December 02, 2020 • 3 min read
What role can predictive analytics play in improving safety?
The best time to start applying predictive analytics to your safety data is now
If you think your safety data isn’t refined enough, complete enough or clean enough to use predictive analytics, think again. Heather Stewart, Vice President, Products shares why.
Start with your safety data
When it comes to workplace injuries, organizations are always looking for new ways to understand what’s causing them and how to stop them from occurring again. So, they record, report and store information on the incidents that happen, from informal one-on-one conversations to in depth root cause analyses.
But organizations don’t always make the most of the insights this data can provide. Often fearful that their data is not refined enough, complete enough or clean enough, they don’t take advantage of technologies like predictive analytics or artificial intelligence. Instead, they continue to rely on traditional reporting to deliver one dimensional information about incidents. After people have been hurt – or worse.
Getting your safety data ready
Data quality and consistency are two factors that repeatedly limit the use of an organization’s safety data. To achieve data quality and consistency, organizations must ensure they’re collecting data regularly and that it’s the right data.
Our work with energy, chemical and resource companies often identifies significant opportunities to improve the quality and consistency of data collection. This includes a more detailed collection of external variables, such as the ‘phase of work’ or ‘activity being performed’ through to variables related to the people who were injured, such as ‘trade/craft’, ‘time since started shift’, ‘how long in current role on current project’, ‘years of total experience in current role’, and various injury classifications.
If we don’t consider all the variables involved, we often miss the relationships and trends that cause incidents. For example, the length of someone doing certain activities can often influence the propensity for an incident to occur. But not all data has to come through completing an incident form – data related to a person’s tenure and job role can be taken from a human resources system.
Additionally, companies typically fail to collect data in a mutually exclusive, collectively exhaustive (MECE) way. That is, ensuring each variable has a purpose in driving insight, the purpose for each variable doesn’t overlap, and the fields in each variable represent an exhaustive list.
This isn’t saying that there will be hundreds of fields to complete in an incident form. But that there is an ideal spot where the fields aren’t too high level that they’re not actionable. Or, too detailed that they limit analysis and are too difficult for people to manage. To overcome this, standard MECE taxonomies and list boxes and checkboxes can help collect data consistently.
The more detailed and consistent your data, the more informed you are to influence safety decision making through specific interventions and safety programs.
Benefiting from predictive analytics
There may be tension around asking colleagues in the field to report more detailed data, particularly around near misses and lower severity incidents. But collecting safety data diligently and consistently across employees, contractors, joint venture partners and even customers is critical to unlocking the power of predictive analytics.
Improving enterprise safety data is a process. It isn’t possible overnight. But the quickest way of getting it done is to subject your data to predictive analytics. Applying more in depth analytical techniques will mean that opportunities for improvement in data quality and consistency will immediately appear, allowing you to fix them and improve your safety insights.
So, when is the right time to subject your data to predictive analytics? The answer is yesterday.
But if you haven’t done it yet, I’d recommend today.