Ian Travers of HSE presented at the IOSH Hazardous Industries specialist group, 13 February 2007. He was referencing a new guidance document HSG254 on leading indicators.
A good analogy was presented. If someone arrives late for a meeting you have a lagging indicator of failure. Arriving on time may be a leading indicator. Thinking about this, I would say arriving on time is not a true indicator. However, if we were to know what speed that person had driven on the journey we would know that either they had plenty of time to arrive safely (kept below speed limits) or if they had been in a rush (above speed limits at times) we would have leading indicator of how safe the arrangements really had been, even though success had been achieved.
Unfortunately putting this into practice in an industrial setting seems very difficult. It was notable that Ian did not provide any good examples in his presentation. He did give some idea of how to go about identifying indicators, with a key steer being to have a good idea of "what success looking like." This needs to be considered carefully, because for example it is not a case of saying what a permit-to-work looks like but what is the system intended to deliver.
Ian said there was a need for better indicators because audit is usually too infrequent and compliance focused, whilst work place inspection rarely addresses critical controls. However, before going down the route of developing leading indicators it is important to answer the following 3 questions:
1. Why do it?
2. What will you do with the data?
3. How will it influence safety performance?
It is unlikely that there will be any generic leading indicators. Even company wide are unlikely to be effective as having enough indicators to cover all the requirements is likely to be too many for them to be effective.
At the end of the presentation I concluded that leading indicators may not be the solution we are looking for (at least at this time). Whilst in theory they are exactly what is required, putting it in to to practice is very difficult. We may do better by using lagging indicators better, especially learning from near-misses and process disturbances.