Food safety culture moved from concept to requirement when it was written into the Codex General Principles of Food Hygiene and adopted into GFSI benchmarking requirements. Most sites responded the way organisations respond to any new soft requirement: posters, a slogan, a toolbox talk, and an annual survey asking whether staff think food safety is important.
Everyone says yes. The survey is filed. Nothing changes.
The problem is not that culture is unmeasurable. It is that the wrong things are being measured. Culture is not what people say they believe. It is what happens when believing it is inconvenient.
Measure behaviour, not opinion
The most direct culture measure available is observed conformance — what people actually do when the procedure and the pressure point in different directions.
Practical behavioural measures:
- Hygiene and PPE conformance rate from structured observation rounds, recorded as a percentage rather than as pass/fail
- Handwashing compliance at entry, observed rather than self-reported
- Record completion timing — are records completed at the time of the activity, or back-filled at end of shift? Back-filling is one of the clearest culture signals available and is directly observable
- Changeover and cleaning discipline under schedule pressure specifically, not under normal conditions
The discipline here is to observe at times when conformance is hardest, not easiest. A 98% hygiene conformance rate measured on a quiet Tuesday says less than an 80% rate measured during a rushed changeover.
Measure whether the system makes the right thing easy
Culture is frequently blamed for what is actually a system design failure. If an operator has to walk 40 metres to find a clean scoop, the scoop will be reused. That is not a culture problem, and no amount of training will fix it.
Enabler measures worth tracking:
- Availability of what the procedure requires — consumables, cleaning materials, verification supplies, correctly sized PPE. Stockouts of these are a leading indicator of conformance failure
- Time allocated versus time required for cleaning and changeover. If the schedule allows 30 minutes for a 45-minute clean, the clean will be shortened
- Equipment condition — how much of the estate is hard to clean, and how long items sit on the maintenance backlog
- Procedure usability — can the operator find and understand the relevant instruction in under a minute?
Measure reporting, and read it backwards
Near-miss and concern reporting rate is one of the most informative culture metrics, and it is almost always misread.
A low reporting rate is treated as good news. It is usually the opposite. In a site with a genuinely open culture, people report more, because reporting feels safe and worthwhile. A near-zero reporting rate typically means either that nothing is being noticed, or that people have learned that raising something produces blame or produces nothing.
Track:
- Reports raised per month, by area and by role — front-line reporting rates matter more than management ones
- Time to response — how long before the person who raised it hears back
- Closure rate and visible outcome — what proportion led to a change the reporter could see
That last one determines whether reporting continues. People stop reporting when reports disappear.
Measure what leadership does under conflict
This is the measurement that matters most, and the one almost nobody records.
Food safety culture is established by what happens when food safety and throughput conflict directly. Every other signal is downstream of it.
Two things are measurable here:
- Escalation outcomes — when a food safety concern was raised that would delay or stop production, what was decided, by whom, and how quickly? Log these. The pattern is the culture.
- Hold and release decisions — how often is product released under concession, on whose authority, and on what evidence?
A site that has never stopped a line for a food safety concern has either an exceptionally capable system or a culture where stopping the line is not genuinely available. The escalation log distinguishes the two.
Designing a survey that returns something useful
If a survey is used, the question design determines whether it returns anything.
Questions that produce no information: Is food safety important here? Do you understand your responsibilities? Everyone answers correctly.
Questions that produce information:
- Have you ever seen something that worried you and not reported it? What stopped you?
- If you needed to stop the line for a food safety concern, could you? What would happen next?
- Do you have the time and materials you need to do the cleaning as written?
- When was the last time something you raised led to a change?
Run it anonymously, segment results by area and shift rather than reporting a single site-wide score, and publish what changed as a result. A survey with no visible consequence trains people not to answer honestly next time.
Practical takeaways
- Stop measuring belief and start measuring behaviour under realistic pressure.
- Treat consumable stockouts and maintenance backlog as culture metrics — they measure whether the system permits conformance.
- Read a low reporting rate as a warning, not an achievement.
- Keep an escalation log recording every food safety concern that conflicted with production, and what was decided. This is the highest-value culture record available.
- Segment every measure by shift and area. Site-wide averages hide exactly the variation you need to see.
- Close the loop visibly. Culture measurement that produces no observable change actively damages the culture it was measuring.
Culture is not the posters. It is what the site does on its worst day, and that is entirely measurable if you are willing to record it.