A rota problem you can feel before you can see it
Most care providers don't have a retention problem they can see. They have a rota problem they can feel: the same shifts going hard to fill, more short-notice gaps, more unfamiliar cover in front of residents, managers spending evenings on the phone. By the time the resignation letters confirm it, the damage to continuity is months old.
This group was losing people faster than the sector average, 36 leavers in six months across six homes, and nearly a quarter of those were gone within their first 90 days. Replacing each leaver costs an estimated £3,000 to £6,000 (Care England), and that direct figure is only the visible cost. The hidden cost lands next: manager hours lost to recruitment, unfamiliar agency cover, and residents meeting another new face while continuity breaks and care quality dips. Exit conversations, where they happened at all, came too late to change anyone's mind.
What the signals showed
In month one, Baton baselined the workforce: tenure, training status, hours patterns, short-notice gap frequency, and regular sentiment capture from carers, surfaced to managers rather than buried in a spreadsheet. The picture sharpened quickly:
- 47 training lapses outstanding or due within eight weeks, concentrated in two homes: carers quietly becoming unable to work the shifts they wanted.
- Short-notice cover clustering around the same units and shift patterns, pressure points rather than randomness.
- Sentiment flags and risk indicators pointing at a group of 23 carers whose patterns, hours volatility, missed shifts, low sentiment, matched people on the way out.
- New starters leaving early: nine joiners gone inside their first 90 days, one of the sharpest leaks.
None of this required anyone to announce they were leaving. The carers most likely to go rarely say so out loud.
Acting on the signals
Acted on every flag. Each at-risk indicator prompted a structured conversation. Baton carries built-in best practice for one-to-ones and exit reviews, so managers had something better than instinct. Of 23 carers flagged, 17 were still in post at month six.
Closed the training gap. Expiry alerts turned the 47 outstanding lapses into a worked queue: placement-led training delivered directly to carers, proactive recertification before anyone dropped out of shifts. By month four, lapses were down to under five, none safety-critical.
Backed the people stepping up. Development routes and progression conversations for carers ready for more, the single strongest retention lever in the evidence base, plus recognition that gave experienced carers a visible reason to stay.
Settled the new starters. New joiners got structured early-tenure support, with Baton's assistance helping them settle in more effectively. Early-tenure exits halved, from nine to four.
Supported what only the provider could change. Pay levels and team culture belong to the provider. We Are Care benchmarked pay against the local market and surfaced the conflict signals; the provider acted. The countermeasures worked because both sides did their part, worth being honest about.
What changed
By month six, turnover was trending at 21% annualised against 34% at baseline. That is 14 fewer leavers than the previous six months, roughly £42,000 to £84,000 in avoided replacement costs at £3,000–6,000 per leaver, before counting steadier rotas and fewer unfamiliar faces. Short-notice cover requests fell by 38%. Familiar-cover share, the shifts worked by carers the residents already knew, rose from 61% to 84%.
What the provider watches now is the flag count: how many carers are showing risk patterns this month, and what is being done about each one.
Why it worked
Signals before resignations
The carers most likely to leave rarely say so. Capturing sentiment and risk patterns continuously means the conversation happens while it can still change the outcome.
People make every call
Baton flags and suggests; managers and We Are Care coordinators decide and act. No carer was ever managed by an algorithm.
Your data stays yours
Workforce data this sensitive only works under real confidentiality, which is why mutual NDAs are standard on every Baton partnership, and why this case study carries no names.