The system that isn't working yet
You lead a team of 12 processing housing benefit claims in a shared service centre. A new case-management system went live three weeks ago. Since then the weekly error rate has risen from 4% to 11% and throughput has fallen from about 900 claims a week to 620. Your manager wants a recommendation by Friday on whether to revert to the legacy system for a month. Reverting costs roughly £18,000 in contractor time and takes five working days, during which no claims are processed at all. The vendor says a fix is due 'within two weeks' but has missed two previous promised dates. Two of your most experienced processors are on leave until Monday. Legacy data is still being mirrored nightly, but nobody has checked whether the mirror is complete.
1. What do you need to know before Friday, and how do you find it out in the time available?
Read the worked answer
Only three facts can actually flip the recommendation, so I would spend the week on those and ignore the rest.
- Is the mirror usable? Nobody has tested it. A reconciliation of 30 recent claims takes one person half a day, and it is the difference between reverting being a real option and a fiction. Do it today.
- Are the 11% errors material? Split them into errors that misstate a payment or entitlement and errors that are cosmetic. Material errors drive cost, complaints and statutory risk; cosmetic ones don't.
- Is throughput trending up? Ask for week-by-week figures. A team climbing a learning curve looks nothing like a system that is failing.
I would not commission a full review, wait for the two experienced processors to return, or accept the vendor's verbal 'two weeks' - I'd ask for a date in writing and what happens if it slips.
The rule I'm applying: gather only information that could change the decision, and take a decision at 70% information now over a perfect one in three weeks.
2. What do you recommend, and under what conditions would you change your mind?
Read the worked answer
Recommendation: keep running on the new system for one more week, on hard conditions - with reversion pre-agreed if they fail.
The arithmetic decides it. Five days of reversion with nothing processed loses roughly 620 claims at the current rate, which is about two weeks of the shortfall you are already carrying, plus £18,000 and a second migration. Reversion only pays if it buys you more than about three weeks of clean running. That makes it the right answer for a broken system and the wrong answer for a steep learning curve - so the decision should hang on which of those we have, not on how anyone feels on Friday.
Conditions, checked Friday: the mirror reconciles; the errors are mostly cosmetic; the vendor gives a written fix date inside two weeks. Any failure, and we revert on Monday.
I would also ask whether a partial revert is possible - one team back on legacy, the rest forward. That caps the cost, keeps the mirror honest, and spreads the risk.
Say the trigger out loud: 'if the error rate is still above 9% on Friday, we revert.' A decision that can't be tested afterwards isn't a decision.
3. Your manager says: 'Just give me the answer, not the caveats.' How do you respond?
Read the worked answer
'No caveats' is a fair request from a decision-maker; 'no conditions' is not, and I would say so plainly rather than either hedging or over-promising.
What I'd hand over is one page: the recommendation, the cost of it, the main risk, and the two trigger conditions. Then one sentence in person: 'Here's the call. It holds unless the mirror fails on Friday, in which case I revert and tell you the same day.' That is a decision, not a hedge - it tells them what will happen and who is accountable for it.
I'd also record in writing what was known on Friday and what wasn't. If this goes wrong, the question asked afterwards will be what we knew and when, not what we felt. Documenting the basis of a decision is not covering your back; it's the difference between a professional judgement and a guess.