Big Data
Becoming a Practical Data Scientist
If you thought of becoming a data scientist, you probably took one of those fancy courses where they taught you how to train deep neural networks.
Your data makes a difference. when you review your historical incidents, you will see obvious opportunities to provide guidance and improve reporting.
These are real examples…..
Date of Incident: “Last week”
Name: “Big Willy”
Sex:” Sometimes”
Location: “Out back”
Cause: “Operator is f**king stupid”
We work with a lot of companies and they tell us about their funny but problematic issues with getting workers to write good health & safety reports. Getting your workers to report is always a challenge and these small but funny examples are not the real problem, the real problem is the compounding of subtle inaccuracies throughout the recorded data set that happens when workers don’t understand the value of these reports.
Because reporting got its start as record keeping and not data, workers see reporting as a chore, lower in value than sweeping the floor, and not a crucial set of data that could save their lives.
Traditional processes see them filling out the same report every day, 365 days a year. I’m sure you’ve heard many times how “useless” this is and how they think it just goes into the void. The result is your workers report inaccurately, inconsistently and incompletely.
It’s obvious that lack of guidelines yields flawed descriptions.
Your data can make a difference and the ‘magical’ first step is in improving what you get now. Go ahead and assess your collection standards. Ask yourself “What do I want to get out of this data and how can I make it happen?” You may not have a good idea but that is ok. Start with what you believe are the stresses the workers face and how you might find identifiers which are leading into risk.
If you haven’t done a reporting lookback, you should. I’m sure you look at every incident that comes into your organization but when you review your historical incidents and the meta-data associated with them, you will see obvious opportunities to provide guidance and improve reporting. Aim for a process of continual improvement – repeat your reviews quarterly or annually.
Rules for reporting:
- Keep reporting simple, and expectations simple.
- Ensure reports are broken into small, achievable chunks.
- Make sure you are not asking the same thing twice (or that is how it is perceived)
- Build reporting around what you need and what will help with analytics.
- Good controls on your software. Not just open fields. Discourage the “other” box.
How to help workers report:
- Have a users guide to reporting (Just a one page doc can really help). Give examples of good reports.
- Make reporting have pride – have supervisors review an example of a poor report (perhaps from an anonymous – different team). Also, discuss a quality report that shows best practices.
- One on one with workers is essential with simple, focused instructions.
- Broadcast clear examples of the value to your company, not just your quarterly metrics. Use emotion and humour.
Good adherence to standards doesn’t just help one report, it cascades across all of your data. Modern analytics is coming to you and the data you collect now is critical feedstock for this. Even though it may not seem that Machine learning and powerful Artificial Intelligence can help your company, it will in very short order. Get prepared, make sure you have the best data you can.
The lowest hanging fruit we see in safety is data standardization. We have been able to help industries significantly improve their quality of reporting by finding and creating thoughtful standards. It opens up many avenues for improvement. Your company can benefit greatly from pressing forward with developing standards and you might be surprised what comes out of it.