Now that we have a shorter list of potential causes (factors), ranked by likelihood of impact, we need to determine which is the root cause. Each factor can be classified into one of four groups:
- Correlated Result. These are factors which are other symptoms of the same root cause. For example, if revenue is down and our sales tax collection is down, the reduction in sales tax is not a cause of the reduced revenue, but another side effect of the root cause (whatever drove down revenue).
- Unrelated Factor. These are factors that look suspicious but are, in reality, unrelated to the change in question.
- Contributing Factor. These are factors that, while part of the chain of events that caused the chain, are not the root cause. For example, if revenue is down and total number of purchases are down, the reduction in total number of purchases is likely a cause of the drop in revenue but not the answer to why purchases dropped themselves.
- Root Cause. This is the factor that initiated the chain of events that resulted in the change. Remember, there may be more than one!
Our first step is to arrange all of the high likelihood factors into a timeline. Establishing the order that factors occur on your timeline is not always as easy as checking the time of day when they happened, sometimes you’ll need to rely on your knowledge of the business and your internal processes.
If you remember our example of Sean’s Snowboards, it was clear that Store #456 was at the core of our drop in revenue on January 21st. Below is the timeline of factors we identified which are related to Store #456.
Organizing the factors by time makes it easier to classify the factors as they start to tell us a story.
- January 10th: Competition lowers prices at store nearby. This is almost two weeks before the drop and unlikely to be the root cause as we would have seen revenue changes much earlier. Verdict: Unrelated.
- January 18th: Assistant manager resigns. This is likely a result of the new manager hiring on January 17th. At best that makes it a Contributing Factor but it’s more likely another symptom of whatever went wrong. Verdict: Correlated Factor.
- January 20th: Parking lot construction starts. This is a likely culprit, as customers may not have been able to get to the store if they could not park. However, construction like this is the result of another decision as construction crews do not show up on their own. That makes this part of the chain of events that led to the change, but not the root cause. Verdict: Contributing Factor.
- January 17th: New manager hired. This is our most likely culprit, as it happened soon before the drop and the new manager would have had to approve the start of construction on the parking lot. Verdict: Root Cause.
This is, obviously, a simplistic example but I hope it gives you a sense of how the process of recreating the timeline and classifying the factors would work for you.
As you probably have noticed, root cause analysis is a lot like detective work. You start with some evidence, eliminate possible suspects and hopefully reconstruct the timeline of the event. Just like with detective work, there are things you can start doing today that will help you be better at root cause analysis in the future. We’ll review some of them tomorrow.