Have you chosen sides in the Heinrich/Dekker debate yet? Zero Harm vs. Safety Differently? Focussing on the incident/accident data of what’s gone wrong rather than working on what’s gone right? People are the problem vs. people are the solution?
I do hope not because the chances are that they’re both right and that breaking through the statistical plateau (that is Great Britain worker fatality rates which have been flat for the past 5-6 years) will require a combination of both.
In a blog post from this time last year, Sidney Dekker related the case of a large health authority that experienced a 1 in 13 likelihood of causing further harm to patients [beyond whatever they were admitted for]. In analysing their safety record, the health authority could produce a definitive list of “what went wrong” but ultimately were unable to improve safety performance.
What jumped out at me was Dekker’s phrase, “After all, they had plenty of data to go on” with its implication, to this reader anyway, that data was not the solution.
Fast forward to this year and the APPG report on Working at Height. One of the major recommendations that came out the report was a call for better inspection and reporting through RIDDOR. This was amplified by expert Peter Bennett OBE, MD of PASMA who said, “we know that data collected does not accurately represent the true scale of ‘near misses’ in the workplace which is why we are calling for enhanced reporting methods, and an independent body who would confidentially collect data to inform industry and Government.”
More data. So which is correct?
The answer is that it all depends on what you do with the data, i.e. how is it used? If you ever studied statistics in school or University, you’ll doubtless have used stats tools from SPSS. Jack Noonan, one-time CEO of SPSS said that, “running a business using statistics is like trying to drive a car while only looking through the rearview mirror”.
Dekker was correct in that the Health Authority was trying to improve their safety performance by looking through the rearview mirror of their accident data. But, that’s not to say that data should be ignored…
The more data that is available, the more we can do with that data beyond simply reporting and categorising what has already occurred. For example, in the Working at Height example above, Peter Bennett proposed gathering information not just on the accidents but near misses, and not just recording the fact that a fall had occurred but other attributes of the fall such as the height fallen. The richness of data then allows us to move from the rearview mirror approach to one of predictive modelling – instead of simply using the data to tell us what has happened, using it to predict and prevent future events.
An example from the mobile phone industry helps illustrate the point. In working in the analytics industry in 2006, a client (one of the major UK mobile operators) was struggling with customer churn. Stats told them the quantity and demographics of customers leaving by plan type etc., but it was predictive modelling that allowed them to take the same data set and identify the single, clear indicator of a customer about to leave.
A customer requesting a new battery. As a result, if during a call with a customer service person a customer asked for a new battery, it was flagged as an indicator of churn and the agent could offer a favourable deal/promotion in the hope of keeping that customer.
Of course, that necessitates keeping a broader data set than otherwise would be thought necessary. After all, it’s difficult to know what the indicator of an incident/accident might be – who’d have thought that a new battery request was an indicator of churn?
Does predictive modelling work in improving workplace safety performance? Yes it does. There are many examples and case studies of improved safety performance arising from the deployment of predictive solutions. However, the point here is not to promote a specific solution but to try to highlight the fact that there are two sides to the Heinrich/Dekker debate and that each has merit.
Indeed, as we said in the opening, they’re probably both right. Breaking through the statistical plateau will not be achieved by rigidly supporting one camp or the other but by taking the best of each and leveraging the best that technology has to offer in delivering innovative solutions.If You Like This Post, Please Share It!