Buried deeply in the recesses of our kitchen cupboards is a Le Creuset Cast Iron, Lifetime Guarantee, Fondue Set (somewhere near the automatic bread maker and the juice making thing). It was a wedding present and represented the height of 1980’s sophistication… apparently.
It’s never been used as far as I can remember. All that boiling hot cheese dripping on you best chunky knit jumper, whist engaged in ‘polite conversation’, wasn’t really our scene. The fondue (and other useless kitchen equipment) nostalgia will hopefully make sense in a minute.
Swiss Cheese and Failure. Previously I’ve written about the James Reason Swiss Cheese Model which is widely used to illustrate how failure happens in complex systems. I’ve even had a go at trying to explain it in 300 seconds (link here).
I really like the Swiss Cheese Model (SCM), but it does feel a bit uncomfortable around the idea that successive ‘layers’ in a complex system are fixed and rigid. Everything I’ve experienced tells me that human systems are far more fluid as they change and respond to the circumstances.
Secondly, I never really worked out how you could extract learning from an incident using the SCM. It always seemed that the learning was drawn out as part of a post incident review. From practical experience, I would prefer to extract learning and make adjustments as you go along, rather wait for a ‘project post mortem’.
The Hot Cheese Model of Failure. This is a lovely paper by Karen Li and Harold Thimbleby from Swansea University, published in The journal of the Royal College of Physicians of Edinburgh, June 2014, Hot Cheese: A processed Swiss Cheese Model.
The paper does a great job of building on the James Reason SCM of failure and answering the questions I had about the flexible nature of the ‘layers’ and how you extract learning from failure.
Briefly what I took from the paper:
- Some of the ‘defensive’ layers introduced in a complex system to prevent failure, actually lead to more failure. This is due to the unpredictable nature of a complex system. There are some startling examples in the paper to illustrate this; from nuclear power plants to the administrative controls around prescribing medicines.
- The ‘layers’ in a complex system described by the SCM are actually very flexible and can be usefully described as ‘hot cheese’. The layers change shape, with ‘holes’ appearing and bits of melted cheese dripping onto the layer below. This gives us the Hot Cheese Model (HCM) description.
- In many organisations the reporting of failure, incidents and near misses though ‘theoretically promising’ is often inaccurate and ineffective. ‘Unremarkable errors’ rarely get reported and overall the reporting of incidents and failure is very far from 100%. They reference a number of studies which have demonstrated this.
- The ‘Fondue Pot’ is where the Hot Cheese ends up. This is the point were a lot of the learning from incidents typically takes place in many organisations – The Post Incident/Project Review.
- I’ve been involved in numerous post project learning events and incident reviews and the ‘Fondue Pot’ describes it perfectly. A gloopy blend of everything than has happened beforehand, all mixed together in the pot. The trouble is that most of the time you cannot distinguish anything specific, just ‘general lessons’.
- This seems like a good point to link to the description and history of the Cheese Fondue
Learning from Failure, Rubbery Cheese and the Fondue Pot. The paper introduces a final idea a round the concept of ‘rubbery cheese’. Basically this is about identifying failure and learning from it well before you get to the gloopy ‘Fondue Pot’ of the post incident review.
Each successive ‘layer’ in the system needs to be ‘rubbery’ so that less serious incidents, failures, learning points etc are ‘bounced’ out to a place were you can learn from them.
“To improve things, we need mechanisms that enable data capturing and monitoring within every system defence layer, so ‘unremarkable errors’ and near misses, along with the gradual development of an incident and its contributing factors, can be made tangible. In other words, we need our hot cheese to be rubbery, so forces and drips that didn’t made it through a defence can be bounced into the recording system as well, instead of disappearing without a trace.” (Figure 4)
Reading this I had Dave Snowden ringing in my ears. This sounds very much like the description of a process for the continuous capture of narrative into SenseMaker...
So What’s the PONT?
- Post project reviews can be a bit of a Fondue Pot. A gloopy mixture of learning points where the best you can expect are ‘generalised lessons’.
- To extract more meaningful ‘learning from failure’ it is better to have a system of continuous capture and sharing throughout the process, and learn as you go along.
- The Swiss Cheese and Hot Cheese Models provide a useful concept to start talking about failure, and the processes by which you might most usefully learn from it.