A recovery dashboard can show how many cases remain without explaining what would move them forward. Some cases need a correction, some need information from another organisation, and others need a business decision. Treating them as one queue makes it difficult to understand progress or agree what the delivery team should change next.

The Cabinet Office’s 16 November 2026 Civil Service pensions recovery update illustrates this distinction. It describes cases where resolving a root cause still leaves a dependency on employer or member information. It also explains that related cases are being consolidated around a primary case. Those are different kinds of movement, and a single closure count cannot explain both.

For a pensions change team, the practical question is: can each important case state be connected to a clear next action, an owner and evidence that the action worked? The following approach is a suggested analysis method for that discussion, rather than a description of the recovery programme.

Agree what the queue is counting

Start with the unit of work. Is a row a member, an enquiry, a calculation, a document request or an administrative task? One member may have several related enquiries. One enquiry may depend on several tasks. Until those relationships are explicit, totals can invite competing interpretations.

Write a short definition of the primary case and its linked work. Describe how duplicates, reopened cases and transferred items should appear. Retain the connection to the original records so an administrator can reconstruct what happened. A consolidation should leave a visible explanation of where the outstanding work now sits.

Make waiting states useful

“Awaiting information” is a starting point. Add the particular information needed, who can provide it, who owns the request and what happens when it arrives. Distinguish an unanswered request from information received but awaiting validation. They need different actions.

Consider a fictional case where an administrator needs an employment-history clarification before progressing a quotation. The useful requirement is more precise than adding a waiting status: an authorised user should be able to see the outstanding question, its request history, the responsible team and the next review date. The team should agree what evidence allows the case to leave that state.

This example makes no assumption about a particular scheme’s calculation rules or service commitments. Those decisions belong with the relevant business owners.

Turn recurring blockers into requirements

Review a small, agreed sample of stalled cases with operations, data and delivery colleagues. Group them by the next decision or action required. Keep uncertain classifications visible rather than forcing every case into a convenient category.

A repeated blocker might point to unclear evidence requirements, an ambiguous handoff or a missing validation step. Record the observed problem before proposing a system change. Sometimes the useful intervention will be a clearer request or an agreed decision route; sometimes a workflow change will be justified.

For each proposed change, connect four things: the observed problem, the required behaviour, the owner who can settle its meaning and the acceptance example. That gives developers and testers a shared basis for discussion and helps operations challenge a requirement that would leave the original problem unresolved.

Test the transitions between states

Acceptance criteria should explain how the work progresses. For the fictional information request, test what happens when a complete response arrives, when the response conflicts with existing information and when only part of the request is answered.

Include linked-case behaviour. If two enquiries are consolidated, can an authorised administrator find the surviving case and its remaining actions? Can a reviewer distinguish an administrative closure from completion of the member’s underlying request? Agree the expected result before testing.

Capture the relevant business decision and use approved test data. Where an interface or external response is simulated, state that limitation in the acceptance evidence. The reviewer needs to understand which part of the process has actually been demonstrated.

Report movement and unresolved dependencies together

Build the recovery view around questions that help owners intervene: what arrived, what progressed, what completed and what remains blocked? Agree how changes in classification or consolidation affect the totals so reviewers can reconcile one reporting period with the next.

Show waiting categories alongside age and ownership. A case that has progressed internally may still need an external response. Recording that progress can be useful, provided the outstanding dependency remains visible and the member outcome is described accurately.

Avoid adopting a new measure merely because it is easy to extract. Ask which decision it supports, who will act on it and what supporting detail they would need to investigate an unexpected movement.

Begin with one decision path

Choose one recurring blocker and map its route from identification to resolution. Agree the states, ownership, required information and acceptance examples with the people who do and accept the work.

The useful output is a small, reviewable chain of evidence: this is the obstacle, this behaviour should address it, this test demonstrates that behaviour, and this person can decide whether it is sufficient. That gives a recovery team something concrete to improve while keeping the wider backlog and remaining dependencies in view.

For related evidence questions, read Pensions data quality: what evidence closes a defect? and Data migration readiness: evidence before optimism.

← Back to all insights