What is a Root Cause, and How is it Identified?
This has probably never happened to you, but what if you created a solution that cost your organization a lot of money, and it didn’t have much effect? Or shaken your head when somebody proposed an expensive project that you felt wasn’t a big priority? These may seem like two different issues, but perhaps they both share something in common – not addressing the verified root cause.
What is a Root Cause?
A root cause is the most basic, system-related, underlying element that contributes to an undesirable condition (effect or problem), and when removed, the condition improves. To be a root cause, the cause and effect must be present at the same time, and when the suspected cause is removed, the effect either improves or disappears.
How are Root Causes Identified?
The best tool or technique to use depends on the situation. The process of root cause analysis begins with an undesirable event, problem, or opportunity. An initial review of the event may first uncover symptoms or potential causes, then the identification of most likely causes, and finally, validated root causes. There are many systematic approaches to facilitate this process, such as (in general order); brainstorming, affinity diagramming, Cause and Effect diagramming, “5 Whys,” Scatter Diagrams, Chi Square Test, Design of Experiments, etc. The goal is to separate symptom from root cause, and to ensure preconceived conclusions are not falsely justified. This calls for methods for understanding how facts relate to each other and to the problem. Effective root cause analysis requires a pragmatically applied set of statistical tools to separate hunches from facts, develop cost effective countermeasures, and prevent problem recurrence through standardization.
Following is a simple but effective approach for identifying root causes:
- First, quantify the issue, and stratify the collected data for the measurement of concern. This is often a Line Graph showing a trend over time, and a gap between actual and desired performance.
- Stratify the Line Graph data which requires understanding the process (flow chart, Gemba Walk, and 8 Wastes analysis) and knowing if your data is continuous or discrete. If the data is continuous, such as time, distance, or money, then consider using a Histogram first. The Histogram can often point you in the direction of the problem.
- If the data is discrete, or following the Histogram, an effective tool for identifying the most significant problem driving the general issue and gap is the Pareto Chart. Select the biggest bar on the Pareto Chart and set a target to reduce it.
- Select potential causes of the significant problem (biggest bar on the Pareto Chart) and test them to determine if there is a cause and effect relationship (see paragraph #2, above) with the big bar on the Pareto Chart. If so, develop and implement countermeasures.
When searching for root causes, it may be helpful to know that most root causes stem from four basic sources: (1) Procedures which are non-existent, out of date, or poorly written (2) People who have not been properly trained or monitored to use the procedures (3) Performance standards or targets set which are unachievable using the current procedures and given the workforce’s capacities and capabilities (4) Work environment in which people are put in a position of stress, burnout, and unreasonable workloads.
Although “problem solving” is documented as one of the most important employee skills in all industries, most organizations do not teach it to their workforces. Rather, employees are often expected to “just know how to improve their work”. Also, organizations often rely on abbreviated logic models and anecdotal techniques amounting to little more than “trial and error” approaches. Others may include additional techniques such as process maps, Cause and Effect Diagrams, surveys, and “5 Whys”. These alone are not enough because they are subjective tools and require no data to use. To get to the root cause, one must have data from the process producing the gap and know how to analyze it to separate the signal from the noise.