How many air samples do you really need?

This page explains what the rules and common practice call for. Every site is different, and what a survey or a filing needs at your facility is something we work out with you after we see it.

The short answer: more than one, and fewer than you fear. The number depends on how many different jobs you have, not on how many workers.

That second point is why proposals for the same plant can differ by a factor of 3.

Headcount is the wrong place to start

The question usually arrives as "we've got 40 people. How many samples?"

Headcount hardly matters. What matters is how many different exposure situations there are in the building.

A 100-person plant where everyone runs the same machine, doing the same thing on the same shift, is one exposure situation. A 20-person job shop where one person welds, one grinds, one runs a saw, one mixes coatings and one does everything depending on the day is five or six, and it needs more samples than the larger plant.

You are not sampling a room. You are finding out what specific people breathe while they perform specific tasks.

Similar exposure groups: the concept that sets your bill

Air monitoring is organized around the similar exposure group (SEG). An SEG is a set of workers whose exposure is expected to be about the same, because they do similar work with similar materials under similar conditions.

You sample a few people from each group, and the results represent the whole group. That is what keeps the cost of air monitoring down: you do not sample everybody, only enough of each group to be confident about all of them.

Defining the groups correctly is therefore the whole game. If the groups are wrong, no amount of sampling will save you, because you will have precise data about a group that does not really exist.

What separates one group from another

• Task. Welding and grinding are different exposures, even in the same bay.
• Material. Welding stainless steel produces hexavalent chromium, and mild steel largely does not. The welder and the booth are the same, but the exposure is completely different.
• Process. Enclosed and ventilated, or done at an open bench.
• Location. The same task near a unit that supplies fresh air, or in a corner.
• Shift. Second shift often runs different products and different volumes, and sometimes different ventilation.
• Time on the task. Someone who welds 2 hours a day and someone who welds 8 are not in the same group, even when the work is identical.

What does not separate them

Job title, department, pay grade and reporting line do not define an exposure group. Org charts are built for managing people, not for tracking exposure. When the groups follow the org chart, the monitoring usually measures categories that have nothing to do with what people breathe.

How many samples per group?

There is no fixed number. The count depends on how many distinct jobs you have and how much the work varies from day to day, and it is settled after a walk of the floor, not from a formula.

The reasoning is worth understanding. Workplace exposure varies. The same worker doing the same job on two different days can get quite different results, sometimes by a factor of 2 or more. That spread is normal, and it is not a sampling error.

Several things follow from that:

• One sample tells you almost nothing. It shows what one person breathed on one day, and it cannot tell you whether that day was typical, high or low. If it comes back at 60% of OSHA's permissible exposure limit (PEL), you truly do not know whether you are comfortably within the limit or often over it.
• Even a couple of samples on a job begin to show the range. Two results that agree, well below the limit, tell you more than one result ever can.
• More samples sharpen the picture of the upper end, which is what OSHA cares about, because the standard is concerned with the exposure workers actually receive, not the exposure they receive on average. How many more depends on how much the results scatter.

When you can take fewer

• The result is clear-cut. If samples come back at 3% of the PEL, more sampling will not change the conclusion. It costs little to be confident that you are nowhere near the limit.
• Worst-case sampling. Deliberately sample the worker with the highest exposure on the day with the highest exposure, and compare the result to the action level, which is set below the PEL. If that result is below the action level, everyone below that worker is too. This is a sound and efficient strategy.
• Small groups. A group of 2 people needs far fewer samples than a group of 40.

When you need more

• Results cluster near the action level or the PEL. This is the expensive case, and it is common. When your data falls on both sides of a limit you cannot conclude either way, and the only way out is more data.
• Results vary widely between samples. Wide scatter means your group probably is not really one group.
• You have to defend something. That could be a contested OSHA citation, a workers' comp claim, or a decision to stop monitoring under a standard that requires you to justify it.

Worst-case or random sampling: choose deliberately

Worst-case (targeted) sampling deliberately goes after the highest exposures: the worst job, the worst day and the worst conditions. It is efficient and conservative, because if the worst case is clean, everything is clean.

Its limitation is that it describes the top of the range, not the full spread of results. You cannot use it to make claims about typical exposure, and you cannot build a reliable statistical picture from it.

Random (representative) sampling takes samples across the group without targeting anyone. It gives you a true spread of results and supports statistical conclusions.

It has limitations too. It costs more, and if you happen to sample only on quiet days you can miss real exposures.

A well-designed survey usually uses both, with worst-case sampling to establish the top of the range and representative sampling to describe the group. What matters is that the choice was made deliberately and documented. Sampling whoever happened to be available is not a strategy, and the report will show it.

Full shift means the full shift

A valid 8-hour time-weighted average (TWA) sample covers the full shift, or close enough to it that you can justify the part you did not sample.

The lead standard spells this out by requiring full-shift samples of at least 7 continuous hours. Other standards are less specific, but the principle holds: you cannot sample for 2 hours, multiply by 4 and call it a shift average. Exposure is not uniform across a shift, and that shortcut is one of the first things people challenge.

Where a task lasts less than a shift, the right approach is task-based sampling. You sample the task, record how long it lasted, and then calculate the TWA with the unexposed time counted in. That is sound, and it often tells you more. Cutting a shift sample short is a different thing, and it is not sound.

Short-term limits are a separate obligation. Beryllium, benzene, formaldehyde and methylene chloride each have a 15-minute short-term exposure limit (STEL), and asbestos has a 30-minute excursion limit. These need their own short samples. A full-shift TWA sample cannot show that you meet a short-term limit, because averaging over 8 hours is exactly what hides a 15-minute spike.

What defensible air monitoring looks like

Your air monitoring may be challenged by an inspector, by an expert hired by someone suing you, or by an insurer. These are the things they will look for:

1. A written rationale for the exposure groups. Why these groups, and on what basis.
2. Enough samples per group to support the conclusion drawn.
3. Full-shift sampling, or properly calculated task-based sampling, with the times recorded.
4. Analysis by a nationally accredited laboratory (accredited under the federal NVLAP program), with the analytical methods named.
5. Calibration records for the sampling pumps, from before and after the shift.
6. Field notes written at the time, recording what was running, what controls were operating, and whether anything was out of the ordinary that day.
7. Employee notification within 15 working days, where the substance has its own standard.

Two notes on item 7. For hexavalent chromium, the clock runs 15 work days from the day you make the exposure determination, meaning the day you decide what the exposure is. That is not the day the results arrive, which is the trigger most of the other standards use.

Also, several hundred substances are covered only by OSHA's general air contaminants table (Table Z). They carry no notification requirement at all, because Table Z sets a limit and nothing else.

Air monitoring that is missing items 1, 5 or 6 is the kind that falls apart under close review. The numbers may not be wrong, but nothing shows what the numbers were measuring.

The honest answer to "can I take only one sample?"

You can, and nothing stops you.

What you will have is one number, and one sample cannot support a decision either way. If it comes back low, you cannot defend a decision to stop monitoring, and if it comes back high, you do not know whether that was typical or a bad day. You will be sampling again anyway.

The added cost of a second or third sample is small on a visit you are already paying for, because the field day is the expensive part, not the sample. The added value is the difference between data and a single anecdote.

→ What an air monitoring survey costs, with the sample count math shown

How the job gets planned: the exposure groups are defined with you, after a look at your process, and the sample count goes in writing before the sampling starts. The group definitions usually come together in one conversation.