Governance
Most improvements look solid on the day you hand them over. The Long-Term Capability Review tool in the Sustain phase of RDMAICS re-checks Cpk and Ppk period after period so you find out early if the gain is slipping.
The Long-Term Capability Review tool does one job. It periodically re-assesses process capability over the long term to confirm gains have not eroded. You enter a review period, a characteristic, a Cpk, a Ppk and an optional sigma level, and the tool judges each row against the 1.33 threshold.
You reach for it after control charts and handover documents are done. Control tells you whether the process is behaving right now. Sustain asks a harder question. Is the process still as capable as it was when the project closed, measured again months later, by someone who was not in the room when the change was made.
The right rhythm is a fixed cadence. A quarter is common. The point is that the review happens whether or not anyone is worried. Erosion rarely announces itself. It shows up as a slow drift in the limiting value while everyone assumes the gain is banked.
Everything below uses the sample data built into the tool. The case is fictional. It is there so you can see the mechanics without needing your own numbers first.
The case covers two characteristics, cycle time and error rate, reviewed across two periods, Q3 2025 and Q4 2025. That gives four capability reviews in total. Cycle time and error rate are the two measures the original project moved, so they are the two measures the Sustain phase keeps watching.
Two characteristics and two periods is a small set. It is also a realistic one. Most teams track a handful of characteristics, not dozens. The value comes from repeating the same small set at the same cadence, not from breadth.
Each row needs a review period, a characteristic, a Cpk and a Ppk. Sigma level and assessment are there to add context. In the sample, the four rows read as follows.
Q3 2025 cycle time carries Cpk 1.41, Ppk 1.35 and sigma level 4.2, assessed as Sustained. Q3 2025 error rate carries Cpk 1.36, Ppk 1.34 and sigma level 4.0, assessed as Sustained. Q4 2025 cycle time carries Cpk 1.48, Ppk 1.42 and sigma level 4.4, assessed as Improving. Q4 2025 error rate carries Cpk 1.39, Ppk 1.35 and sigma level 4.1, assessed as Sustained.
Two habits make this table useful rather than decorative. Keep the characteristic names identical across periods, because the tool groups on them. And keep the period labels in a consistent format, because the trend reading depends on them lining up in order.
The tool judges each review on its limiting value, which is the lower of Cpk and Ppk. That choice matters. A strong Cpk with a weaker Ppk means the process behaves well within short windows but drifts across the longer one. Taking the lower number stops that drift being hidden.
For the sample data the output reads: 4 reviews, 4 of 4 capable at 1.33 or more, 0 marginal or not capable. The lowest limiting value is 1.34, from error rate in Q3 2025. The latest period is Q4 2025.
The tool also summarises movement in the limiting values from Q3 2025 to Q4 2025. Cycle time moves from 1.35 to 1.42. Error rate moves from 1.34 to 1.35. The chart plots Cpk and Ppk for each of the four reviews against a marked 1.33 threshold line, so you can see the gap between each bar and the line rather than only the pass or fail label. An AI review is available to explain the result in words.
In the sample, nothing is failing. Every one of the four reviews sits at or above 1.33. But a clean pass is not the same as a comfortable one, and the tool gives you the numbers to tell those apart.
Error rate is the characteristic to watch. Its limiting value was 1.34 in Q3 2025 and 1.35 in Q4 2025. That is above the threshold by a small margin in both periods, and the movement between periods is one hundredth. Cycle time, by contrast, moved from 1.35 to 1.42 and is labelled Improving. Same pass status, very different margin and very different direction.
So the slip signals are not only a number falling below 1.33. They are the pattern around the threshold.
The first decision is where to spend attention. With four of four capable, there is no fire to fight. The sensible move is to keep cycle time on routine monitoring and give error rate a closer look, because its limiting value of 1.34 and 1.35 leaves the least room.
The second decision is cadence. Characteristics that hold a wide margin can move to a longer review interval. Characteristics sitting near the line keep the shorter one. The tool gives you the evidence to make that call rather than reviewing everything at the same frequency out of habit.
The third decision is closure. A sustain plan cannot run forever. When a characteristic has held above threshold across repeated periods with a margin that is not shrinking, you have a defensible basis to hand it fully to the process owner. With only two periods in the sample, the data does not yet make that case. It makes the case for one more review.
The most frequent error is reading Cpk alone. In the sample, Q3 2025 cycle time has a Cpk of 1.41, which looks healthy. Its Ppk is 1.35. The limiting value is 1.35, and that is the number the tool uses. Quoting the higher figure in a report overstates the position.
The second is inconsistent labels. If a characteristic is called one thing in Q3 and something slightly different in Q4, the tool cannot pair them and the period comparison breaks.
The third is treating a pass as a finish line. Four of four capable is a good result for this quarter. It is not evidence of a permanent change. The value of the tool is in the next review, and the one after that.
The fourth is changing the measurement method between periods. If the way Ppk is calculated shifts, the movement from 1.34 to 1.35 means nothing. Freeze the method before you freeze the cadence.
It will not tell you why a number moved. The tool reports that error rate went from a limiting value of 1.34 to 1.35 and that cycle time went from 1.35 to 1.42. It has no view on whether that was a genuine process change, a staffing change, a seasonal effect or a measurement artefact. Finding the cause is your work.
It will not validate the inputs. Cpk and Ppk are calculated elsewhere, from your own data, with your own specification limits and subgroup choices. If those are wrong, the output is confidently wrong. The tool checks numbers against 1.33. It does not check how the numbers were produced.
It will not forecast. Two periods is a comparison, not a trend. Nothing in the sample output predicts Q1 2026. Treating a one hundredth movement in error rate as a direction would be reading more into the data than it holds.
And it will not replace a control chart or a conversation with the people running the process. It gives you a repeatable, dated record of whether capability held against a stated threshold. That record is useful because it is boring and regular. Everything else around it still needs judgement.
RDMAICS, A3, Just Do It and Kaizen projects with the full Lean Six Sigma toolset, AI interpretation and tollgate presentations built from your real evidence.
Savings Validation puts your reported benefit in front of finance and records what survives the check. This guide walks through a worked example where $120,000 of $121,000 reported is validated, and shows what a slipping number looks like.
Read the article
The Benefits Realisation Tracker compares what your business case promised against what the ledger actually shows. This guide walks through a worked accounts payable case and explains what an early slip looks like before it becomes a reversal.
Read the article
The Long-Term KPI Dashboard tracks a handful of measures well past project closure, each read against its own target and direction. This guide walks through a worked invoice processing example where all five KPIs hold, and shows the early signs that a result is slipping.
Read the article