When we compile incident charts for a chemical packaging line in Osong, the first question from the plant director is never about the software used to draw the curves. It is whether the spike in week fourteen still carries the forklift aisle constraint that the night supervisor wrote in the margin.
Severity curves work when each data point retains a pointer back to the original log entry. We annotate charts with entry numbers, not employee names, so reviewers can open the source record during a briefing without flipping through unrelated files.
A common mistake is collapsing near-misses and recordable incidents into one line because the chart looks cleaner. Floor teams then distrust the graphic because it hides the reporting behaviour they worked hard to improve. We keep separate traces and show where classification changed mid-year.
For Korean plants aligning with Ministry of Employment and Labor reporting expectations, we mark statutory notification thresholds directly on the axis. Executives see when an internal severity label crosses into an externally reportable band.
The briefing packet ends with three questions tied to visible chart features: which zone contributed the longest flat segment, which corrective action correlates with the first downward slope, and which month still lacks complete near-miss entries. Those questions come from the logs, not from a template.