The old adage goes: "A problem that is well stated is a problem half-solved" (Charles Kettering). The other half? It hinges on the ability of problem-solving teams or continuous improvement advocates to determine the factors that influence a well-defined problem statement. Filtering these factors is very challenging; a miss in the determination process could decide the outcome of the solution(s).
The process of determining root causes has taken varying forms from the early stages of the Industrial Revolution to the present. In the Six Sigma or Lean Six Sigma discipline, determining the significant factors usually starts in the Measure phase (the M in DMAIC). During this phase, uncovering the root causes means a rigorous walk-through of the process flow, deep dives on the identified factors, and characterizing the process behaviour with statistical tools. It is critical at this stage to exhaust all potential root causes affecting the focus area of an LSS project, and then to probe these "significant X's" in the Analyze phase (the A in DMAIC). A question LSS practitioners commonly ask is: how many significant X's should there be going into the Improve phase (the I in DMAIC)? Is there a magic number?
One project leader in a big FMCG manufacturing company proved that following the DMAIC process will lead you to the right root causes, and that the power of hypothesis testing helps in the sifting process. The project was about streamlining a process that had been established because of quality issues in the past. Six years earlier, the company's quality group had collaborated with the engineering team to set up an off-line bottle rinsing line to mitigate (if not eliminate) the presence of foreign particles in their bottles. Substantial time and effort had gone into designing, commissioning and installing the equipment, and then there was the labour involved in the whole process, plus operations management time. For years it was part of the end-to-end process, until, just recently, a Lean Six Sigma project leader challenged whether it was necessary. The LSS project team realized that the decision to set up the line was a response to quality issues, and that the root causes of the problem had never been fully exhausted, hence the "band-aid solution". The team collected a tremendous amount of data, applied statistical analyses to it (e.g. hypothesis testing), and concluded which of the root causes were indeed causing quality issues. To the team's surprise, out of the fourteen (14) identified root causes, only two (2) factors had a statistically significant effect on the problem at hand. At first, the LSS project lead was a bit doubtful about the results, saying it could not be possible that there were only two (2) significant X's. But nobody had tampered with the data, and the team had strictly followed the LSS process. The team's subsequent actions to address those root causes included redesigning the packaging, reconfiguring the orientation of raw materials during delivery, and working with vendors on complying with quality standards, among others. By the end of implementation, the significant X's had been addressed, which eventually reduced the need for off-line bottle rinsing. The project team also achieved 140% of the committed annualized hard savings.
Is there a real magic number for significant X's? There's none! There's no such thing as too few or too many significant X's… Following the LSS process will lead you to the "valid" root causes.