Process integration gaps at Caldera Health
Caldera Health, a health insurer with 4.5 million members, has run a SIAM model with six providers for 18 months. End-to-end SLAs exist on paper, but only two of the six underpinning contracts contain operational level agreements that map to them. Each provider runs its own priority matrix: the network provider defines P1 as 'total loss of a data centre', the application provider defines P1 as 'any member-facing transaction failure', and the service desk uses a five-level matrix of its own. In the last quarter 30% of P1 incidents were reopened. Problem management is performed inside each provider, with no cross-provider problem record, so recurring failures at the boundary are never owned. Around 220 changes a month go to a single change advisory board with 18 attendees and a four-hour meeting, and 11% of changes fail. There is no documented conflict-resolution process, so disagreements are settled by the provider with the loudest account director.
1. Prioritise the process integration gaps and explain the order in which you would address them.
Read the worked answer
Prioritise by customer impact, dependency and speed to fix, not by which process is easiest.
- Priority matrix and end-to-end incident ownership. Three different definitions of P1 mean the wrong incidents get the wrong response, and 30% reopened shows the handoff is failing. A single matrix agreed with all six providers is low-cost, fast and a prerequisite for meaningful reporting. Fix this first.
- Conflict resolution and escalation. Disagreements currently escalate by volume, which undermines every other process. A documented mechanism is a dependency for enforcing the priority matrix and for arbitrating reopening disputes.
- Operational level agreements back to back with end-to-end SLAs. Four providers have no OLAs, so the end-to-end SLA cannot be delivered or proven. This takes longer because it involves contract change, so start it early but expect it to land after the first two.
- Cross-provider problem management. Boundary failures need a single problem record with an owner and a route to a permanent fix. This depends on agreed priority and escalation, which is why it follows.
- Change management. Eleven per cent failure and an 18-person, four-hour CAB indicate a process designed for attendance rather than risk. It matters, but it is not causing the current customer-facing failures.
Sequence the first two within a quarter, OLAs within two, and problem and change redesign over six months.
2. Design the conflict-resolution and escalation process at a high level, naming the roles involved.
Read the worked answer
The process must be time-bound, evidence-based and documented in contracts, or it will revert to whoever escalates hardest.
Tiers. Tier 1: operational leads from the affected providers and the service integrator attempt resolution within an agreed window (for example 30 minutes for a P1, one working day for lower priority). Tier 2: the SIAM process owner for the relevant process arbitrates on process and evidence, with authority to decide the sequence of investigation and the single incident owner. Tier 3: the SIAM head or service governance board arbitrates cross-provider disputes that Tier 2 cannot settle. Tier 4: vendor management handles contractual consequences; this is deliberately separate so commercial leverage is not used to win a technical argument.
Rules. The dispute must be recorded with the evidence each party relies on. Timers run from the moment of disagreement, not from when someone notices. Providers must continue to deliver and cooperate while a dispute is open - no withdrawal of effort. Decisions and rationale are logged and reviewed monthly for patterns. The service integrator owns the register of open disputes and reports it to governance.
Embedding. Publish the process, include it in onboarding, reference it in every statement of work, and test it in a tabletop exercise with all six providers before relying on it in a live P1.
3. How would you redesign change management to reduce the 11% failure rate without losing control?
Read the worked answer
An 18-person, four-hour CAB is a symptom of approval by attendance. Control comes from categorisation, standardisation and evidence, not from meeting size.
- Introduce standard changes. Pre-authorise low-risk, repeatable changes with a documented procedure and a post-implementation review sample. This removes volume from the CAB and lets it focus on risk.
- Risk-based categorisation. Define change types by impact, reversibility and customer exposure, with matching approval routes: standard, normal (delegated to provider and SIAM change manager within thresholds), and significant (CAB or emergency CAB).
- Delegate approval. Give named roles authority within thresholds so decisions are not batched weekly.
- Share the calendar. A single forward schedule of change across all providers, visible to everyone, is the practical defence against conflicting changes at the boundary.
- Reduce attendance. The CAB should include decision-makers and technical advisers for the changes under review, not all 18 stakeholders; others receive decisions and the calendar.
- Manage failed changes. Analyse the 11% by provider, change type and root cause. Failure is often a testing or backout gap rather than an approval gap, and that requires different interventions.
Measure: proportion of changes handled as standard, CAB lead time, change success rate and unauthorized change count, reviewed monthly at governance.