By Jack Jager
The ability to demonstrate the success or failure of Root Cause Analysis (RCA) is a crucial stage of incident investigation that is often missed. After all, if you don’t measure, how do you (and others) know if your program is working and whether it’s worth the effort?
Measuring the success of an RCA program is important both for the short-term and the long-term. In the short term, you need to know if the changes implemented as a result of the RCA findings are effective. Longer term, you need the proof that your RCA program works – so that you gain ongoing support from management for this important tool.
Yet some companies fail to measure the success of their RCA programs. Here, we look at what’s stopping them, and how to overcome it.
First, why is measurement so important?
The goal of RCA is to improve processes or reduce the severity or impact of incidences. In RCA, you typically generate a raft of possible solutions and only implement some of them.
Implementing RCA solutions without measuring their effectiveness is akin to trial and error.
Measuring the changes caused by RCA solutions – good and bad – is critical to knowing whether your RCA efforts were successful. You need to know the “before” and “after” states of whatever the RCA is trying to improve, and assess how effective the solution is.
Measurement is probably more important for the “bad” results. You need to know if a solution isn’t working. A negative result will show you that the problem was not understood well enough (in which case you can go back to your cause and effect chart) or that poor choices were made in terms of which solutions to implement.
In this way, a “bad” measure still leads to a positive outcome. It allows positive decisions to be made to revisit the issue. After all, if the problem was significant enough to warrant an investigation in the first place, you need to know whether to revisit it or not.
What’s stopping you from measuring RCA success?
In some cases, the lack of measurement boils down to the fact that the RCA process is still relatively immature, and has not yet evolved into a complete process. In these instances, you need to deepen your commitment to grow your RCA program so that it captures this crucial step.
In other cases, unfortunately, measurement is simply shelved in the “too hard” basket.
Yet it doesn’t have to be hard. To measure the success of RCA, you simply need to set some parameters or criteria. Identify what you are trying to achieve – both the “big picture” goals and those relating to the RCA program itself.
What measures would indicate success of your RCA process?
The big picture will show:
- Improved availability of plant (less downtime) – either mobile plant or fixed plant (production infrastructure)
- Improved production data – weekly, monthly, per quarter, biannual and annual
- Less downtime when things go wrong
- Lower frequency of problem occurrence or similar types of problems
- Less impact of problems – problems are less severe or the ramification of these problems are less severe
- Less time spent reacting to problems and more time available for planning and making improvements
At a more local level, RCA program measures will show things like:
- Ratio of total number of incidents which should trigger an RCA against how many RCAs were performed
- Percentage of solutions generated against how many were implemented
- Percentage of people who have been trained in the process against those who have actually conducted RCAs or are using the process informally
- Indication of the timeframe needed to begin investigations (shorter is better)
- Indication of the timeframe required to implement solutions
By collecting information of this nature, you will be able to demonstrate how successful the RCA program is. In doing so, you will gain valuable support from management and co-workers.
Remember, you can easily tell someone a story of how good the RCA process is – but if you can’t show them the actual benefits in terms of production, availability or dollars, then the story counts for nothing. You have no evidence to prove it.
Instead, let the data from your measurement tell its own story.
