A safety incentive program rewards employees for safety performance. Programs that reward low injury rates can backfire by discouraging workers from reporting injuries, which OSHA's recordkeeping rule (29 CFR 1904.35) treats as unlawful when it suppresses reporting. Programs that reward positive, leading behaviors avoid that trap.

The pizza party for a month with zero recordables is one of the most common safety programs in American manufacturing, and one of the most quietly counterproductive. The intent is good: make safety visible, celebrate good outcomes. The effect, too often, is a crew that quietly stops reporting so it does not be the one that costs everybody the party. This post covers why rate-based incentives go wrong, what OSHA actually says about them, and how to build a program that rewards the behaviors that prevent injuries instead of the silence that hides them. It is educational, not legal advice.

What is a safety incentive program?

Any program that ties a reward to safety, and the reward's design decides whether it helps or hurts. Broadly there are two kinds. Rate-based (or outcome-based) programs reward the absence of injuries: a bonus for an injury-free month, a prize drawing tied to a low incident rate, a manager's evaluation pegged to their unit's recordable count. Behavior-based (or participation-based) programs reward the things people do that make injuries less likely: reporting a hazard, turning in a near miss, completing training, taking part in a safety observation. Same word, "incentive," opposite effect on reporting.

Why do rate-based incentives backfire?

Because they put a price on reporting, paid by the whole crew. When the reward depends on zero injuries, a reported injury is the thing that costs everyone the bonus. Nobody has to write a policy against reporting; the incentive does it quietly. The hurt worker weighs a sore shoulder against being the person who broke the streak, and some of them choose to stay quiet. The injury still happened, you just lost the data, and with it the chance to fix whatever caused it.

How a rate-based incentive suppresses reportingThe suppression cycleREWARD FOR ZERO INJURIESPRESSURE NOTTO REPORTHAZARD STAYS HIDDENno report, no fixWORSE INJURYthe one you can't hide
A rate-based incentive doesn't remove injuries; it removes reports. The hazard that was never written down is the one that comes back bigger.

The deeper damage is to the data. Safety improves when hazards surface early and get fixed. An incentive that suppresses reporting starves the whole program of its inputs: fewer near misses, fewer first-aid cases, a quieter log that looks like progress and is actually just darkness. You cannot fix what you cannot see, and a rate-based bonus is a machine for not seeing.

There is a second, uglier failure mode. When the prize is large enough, the pressure to stay quiet does not stay individual. Co-workers start leaning on the person who is about to break the streak, a supervisor's bonus rides on their unit's clean number, and reporting an injury becomes a social act with consequences well beyond the injury itself. That is precisely the environment OSHA's anti-retaliation rule was written to prevent, and it is why a well-meaning program can drift into something that looks a lot like coercion without anyone intending it. The incentive did the coercing.

What does OSHA say about safety incentives?

They are not banned, but they cannot be run in a way that discourages reporting. OSHA's 2016 recordkeeping rule amended 29 CFR 1904.35 to prohibit retaliating against employees for reporting work-related injuries and illnesses, and 1904.36 ties that to the anti-discrimination protection of the OSH Act. In an October 2018 clarification memo OSHA made its position practical: rate-based incentive programs are permissible under 1904.35(b)(1)(iv) as long as they are not implemented in a way that discourages reporting. If withholding a prize because of a reported injury is paired with adequate precautions to ensure employees still feel free to report, OSHA generally will not cite it; but the memo warns that a bare statement of "we don't retaliate" may not be enough when reporting costs someone a substantial reward.

Read plainly: the law does not tell you to scrap incentives. It tells you the incentive cannot be the reason an injury goes unreported. The safest way to satisfy that, and to get a better program in the bargain, is to stop rewarding the number and start rewarding the behavior.

What should you reward instead?

Reward the leading behaviors that predict fewer injuries, not the lagging count of injuries avoided. This is the same logic behind leading versus lagging indicators: a low injury rate is an outcome you can only measure after the fact and can game by hiding, while hazard reports, near misses, observations, and training completion are activities you can count honestly and grow on purpose. Reward those and you get more of them, and more of them is exactly what prevents the next injury.

Rate-based versus behavior-based incentivesReward the behavior, not the numberRATE-BASED (lagging)BEHAVIOR-BASED (leading)rewards: zero recordablesmeasured: after the factgamed by: hiding reportseffect on reporting:rewards: hazard reportsnear misses turned inobservations, training doneeffect on reporting:↓ pushes it DOWN↑ pulls it UPquiet log, hidden hazardsmore signal, earlier fixes
Rate-based incentives reward an outcome you can hide; behavior-based incentives reward activity you can count and grow. Only one of them makes hazards more visible.

Concrete rewards that pull reporting up: recognition for the near miss that led to a fix, points or small rewards for hazard reports and safety observations, team recognition for training completion and audit findings closed on time, and celebrating the person who stopped a job over a safety concern. Pair this with a genuine behavior-based safety observation process and the incentive reinforces the exact activity that program depends on.

The mindset shift is worth stating plainly. A rate-based program treats the injury count as the thing to move, and the fastest way to move a count is to change what gets counted. A behavior-based program treats injuries as the output of a system and rewards the maintenance of that system, the reporting, the fixing, the watching out for each other. One rewards a scoreboard; the other rewards the work that actually changes the score. And because the rewarded behaviors are visible and countable, you never have to wonder whether the program is being gamed, growth in hazard reports is the point, not a red flag.

How do you design a program that doesn't backfire?

Build it around participation, make reporting safe, and keep the rewards small enough that honesty never costs too much.

  1. Reward inputs, not outcomes. Tie rewards to hazard reports, near misses, observations, training, and closed corrective actions, activities people can do honestly, not to an injury count they can suppress.
  2. Never penalize reporting. No reward, streak, or evaluation should get smaller because someone reported an injury. That is the line OSHA draws, and the line that keeps the data honest.
  3. Keep the stakes modest. The bigger the prize riding on silence, the stronger the pull to stay quiet. Frequent small recognition beats one large annual jackpot tied to a number.
  4. Recognize the stop-work call. Publicly reward the person who halted a job over a safety concern, even, especially, when the concern turned out to be minor. That is the behavior you most want repeated.
  5. Make it team and participation based, so a single injury does not wipe out a group's reward and turn co-workers into pressure on the person who got hurt.
  6. Measure the leading behaviors you are rewarding, and feed them into the plant's safety KPIs so you can see participation rising.
  7. Ask the crew. Workers can tell you instantly whether a program feels like encouragement or a trap. If reporting an injury feels like letting the team down, redesign it.

How do you know if your program is working?

Watch reporting go up, not down. This is the counterintuitive part: a healthy incentive program often makes the near-miss and hazard-report numbers climb, because people are surfacing more, not because the plant got more dangerous. Rising reports with falling severity is the signal you want. If your incident log goes silent right after you launch a big rate-based bonus, that is not success; that is the suppression cycle starting. Track participation, hazard-report volume, and time-to-close alongside the lagging injury rates, and trust the leading numbers to tell you first. Give it a quarter or two before you judge it, because the early spike in reports is the program working as designed, not the plant falling apart. The steady state you are aiming for is high participation, a short list of open hazards, and injuries that keep drifting down because their causes keep getting caught upstream.

What do the standards say?

From the primary sources:

The whole debate comes down to one design choice: do you reward the number or the behavior? Rewarding the number quietly buys silence; rewarding the behavior buys visibility. And visibility only helps if the reporting itself is frictionless. Harmony connects machines, software, and paperwork into one operational layer with no rip-and-replace: a hazard report or near miss takes thirty seconds on the tablet already at the station instead of a paper form and two signatures, so the leading behaviors you are trying to reward are easy to do and easy to count. Live dashboards show participation and open fixes, and AI search returns cited answers across reports, observations, and corrective actions, part of the everyday shape of connected worker technology (see how it works). Harmony is not a safety-compliance product, but it makes the honest behavior the easy one. A periodic safety audit can then check that the program is lifting reports, not burying them.