged somewhere.

Leading indicators can be gamed as well. Inspection completion rises when inspections get shorter, and action closure rises when actions get re-dated. Pairing each count with a quality measure is the defence on both sides.

Owners, sources and refresh cadence

Every indicator needs one named owner. The owner is accountable for the definition, for checking the number before it is published, and for explaining it when it moves. The owner is usually the person closest to the data source: the site EHS lead for inspection completion, the HR or payroll contact for hours worked, the EHS manager for the recordable rates.

Refresh cadence should follow the decision, and the data should be at least as fresh as the meeting that uses it. A weekly site huddle working from a monthly extract is discussing last month. Write the "data as of" date on every view so readers know what they are looking at.

Rebuilding a dashboard from source records on a schedule removes manual copying. A scheduled rebuild still reflects only what the source holds at run time. In the public SoterAI product webinar of 13 October 2026 on scheduled custom workflows, the presenter built a policy gap-analysis workflow that produced a dashboard. Discussing a monthly schedule, he advised uploading new versions of source documents before each manual or scheduled run (webinar, 22:20). The same applies to any automated dashboard: assign someone to confirm the inputs are current.

A sample executive dashboard, described in words

Describing the layout in plain language before building it lets reviewers agree on the content first. Here is a monthly executive view for a multi-site manufacturer.

Top row, four headline cards. Rolling twelve-month TRIR with the case count beneath it. Rolling twelve-month DART rate. Corrective actions overdue across all sites. Hazards and near misses reported per 100 workers this month against the twelve-month average. Each card shows the data-as-of date.

Second row, two trend charts. Recordable cases by month for the past 24 months, with near-miss and hazard reports plotted on the same time axis. Corrective actions opened and closed by month, so a growing backlog is visible as a widening gap.

Third row, a site table. One row per site: inspections completed on schedule, findings per inspection, actions overdue, days from hazard report to interim control, and the site owner's name. Sorted by actions overdue.

Bottom section, top risks and commentary. The five highest-severity risks from the risk register with the status of their controls and when each was last verified. Below it, three or four sentences from the EHS manager on what changed and what is being asked of the executive team.

The board version keeps the headline cards and the top risks, drops the site table, and expands the commentary. The site version replaces the trend charts with lists of overdue items and named owners.

The top-risk section depends on having a connected history of hazards, controls and incidents to draw on. Our short definition of Risk Intelligence in safety covers what that history makes possible.

Where SoterAI fits

The Soter Help Centre describes interactive dashboards that SoterAI builds on request in the chat. You describe the data and the view, such as incident counts by month and by type, and SoterAI assembles charts and summary cards in a panel beside the conversation. The data can come from your records, a connected app or an uploaded file, and you refine the dashboard by asking for changes, for example adding a severity breakdown (Creating Interactive Dashboards).

Two documented limits matter for the practices above. Each dashboard is a snapshot of the data at the time it was built and does not refresh on its own, so the data-as-of date still needs to be visible. The dashboard is a read-only view and does not change records. Separately, custom workflows can be scheduled hourly, daily, weekly, monthly or on a custom cron expression, and the Help Centre lists monthly safety KPI dashboards among its common examples (Scheduling Workflows). The definitions, owners and pairing rules in this guide remain your decisions; a tool can only calculate what you have defined. The 4 August 2026 interactive dashboards webinar walks through dashboards built from an incident log with date and severity filters.

Check which fields your corrective action records hold before you define the closure indicators. Open Corrective Action Plans.