'Managing a quality service': example answers for the civil service behaviour

Our contact centre's quality-assurance scores dropped across two consecutive months — not dramatically, but our lead measure had been stable for a year, so the drift was signal. The tempting explanations were seasonal pressure or the new starters; neither survived contact with the data.

I pulled the failed assessments and read forty of them properly. The failures clustered on one criterion — accuracy of next-step advice — and traced to a policy change six weeks earlier whose guidance update was technically published but buried four links deep. Advisers weren't getting worse; they were confidently giving the old answer. That distinction mattered: a coaching response would have insulted thirty people and fixed nothing.

I had the guidance page rebuilt so the changed advice appeared on the screen advisers actually used, added the change to the daily team brief for a fortnight, and — because some callers had been given outdated advice — worked with my manager on a call-back list for the 60 or so affected cases rather than waiting for them to discover it themselves. Scores recovered to baseline within a month, and the lasting fix was procedural: policy changes now come with a named check that frontline screens reflect them before the change goes live.

Our planning-validation team hit a spring surge that pushed the backlog from five days to eighteen, and the pressure from above was blunt: get the number down. The fastest route was skimming the checks — and I'd watched a neighbouring authority try exactly that the year before, then spend the autumn re-validating half of what they'd waved through.

I split quality into what it actually was: three of our fourteen checks caught almost all the downstream failures — the ones that void an application late and restart the clock. I proposed a triage standard, in writing, to the service manager: full checks on the application types with history, the critical three on everything else during the surge, and a fortnightly sample of 20 fast-tracked applications re-checked in full so we would know, not hope, that the shortcut was safe. She signed it, which mattered — a quality standard lowered quietly is a standard abandoned.

The backlog came back to six days over five weeks. The fortnightly samples caught one recurring miss, which we added back into the fast track as a fourth check — the mechanism worked exactly as designed. When the surge passed we returned to full checks by the date in the original note, and the triage standard sat on the shelf, documented, for the next spring.

Our service's only route for resolving complex queries was a phone line, and our failure demographics told a quiet story: cases from deaf and hard-of-hearing customers, and from customers who needed an interpreter, were taking three times longer to resolve and escalating far more often — not because the cases were harder, but because the channel was wrong for them and each workaround was being improvised from scratch.

I proposed we stop improvising and build the alternative properly: a bookable appointment route — video with BSL interpretation, or a scheduled call with a language interpreter pre-arranged — reachable from the same starting point as the phone line, not hidden behind a special-needs page. The design principle I held onto was that adjusted routes must be first-class routes: same case owners, same standards, same follow-up, different channel.

We trialled it with the cases already stuck in the escalation queue. Resolution times for those customer groups came down close to the service average within a quarter, escalations from those groups roughly halved, and the unexpected benefit was staff-side: case owners stopped dreading those cases because the set-up work was now done by the booking system rather than by them, ad hoc, under time pressure. The route was made permanent at the next service review, with its own line in the performance report — which is what stops it quietly decaying.

Six months into an interpreting contract, complaint themes started repeating: interpreters arriving late to booked appointments, and occasional no-shows that collapsed the appointment entirely. The contract had service levels, but our side had been recording failures inconsistently — which meant every review meeting dissolved into anecdote against anecdote.

I fixed our half first: a simple failure log with four categories, completed by the case owner within an hour of any incident, so the next review ran on our data rather than our impressions. With three months of clean records, the pattern localised — most failures traced to one subcontracted region — and the conversation changed from 'your service is poor' to 'this region is failing; what's your plan?'. That specificity got a real response where generalised complaint had got apologies: the supplier changed their regional allocation model and offered affected appointments a priority rebooking slot at their cost.

Failure rates in that region came down to match the rest of the contract within two months, and the failure log became standard practice for our other contracts — because the lesson generalised: you cannot manage a supplier's quality with anecdotes, and the discipline usually has to start on your own side of the table.