The ML4 Claim at Northbridge Assurance
Northbridge Assurance employs 900 people in claims handling across nine teams. Two years into a Kanban adoption, every team has a board, a standup, and a cycle-time chart. The CIO has read about the Kanban Maturity Model and wants the organization assessed at ML4 by year end, because the board has asked for 'quantified delivery capability'.
The evidence is mixed. Team-level flow metrics exist, but 40% of claims spend more than five days in a 'pending information' queue that no team owns. There are no agreed service-level expectations with underwriting or the broker channel. Risk review happens quarterly at portfolio level and is disconnected from the boards. Each team reports on-time performance against its own definition; aggregated, the numbers look 85% green, which nobody believes. Retrospectives produce team-local actions and rarely reach management. Cross-team dependencies are tracked in a spreadsheet maintained by a PMO analyst.
The CIO has asked you, as coach, to 'get us to ML4 in two quarters'. Your assessment interviews took eight days and are now complete.
1. What maturity level does the evidence most credibly support, and which specific observations would you present to challenge the ML4 target? Show how you reason, not just the label.
Read the worked answer
KMM maturity is about the depth and consistency of practices and their fitness criteria, not the presence of artifacts. The evidence supports ML2 in pockets with fragments of ML3, and does not support ML4.
Reasoning: at ML2 the fitness criteria are team-level explicit policies, WIP limits, flow metrics, feedback loops, and a service delivery approach within the team's scope. Northbridge has the artifacts but not the policies - on-time has nine definitions, so the 85% figure is a local metric aggregated into a false enterprise one. ML3 requires the organization to be fit for purpose as an enterprise: cross-team dependencies managed by policy, service-level expectations negotiated with customers, risk review connected to delivery, and feedback loops that reach decision-makers. The 40% pending-information queue with no owner, the PMO-owned dependency spreadsheet, and the disconnected quarterly risk review all fail that test. ML4 requires quantitative management of the value stream - reliable forecasting, explicit classes of service, and business outcome linkage - which cannot exist on top of undefined policies.
Present this as a diagnostic hypothesis, not a verdict: show the definition variance, the queue aging, and the dependency data; ask the CIO which customer-facing service level the organization can currently state in one sentence. The answer is the gap. Recommend two cross-team queues be given owners and explicit policies before any maturity label is discussed.
2. The CIO wants Monte Carlo forecasting across all nine teams within six weeks to satisfy the board. How do you respond, and what do you propose instead?
Read the worked answer
Start by agreeing with the intent - the board needs credible forecasts - and disagreeing with the sequence. Monte Carlo simulation is only as good as the historical data and the definition of start and finish. With 40% of items sitting in an unowned queue and nine different on-time definitions, a simulation across all nine teams would produce confident-looking numbers that cannot be defended when the board asks why reality diverged. That is a reputational risk for the CIO, not just a technical one.
Propose a six-week deliverable: one value stream, one explicit definition of start and finish, 8–12 weeks of historical data, and a forecast expressed as a range with the assumptions written down. Include the pending-information queue in the model, because it is the dominant delay. Present the result alongside the data-quality caveats. Then use the forecast's wide range as the argument for policy work: queue ownership, classes of service, and agreed service-level expectations.
Also name the pattern: installing a higher-level practice before the lower-level fitness criteria exist tends to produce a practice that decays under pressure, and the organization concludes that forecasting does not work. Offer the CIO a joint review at week three where she decides whether the data is good enough to extend to a second value stream. This keeps the relationship and the evidence honest.
3. The 85% green reporting is politically sensitive. How do you raise it without becoming the maturity police, and what do you ask the CIO to own?
Read the worked answer
Raise it as a measurement question, not an accusation. In a one-to-one with the CIO, show the definition sheet: nine teams, nine definitions of on-time, some measuring from commitment, some from start, some excluding waiting time. Then ask a single question: what does the board believe the 85% means? The purpose is not to embarrass anyone; it is to show that the number cannot support a claim of quantified management.
Ask the CIO to own three things. First, convening a short service-level workshop with underwriting and the broker channel to define one customer-facing expectation. Second, sponsoring the assignment of an owner to the pending-information queue. Third, protecting the teams while the definitions are renegotiated - teams should not be penalized for a metric they did not design.
The coaching stance matters. I would not declare a maturity level publicly or use the assessment as a scorecard against managers. I would share the evidence with the CIO, let her decide how to communicate it, and offer to co-facilitate the service-level conversation. If she wants the ML4 label regardless of the evidence, I would say plainly that I cannot support that claim and document the advice. The KMM is a diagnostic and a change roadmap, not a certification trophy.