Another day, another horror story about something going wrong. It ranges from the highest threat (human lives being lost and financial meltdowns) to more trivial issues (no bread at the supermarket).
The majority of outcomes are entirely predictable and manageable if you can identify risk at the earliest possible stage.
Traditional approaches to risk management have focused on risk managers using (at best) an industry matrix and process to identify and score risk, along with a set of mitigating and reduction type actions. Technology driven? Nope.
Organisational learning boards (where established) will also implement key actions to embed learning within the people, process and systems of the organisation.
But the horror stories still keep happening. Even with the meetings, discussion, blame, and more action, it all keeps coming. We all say, "This can't happen again!" but it does.
The problem is we only tend to focus on risk when it is high or has already become a reality and by then it is too late.
Often the indicators or predictors for the risk are not identified in an 'evidential and empirical way' (i.e. what drives the risk?). Also, the recency, frequency and gravity interdependencies are not monitored and risk assessed day in, day out, hour by hour, etc. Why?
Collecting data on such things is very difficult for us to achieve due to capacity, capability and technology. If you can capture vast amounts of data from multiple sources, how do you make sense of it in a timely fashion?
What happens when someone is sick or on holiday? We are just not good at this in an operational environment. You need the skills and strengths that technology provides.
People have their strengths. Let them make the decisions and take action from the accurate evidence.
Predictive analytics is a great solution to help identify risk at the earliest opportunity. Before there is financial, reputational or human loss.
First, identify your target. The target is preventing domestic homicides. Identify the most relevant datasets that link to this target internally and externally. Clean and prepare your data. Build your model. Involve the key business practitioners all the way. Train your model on high risk cases and you're nearly there.
Next, flush though real data. Evaluate the results. Work with the 'front line'. Tweak your model some more. Understand the results? Happy?
Good, now implement. Evaluate. Tweak. Pilot. Work with the business to devise prescriptive corrective interventions. Review. Roll out. Evaluate ROI.
In this example, at the earliest point that critical risk is identified and visible, interventions and governance are asserted and a life is saved.
1) Reduce demand through preventing things before they happen. How many resources would be diverted from business as usual (BAU) to investigate a homicide?
2) Reduce risk/threat/harm by presenting an opportunity to intervene before an event happens. Are a number of lower level assaults a precursor to a more serious offence?
3) Reduce financial loss that occurs when things go wrong. How much does a homicide cost?
4) Improve reputation by reducing the number of major incidents and showing appropriate risk control measures and interventions are in place.
5) Improve staff engagement by providing the tools to help them understand and manage risk more effectively (and you will see a reduction in blame culture).
Difficult to achieve? No. Technology is now readily available and even cost effective for small to medium enterprises. Having the right people in place is more difficult, however, larger organisations can build a team, smaller ones can hire data science services through a low risk and cost subscription solution.
Article written by Sean Price
Want more? For Job Seekers | For Employers | For Contributors
