What is the difference between leading and lagging safety indicators?

Lagging indicators count something that has already gone wrong, an injury, a lost workday, a recordable case, and report it after the fact. Leading indicators count proactive activities done to prevent those outcomes, like near-miss reports and corrective actions closed, and they move before anyone gets hurt. Lagging tells you where you have been; leading tells you where you are heading.

Why are leading indicators better for prevention?

Because they give you signal between injuries, when there is still time to act. By the time a lagging indicator moves, the harm has already happened. Leading indicators reach into the wide base of near misses and at-risk conditions beneath serious injuries, so you can fix a hazard before it produces the injury a lagging metric would later record.

Is a rising near-miss count a bad sign?

Usually the opposite. A rising near-miss count generally means people are surfacing hazards that were always there, so you can now fix them, not that the plant suddenly got more dangerous. Near-miss reporting makes the invisible base of the safety pyramid visible, which is exactly why it is one of the most valuable leading indicators to start with.

Should you stop tracking lagging indicators?

No. Lagging indicators like TRIR and DART are still your outcomes and your external scorecard for customers, insurers, and regulators, so keep and benchmark them. The point is not to drop them but to stop managing by them alone. Pair them with leading indicators and steer day to day by the leading side, where the numbers move before injuries do.

Safety Leading vs Lagging Indicators

Q: What are examples of leading safety indicators?

Near-miss and hazard-report frequency, the count and quality of safety observations, corrective-action closure rates, audit and inspection findings closed on time, safety training completion, and preventive-maintenance or job-safety-analysis currency on high-risk equipment and tasks. Each is countable, improvable this month, and tied plausibly to fewer injuries.

Q: What makes a good leading indicator?

It is predictive (a plausible line to fewer injuries), actionable (a supervisor can move it this week), measurable and consistently defined, hard to game by doing less real work, and owned by someone accountable. It should also be balanced for quality, not just quantity, fifty observations that name real hazards beat a thousand box-checking ones.

Why leading metrics predict and prevent injuries, what a good one looks like, and how to build a balanced safety scorecard.

Lagging safety indicators measure injuries that have already happened, like TRIR and DART. Leading indicators measure the proactive activities that predict and prevent injuries, like near-miss reports, safety observations, corrective actions closed, and training completion. Lagging indicators tell you where you have been; leading indicators tell you where you are heading.

Every plant tracks lagging indicators, because customers, insurers, and regulators ask for them. Fewer plants track leading indicators well, and that is the gap that separates a program that reacts to injuries from one that prevents them. OSHA now runs a dedicated leading-indicators resource for exactly this reason. This post explains the difference, why leading indicators predict outcomes, what good ones look like, and how to combine both into a balanced scorecard. It is educational, not legal advice.

What is the difference between leading and lagging indicators?

Timing and direction. A lagging indicator counts something that has already gone wrong, an injury, a lost workday, a recordable case, and reports it after the fact. A leading indicator counts something you do on purpose to prevent those outcomes, and it moves before anyone gets hurt. TRIR and DART are lagging: they are the scoreboard after the game. Near-miss reporting rates, safety-observation counts, audit findings closed on time, and training completion are leading: they are the practice that decides the score. Both matter, but only one of them can be improved this week in a way that changes next quarter's injuries.

You need the mirror, but you drive by the windshield. A plant that manages only by lagging indicators is steering by looking backward.

Because they only report the crash, never the near-crash. By the time a lagging indicator moves, the injury has already happened; the metric is a memorial, not a warning. Worse, lagging numbers are noisy at the plant level. A site can go a year injury-free through luck as much as through safety, and a single bad event can spike a rate that was not actually trending. Managing purely on lagging data means waiting for harm to accumulate before you learn anything, and then over-reacting to whichever injury happened to land. Leading indicators give you signal in between the injuries, when there is still time to steer.

There is also a math problem hiding in lagging rates at the plant level. Most single sites simply do not have enough injuries in a year for the rate to be statistically stable, so a good year and a bad year can look very different when the underlying safety of the plant barely changed. Basing decisions on that noise leads to whiplash: chasing the cause of last month's spike, then declaring victory when the next month happens to be quiet. Leading indicators, because they count far more frequent events, produce a smoother, more trustworthy signal you can actually manage against.

Why do leading indicators predict injuries?

Because serious injuries sit on top of a much larger base of near misses and at-risk conditions. The long-observed pattern in safety is a pyramid: for every serious injury, there are many minor injuries beneath it, and beneath those, a far larger number of near misses and unsafe conditions that produced no injury at all, this time. That base is invisible to lagging indicators, which only count the top. Leading indicators reach down into the base: a near miss is a free lesson about a hazard that has not yet reached the top of the pyramid, and each one you catch and fix removes a path to an injury before it is walked.

Lagging indicators measure the narrow tip of the pyramid; leading indicators measure and act on the wide base of near misses and at-risk conditions that feed it.

This is also why the practice of near-miss reporting is the single most valuable leading indicator most plants can start with: it makes the invisible base visible and countable. A rising near-miss count is not bad news; it usually means people are surfacing hazards you can now fix.

The pyramid also explains a trap plants fall into. When a site posts a clean lagging record, leadership relaxes, and the very activities that produced the clean record, the reporting, the fixing, the watching, quietly wind down because nothing seems to be wrong. The base refills with unaddressed near misses, and a serious injury eventually climbs back to the top, seemingly out of nowhere. It was never out of nowhere. The leading indicators were falling the whole time; nobody was watching them because the lagging number looked fine.

What are examples of leading safety indicators?

The activities that shrink the base of the pyramid, each one countable and improvable. The most useful across manufacturing:

Near-miss and hazard-report frequency the volume of hazards surfaced before they cause harm.
Safety observations the count and quality of structured observations from a behavior-based safety process.
Corrective-action closure the share of safety corrective actions closed on time and verified effective.
Audit and inspection performance findings per safety audit and how quickly they are closed.
Training completion the percentage of required safety training that is current.
Preventive maintenance and job-analysis currency on high-risk equipment and tasks.

What makes a good leading indicator?

Not every activity you can count is worth counting. A good leading indicator passes a short test, and a vanity metric fails it.

It is predictive. There is a plausible line from this activity to fewer injuries. "Posters hung" fails; "near misses closed" passes.
It is actionable. A supervisor can do something this week to move it. If nobody can influence it, it is a statistic, not a lever.
It is measurable and defined. You can count it the same way every time, so the trend means something.
It is honest under pressure. It cannot be gamed by doing less real work, count of hazards fixed resists gaming better than count of forms filed.
It is owned. Someone is responsible for the activity and its number, not just for reporting it.
It is balanced with quality, not just quantity. A thousand box-checking observations are worse than fifty that name real hazards, so pair volume with a quality check.

Run any candidate metric through that test. If it is predictive, actionable, honest, and owned, it belongs on the board. If it is just easy to count, it will make the report look busy and change nothing. The failure mode to watch for is the metric that measures activity for its own sake, forms submitted, meetings held, boxes ticked, with no link to a hazard being found or fixed. Those numbers always climb, which is exactly why they are dangerous: they feel like progress while the base of the pyramid quietly refills.

How do you build a balanced scorecard?

Put leading and lagging side by side, and manage on the leading side. The lagging indicators, TRIR, DART, severity, are your outcomes and your external scorecard; keep them, benchmark them, and report them. But the day-to-day management happens on the leading side, where the numbers move before injuries do. A practical scorecard shows two or three lagging rates next to three or four leading metrics, reviewed on a fixed cadence, so a flat injury rate paired with a rising near-miss count reads as progress rather than mystery. This is the same balanced view laid out in our guide to safety KPIs and it is why a lone TRIR calculation should never be the only safety number a plant reports.

Keep the set small. The temptation is to track twenty leading indicators and drown the signal; three or four that pass the test and get acted on beat a dashboard nobody reads. And give the leading side real teeth by reviewing it where decisions get made, in the safety committee, in the daily production huddle, wherever the plant actually acts, rather than filing it in a monthly report that arrives after the moment to act has passed. A leading indicator that nobody looks at until quarter's end has quietly become a lagging one.

What do the standards say?

From the primary sources:

OSHA's Leading Indicators resource defines them as proactive, preventive measures that reveal problems in a safety program before injuries occur, and it collects examples and research for building them.
OSHA's Program Evaluation and Improvement guidance treats tracking both leading and lagging measures as part of a functioning safety program.
On the lagging side, the Bureau of Labor Statistics' Injuries, Illnesses, and Fatalities program is the benchmark source, private employers reported 2.6 million nonfatal injuries and illnesses in 2023, the very outcomes that leading indicators exist to reduce.

The catch with leading indicators is that they only work if the activity behind them is easy to do and honestly captured. A near-miss program that requires a paper form and two signatures produces a leading indicator of zero, not because the plant is safe, but because reporting is a chore. Harmony connects machines, software, and paperwork into one operational layer with no rip-and-replace: near misses, observations, audit findings, and corrective actions become structured data captured on the tablet already at the station, so the leading behaviors take seconds and the numbers are real. Live dashboards show the leading side climbing and the open-fix list shrinking, and AI search returns cited answers across those records so a rising hazard-report trend can be traced to the jobs and machines behind it, part of the everyday shape of connected worker technology (see how it works). Harmony is not a safety-compliance product, but it makes the windshield metrics easy enough to actually watch.