KOGSYWhat it costsThe office
How much office time goes to work nobody bills for?
Nearly all of it, in the sense that almost no office work is billable to a client. The useful question is the split: how much goes to scheduling, to chasing documentation, to billing, to hiring, to compliance. Most agencies have never measured it, and guessing is unreliable in a specific and well-documented way. The method is below.
Last updated 28 July 2026. No figure on this page is a sector average. The arithmetic is yours to run on your own numbers.
01 · Why
Why you probably do not know
Office payroll is one number. It buys scheduling, call-out cover, hiring, onboarding paperwork, client communication, caregiver support, billing, compliance, and everything nobody else picked up. None of that is separated anywhere, because there is no reason for your accounting system to separate it.
One care director described her day like this: "I have three phones. Sometimes the three phones are going off at once and somebody's walking in the door and I have two documents I'm working on and an email." That is not a scheduling problem or a documentation problem. It is an allocation problem, and you cannot fix an allocation you have never seen.
02 · The method
How to work it out for your agency
Do not ask people to estimate their split from memory. Recall is unreliable for exactly this, because the tasks that consume a day are the ones nobody registers as events.
Do not ask them to log every task either. Nobody sustains it, and the tasks that get logged are the ones with clean boundaries, which are not the ones eating the time.
Use sampling. Pick a random moment several times a day, record what the person was doing at that instant, and the proportion of observations in each category estimates the proportion of time.
share of time in category = observations in that category ÷ total responses
hours in category = share × total paid office hours
cost = hours × loaded office rateNine random checks per person per day over ten working days gives roughly 180 observations for two people, which supports an estimate within about six percentage points on an activity that turns out to be a fifth of the day. Fewer observations, wider range. Report the range, not a point.
03 · The figure
What a typical figure looks like
We do not have one and we have not found one published. This is measured so rarely in home care that we could not find a distribution to compare against.
The one pattern we would flag is that the categories people expect to dominate and the categories that actually dominate are frequently not the same, which is the entire argument for sampling rather than asking.
04 · The limits
What this is not
Sampling estimates how a day divides. It does not measure how long any individual task takes, it does not evaluate anyone's performance, and it should never be used that way. People behave differently when they know they are being sampled, and we do not correct for that because we do not know how large it is.
05 · Next
Next
The full protocol covers randomisation, non-response, consent and the limits, in enough detail to run it yourself or to challenge how we run it.
Not sure this is your biggest one? Rank all eight against your own numbers — fourteen questions, no score, and it asks for nothing before showing you the result.