Improve Without Breaking Things: A Pilot That Saved a Claims Team
By the Improve phase of DMAIC, a project team has usually done the hard analytical work. They have defined the problem, measured the process, and confirmed the real drivers in Analyze. The temptation now is to act fast, especially in early 2021, when a claims-processing team at a benefits administrator was under heavy backlog pressure and eager to show results. They had found their root cause and wanted to deploy the fix everywhere by Monday.
DMAIC, the core problem-solving sequence in Lean Six Sigma, runs Define, Measure, Analyze, Improve, Control. The Improve phase is where you generate, select, and test solutions. The word that matters most there is test. Improvement is a hypothesis until the process proves it, and rolling an untested change across an entire operation is how good intentions become new problems.
The scenario
The team's analysis showed that claims stalled at a manual verification step. Their proposed fix was to let processors approve low-value claims without that step. On paper it would cut cycle time dramatically. The team wanted to switch it on for everyone at once. The Black Belt coaching them insisted on a pilot first.
Pilot small and bounded. They ran the new approach with two processors on one claim category for two weeks, while the rest of the team continued as before, giving a built-in comparison.
Predict the result before you look. They wrote down the expected drop in cycle time and the error rate they would accept, so the pilot could not be rationalized after the fact.
Measure outcomes and side effects. They tracked not just speed but rework, customer complaints, and downstream audit flags, the places a shortcut tends to leak.
What the pilot revealed
Cycle time fell, as predicted. But the pilot also surfaced something the analysis had missed: a small but costly share of low-value claims were being used to test eligibility for much larger future claims, and skipping verification let errors through that surfaced expensively weeks later. Because the change was contained to two processors, the team caught it before it spread. They adjusted the rule to exclude those cases, re-piloted, and only then scaled up, carrying the new standard into the Control phase with a monitoring plan attached.
Pilot in a controlled slice before any full rollout, and keep an unchanged group to compare against.
Decide your success and failure criteria in advance, in writing.
Watch for second-order effects, not just the metric you set out to improve.
Only standardize and hand to Control once the pilot has earned it.
A pilot felt slow to a team under pressure. It was, in fact, the fastest path, because the alternative was an operation-wide error and a far longer cleanup.
If you are weighing a process change and want to test it safely before betting the whole operation on it, XNM's strategic advisory can help you design pilots that tell you the truth early.