Tailoring Safer Systems

Measuring safety

There is a lot of talk about measuring safety. That is something which is easier said than done. This blog shares some reflections.


Measuring what?

Before starting to measure, one needs to know what one is measuring. How you define safety will determine what you measure and how you measure. Let us illustrate the problem with three quite common views on safety. As you will see, none of them covers the subject entirely and all have advantages and disadvantages.

Safety as compliance

A very basic way of thinking: safety is following the safety rules. Being compliant with these rules is being safe. This correspondents to the almost automatic reaction that many people have after an accident: if only they had followed the rules, this would not have happened. Many investigations therefore focus on breaches of protocol and deviations. Also, in ‘normal’ situations there is emphasis on compliance. Wear the mandatory safety gear. Hold the railing. Striving for compliance also appeals to the human tendency towards conformity. We are social creatures, after all.

Safety rules are important. They are a basic form of how we teach safety: “Don’t touch the stove, it’s hot!” “Watch left, right, left before crossing the street.” These things we teach our kids, our workers, etc. Safety as compliance works reasonably well in rather simple, ordered and predictable systems. In these situations, you have a reasonable chance to foresee what can happen and conceive actions to deal with variations. If you are on known territory, you can deal with the things that happen by applying prescribed routines. Following ‘best practice’ means acting safely, while acting outside of these scripts is regarded as unsafe.

Safety rules are not perfect, however. We live and work in a world with a lot of variability and we have a limited amount of foresight. This means that we cannot write rules for every eventuality. If we could, then the rules were impossible to handle because of their sheer volume. Besides, rules depend on context. In London it is smarter to look right, left, right before crossing while this is not the best strategy for Zürich.

Rules are compromises and may sometimes not be enough to keep you safe. Even if you follow all the traffic rules, you can have an accident. For example, when others do not follow the rules. In some situations, following rules is even the unsafe option. One (in)famous example is the Piper Alpha disaster where the people that followed the emergency procedures died while the ones who ignored the procedures and just jumped overboard survived.

Safety as an absence of accidents

Go out on the street and ask a hundred randomly chosen people, “What is safety?” Chances are that many will answer something in the line of “Not having any accidents”. Thinking this way makes intuitively sense to most people. It feels right because in our minds safety and accidents are very much linked. When we do not have any accidents, we have been safe. Or have we? Actually, not necessarily… That nothing has happened does not mean that things are safe. In many cases it only means that nothing has happened yet. Although it can very well be that nothing happens ever.

A simple test is to reverse the definition and see whether it still works. Is “The absence of accidents is safety” true? Absence of accidents can be achieved by other ways. Randomness or luck are possible factors. Your definition of accident is another. Whether people choose to report accidents yet another. However, accidents do give an indication about safety, or rather unsafety. An accident can be regarded as a manifestation of risk, bringing us to the next definition.

Safety as acceptable risk

Whatever you do, there is some risk involved. We cannot avoid this. We even want some risk, but not too much. We need to compromise between various goals (financial, safety, production, quality, etc.), between uncertainty and control. We have only limited resources (money, time, expertise, etc.). Therefore, we must make trade-offs and search for balance.

This view of safety appeals to rational creatures. It suggests deliberation and decision based on ‘facts’. We will always face risks; we just have to make sure that they are acceptably low. The question is therefore what level the right level of risk is. We should obviously try to put as much ‘distance’ as possible between ourselves and the hazard and the possible negative futures the hazard could lead to. But we do not want too much distance either. It has to be practicable and affordable. Besides, some hazards we actually do desire. Just think of drinking coffee. We want our coffee hot, but we do not want to burn ourselves. Therefore, we tend to sip our coffee carefully at first, or maybe blow a bit on it, instead of gulping it down at once.

The view of safety-as-acceptable-risk is useful, but there are also some drawbacks. One is its reliance on knowledge, another is how it can lead to quantitative approaches to risk that look more objective than they are, that it may lead to a static view of safety, and the problem of monitoring the risk level. Then there is of course the problem of who decides what is ‘acceptable’ and based on what. Who determines what is included in the assessment and what factors weigh in (and how much)? Who is allowed to participate in the process and how can they participate in the process? What language is used during the process and in the communication of the results?

One example of the latter is how consequences are selected and expressed. Certain risk assessments focus on fatalities, but those are often not the only bodily consequences. So, what to do with injuries? Should one choose a number of severe injuries that equals a fatality? Or should we, as one often sees, translate fatalities and injuries into monetary units? Is that really a good, and fair measure? Can you put a number on a human life? And if so, what number? Sure, you can estimate one person’s economic contribution to society and his/her family, but a person is so much more than his/her economic contribution.


The above views of safety all bring their own ways of measuring safety. Regard safety as compliance and you may be tracking citations from the inspectorate, or observations of unsafe acts (e.g. not wearing protective equipment). If safety is seen as the absence of accidents, you will naturally follow up on accident and injury reports. Those who adopted a risk view of safety may have some kind of a risk register, present the most important risks in a risk matrix or heat map and follow up on actions to control the risks.

How you define safety will influence your choice of things you measure – and vice versa! What you measure may very well become your definition of safety, consciously or not. If corporate policy, an ISO standard or the regulator requires you to record accidents and near misses as part of your monitoring, it will become very natural to talk about these metrics when someone asks about “How are we doing at safety?”

Another challenge is that management dashboards and scorecards allow only limited space for the presentation of how things are going. Managers are busy people and they would very much like to get clear, concise, unambiguous and short answers. However, safety is a complex phenomenon. Therefore, we need a variety of measures to give a reasonable description. No one view captures everything. Every view shows some elements of safety, but never the full picture. A good answer thus needs rich information and nuances. Here is a tension between space and attention available and what is needed to give a high-quality answer.

Dumbing it down into an easy measure, no matter how intuitive, will not do justice to the subject. A fatality/injury-based metric only captures a tiny part of a very complex phenomenon. It would be like describing a river exclusively by its temperature – which, by the way, rather depends on its surroundings, location and season than on ‘itself’, just as injury rates may correlate stronger with the context than with safety efforts initiated by the organisation. A trade-off between thoroughness and efficiency is inevitable and carefully addressing this in the management system is essential.


This blog is an adapted and abbreviated chapter from the book If You Can’t Measure It… Maybe You Shouldn’t. Reflections on Measuring Safety, Indicators, and Goals.

Available from: https://www.amazon.com/dp/8269037729

Safety Mythologist and Historian

One Comment

Leave a Reply

Your email address will not be published. Required fields are marked *