Median caregiver turnover was 75% in 2024. At that rate, recruiting stops being a growth activity and becomes maintenance — you are refilling a tub faster than it drains, forever. Turn the dials and see it.
Hiring is an inflow. Quitting is an outflow. The number of caregivers you actually have is the stock in between — and a stock is set by both taps, not just the one you control. Most organisations optimise recruiting because it is visible and feels like progress. The stock is governed by the leak.
If replacing one caregiver costs in recruiting, onboarding and lost productivity, the gap is $27,000 a year at this size. Replacement cost varies widely by market and is not something we have measured — treat it as your number, not ours.
This is the oldest result in systems thinking. The first industrial model of its kind — built at MIT in the late 1950s — studied a plant with a three-year cycle of frantic hiring followed by mass quits. The finding was that the oscillation was produced by the plant's own policies. Not by bad workers, and not by bad managers.
That is the claim underneath a worker-owned cooperative, stated precisely: turnover is a property of the structure people work inside, not a defect in the people. Change what the job is — a real wage, W-2 employment, benefits, and an ownership stake in the business — and you are changing the structure that generates the leaving.
It also tells you the sequence. Recruiting is an inflow action; retention is an outflow property. Work the inflow first and the effort goes into refilling. The cooperative's bet is on the outflow loop — which is the honest way to say what it is betting on, and a claim that can be proven wrong.
The 24–36% figure is not our result. It is what established worker-owned home-care cooperatives report — the largest in the US has run under 24%, and it took that organisation decades. co-op.care is early and has no operating history to point at. The dial is set to 30% because that is a reasonable expectation from the literature, not a measurement of us. If we ever publish our own number, it will be labelled as ours.
Last reviewed July 2026. If a figure here goes stale, this page is wrong until it is updated.