KOGSYMethodThe model

How the documentation figure is calculated

Published 27 July 2026. Last updated 27 July 2026. This page matches the calculator's code exactly. If the two ever disagree, the code is what runs and the mistake is here, on this page. Tell us and we will correct it.

The whole model, in the open. It is arithmetic on figures you supply plus one assumption we name. There is nothing else in it.

01 · In words

The formula in words

Multiply your caregivers by shifts per caregiver per week by working weeks per year to get shifts per year. Multiply that by your average shift length to get billable hours, and by your bill rate to get annual billings.

Take your minutes per shift note, divide by sixty, and multiply by shifts per year. That is documentation hours. Multiply those hours by the hourly rate of whoever actually writes the note — the office rate if admin staff do it, the caregiver rate in either case where the caregiver does — and that is your documentation cost.

The modelled reduction is a range. The low end always assumes half the documentation time is removed. The high end uses the assumption you select on screen, which starts at 80%.

Exposure is annual billings multiplied by your estimated discrepancy rate. It is not a loss figure. See the limits section below.

02 · As code

The formula as code

shiftsPerYear     = caregivers × shiftsPerWeek × weeksPerYear
billableHours     = shiftsPerYear × shiftHours
annualBillings    = billableHours × billRate
marginPerHour     = billRate − caregiverRate
annualGrossMargin = billableHours × marginPerHour
docHours          = shiftsPerYear × (minutesPerNote ÷ 60)
docRate           = noteAuthor is office ? adminRate : caregiverRate
docCost           = docHours × docRate
savingsLow        = docCost × 0.50
savingsHigh       = docCost × reductionHigh
exposure          = annualBillings × discrepancyMidpoint
perPointBillings  = annualBillings × 0.01

Nothing is rounded at any intermediate step. Rounding happens only when a figure is displayed.

03 · Worked example

A worked example

An agency with 50 caregivers, each working 3 shifts a week averaging 12 hours, across 50 working weeks. Billing at 35 an hour, paying caregivers 24 and office staff 25. Fifteen minutes per shift note, written by office staff. Discrepancy estimated at 2 to 5 in 100 shifts. Assumption left at 80%.

Every figure here matches the calculator's own assertions.

Value	Result
Shifts per year	7,500
Billable hours per year	90,000
Annual billings	$3,150,000
Gross margin per hour	$11
Annual gross margin	$990,000
Documentation hours per year	1,875
Documentation cost per year	$46,875
Modelled reduction, low end (50%)	$23,438
Modelled reduction, high end (80%)	$37,500
Billing exposure at 2 to 5 in 100	$78,750

04 · The assumption

The one assumption

The low end is fixed at 50% regardless of what you select, which is why the range collapses to a single figure when you set the assumption to 50%.

We could not find a measured figure to use instead, from anyone. That search is now written up: are there any measured results from AI in non-medical home care? It publishes every query, the eight candidates that came closest, and our own commercial interest in the answer.

05 · The bands

The discrepancy bands

You pick a band; the calculator uses a single rate for it.

Band	Rate used
I do not know, we do not track this	none
Less than 1 in 100 shifts	0.5%
1 to 2 in 100	1.5%
2 to 5 in 100	2.5%
5 to 10 in 100	7.5%
More than 10 in 100	12%

One of these needs flagging rather than hiding. The 2 to 5 in 100 band uses 2.5%, which is the conservative end of its own range rather than its arithmetic midpoint of 3.5%. That understates the exposure figure by roughly a third for anyone in that band. We would rather the number be too small than too large, but the label and the rate do disagree, and this is under review.

Choosing "I do not know" produces no exposure figure at all. Instead the calculator shows what a single point of discrepancy is worth on your volume — annual billings multiplied by one per cent — so you get a sense of scale without guessing a rate you do not have. On the worked example above that is $31,500 for every 1 in 100 shifts.

06 · The limits

What this model deliberately does not do

The limits matter more than the arithmetic.

• It does not apply a recovery rate to the exposure figure. We could multiply it by some percentage and call the result recoverable revenue. We do not, because we have not measured what proportion is recoverable and neither has anyone else. Exposure runs both ways: under-billing costs you revenue, over-billing is a compliance problem.
• It does not model anything about your workforce. No claim about recruitment, hiring or how long people stay. There is no evidence for such a claim and we are not going to imply one.
• It does not measure anything. Every output is arithmetic on figures you typed in plus the one stated assumption. If you enter a wrong number you will get a confident wrong answer.
• It does not produce a cash figure when the caregiver writes notes unpaid. In that case the agency has no cash cost, and saying otherwise would be false. What it has instead is a dependency on unpaid work, and the calculator says exactly that.

07 · Try it

Try it on your own figures

The calculator takes about a minute. It asks for nothing before it shows you the result.

Open the calculator

08 · Questions

Questions we get asked