Safety Observations: Quality or Quantity? The Argument Is About the Wrong Thing

At a Glance

• The quantity side is right that observation is a habit, and habits need reps. A crew that files five observations a month is not looking.
• The quality side is right that a quota produces entries written to hit a number, and a thousand notes about housekeeping say nothing about the next serious injury.
• Both camps are arguing over a trade the form forced on them: volume cost detail and detail cost participation. When the record structures itself, that trade goes away.
• The number to watch is not observations per month. It is how many high energy observations reached someone who could act, and how many got closed.

The case for quantity

The people who argue for volume have the field on their side. Observation is a skill, and like any skill it fades without reps. A crew asked to look for hazards every day starts to see them, and a crew asked once a quarter learns to walk past them. Volume also builds the data. Five observations from a site are noise. Three hundred start to show which trade, which task, which hour, and which energy source keeps coming up, and that is how a pattern becomes visible before it becomes an incident.

There is a cultural piece too. When most of the crew is writing observations, the program belongs to the crew instead of the safety department, and everyone on site can tell the difference.

The case for quality

The people who argue for quality have usually inherited a program that chased volume and drowned in it. Set a quota and you get the quota, which is the compliance trap wearing a different hat. Ten observations per person per week produces ten entries, most written at 4:45 on Friday, most some version of "PPE worn correctly" or "good housekeeping." Nobody reads them. The safety manager ends up with 1,200 rows and no way to tell which three matter.

The deeper problem is what those entries are about. Housekeeping and PPE are low energy. They can produce a recordable, but they rarely produce a fatality, and the research is clear that driving down minor injuries does not drive down serious ones. The observation that matters is the one about the crew that skipped the isolation verification, or the load rigged off a single point, or the excavation that got deeper than the shoring was rated for. If the form never asks for the energy source, never asks about SIF potential, and never captures who was involved or what was done, the observation is a note, not a record. A thousand notes are not a safety program.

Both sides are describing the same failure

Notice the two arguments do not actually disagree. Both want observations that get read, get acted on, and change something, the same way a near miss only becomes a leading indicator when someone responds to it. The fight is over a trade the tooling forced on everyone. A complete observation took a trained person several minutes to write, and most of the workforce was never going to do that ten times a week. So programs picked. Push volume and accept thin entries, or demand complete records and accept that a handful of people would write all of them.

Neither choice was wrong. Both were the best available answer to a form that made detail expensive.

What changes when the record builds itself

The form is what has changed, because the tool now sits inside the work instead of after it. A worker on a lift can now say out loud what they saw, in their own words and in whatever language they speak, and the record fills in behind them. The energy source gets identified from the description. The SIF potential gets flagged. Location, people, task, and corrective action land in their fields, along with whatever custom field the company cares about. The worker did not fill out a form. They described a situation, and the situation became a structured record.

Volume no longer costs detail, because detail is no longer something the worker supplies by hand. The safety manager is no longer reading 1,200 rows, because the observations arrive sorted. The one about the unverified isolation surfaces first even though the worker gave it two sentences, and the 900 about housekeeping are still counted without being in the way. Across sites, the same energy source showing up in the same trade at the same hour becomes visible while the work is still being planned, without anyone building a pivot table on a Sunday night.

What to measure instead

Stop counting observations. The same logic that says PTP count is not PTP quality applies here. The number that matters is how many observations carried SIF potential and reached someone who could act on them the same day. Behind that, track how many corrective actions closed and how long each one took. Then look at the share of the crew that wrote at least one high energy observation this month, because that is the participation figure with meaning in it. Those measures reward volume and quality at once, and they quietly retire the Friday afternoon PPE entry without anyone having to police it.

FAQ

Should we keep observation quotas? Keep the expectation that everyone observes and drop the number. A quota produces entries. An expectation that the crew watches for high energy situations, backed by a tool that makes reporting one easy, produces records.

What makes an observation high quality? It names the energy source, makes a call on SIF potential, says who was involved and where, and either includes a corrective action or makes clear one is needed. If someone who was not there can read it and understand what nearly happened, it qualifies.

Will easier reporting just produce more junk? It produces more of whatever the crew is looking for. Train them to look for energy and you get more energy observations, which is exactly what the hazard recognition research found when crews were trained on the energy wheel. A system that sorts by severity keeps the rest from becoming a burden.

The observation itself was never the point. The point was always whether someone with the authority to change the work found out in time. Quality and quantity were both attempts to get there, and for the first time you do not have to pick one.

Share this with the safety manager who is still reading the spreadsheet by hand.