Almost every collection rate a practice owner is shown was built the same way: take the money that came in this month, divide it by the charges posted this month, call the result a collection rate. It is the default in nearly every practice-management report, and it does not measure collections.
The payments arriving in September are for claims billed in June, July and August. The charges posted in September have not been adjudicated by anyone yet. The two halves of that fraction describe different sets of claims, so the answer describes nothing. Run it in a slow month and you look like you are failing. Run it in a month when a backlog clears and you look excellent. Neither is true.
The only honest method is to take a defined set of claims and follow each one forward until it resolves. A closed cohort, not a snapshot.
What the cohort actually showed
In a pilot engagement we took every claim line billed to insurance over a defined window — 2,099 of them, $144,213 in total — and tracked each dollar to where it finally landed.
| Where the billed dollar went | Amount | Share |
|---|---|---|
| Contractually written off — never collectible | $48,376.10 | 33.5% |
| Paid by the insurer | $38,873.24 | 27.0% |
| Still open, awaiting payment | $45,428.07 | 31.5% |
| Moved to patient responsibility | $11,481.17 | 8.0% |
$48,376 — 33.5% — was the contractual write-down: the gap between the posted charge and what each payer had already agreed to allow. No amount of follow-up recovers a cent of it. Counting it as lost revenue is where six-figure “opportunities” come from.
Half of it is still in flight — and that is fine
After the contractual adjustment came out, insurers were on the hook for $95,837. Of that, $38,873 had been paid (40.6%) and $45,428 was still open (47.4%).
On a snapshot measured today, that 47% reads as a collections failure. It is not. The median age of the open balance was 59 days, and none of it was past 116. That is a pipeline that has not finished running, not money anyone lost.
The practical consequence matters more than the arithmetic. Staffing against that 47% would put the scarcest resource in a small practice — someone’s hours — onto claims that were going to pay on their own. The right move is to re-measure the same cohort in ninety days. If 47% is still 47%, there is a problem. If it has fallen, there never was one.
What happened when we ran it twice
The clearest way to see this is to measure the same window twice. In one pilot engagement we pulled the standard insurance-collections report on 31 July 2026, the day after the window closed, and then pulled the identical report five weeks later.
| Pull date | Owed | Collected | Gap | Rate |
|---|---|---|---|---|
| 31 July 2026 — day after the window closed | $47,720.48 | $31,087.03 | $16,633.45 | 65.1% |
| 2 September 2026 — 34 days later | $43,776.80 | $31,214.05 | $12,562.75 | 71.3% |
| Change | −$3,943.68 | +$127.02 | −$4,070.70 | +6.2 pts |
The other $3,944 was the software revising its own estimate of what it thought insurers owed. Nobody collected it, because there was never anything there to collect.
Medicare Part B makes the point beyond argument. It moved from $12,315.78 owed at 83.0% collected to $10,097.78 owed at 102.2% collected. A collection rate above 100% is impossible. What it means is that the original expectation had simply been set too high, and the system corrected it once the claims adjudicated. The entire “Medicare gap” on the first pull was an artefact.
Had the practice acted on that first report, most of the follow-up work would have chased money that was never missing. Staff hours are the scarcest resource in a small practice. Spending them on a reporting artefact is worse than spending them on nothing, because it also teaches the team that the reports lie.
The full write-up of this comparison, with the payer-by-payer detail, is published as a case study: The Second Pull → — Practice A, name withheld, June–July 2026.
When you measure decides the answer
The window matters as much as the method, and one example makes the point better than any argument.
This practice changed practice-management systems in May. That single cutover dumped $247,554 of artefact adjustments into one month — not economic events, just debris from the conversion. A separate migration code carried $11,723 and decayed from $10,747 in June to $80 by August, entirely on its own.
Any twelve-month analysis whose window happens to include May inherits a quarter of a million dollars of noise and reports it as performance. Nobody in that report is lying. The window is simply wrong.
1. Is the denominator net of contractual adjustments? If it is a gross rate being compared to a benchmark in the nineties, the finding is arithmetic, not analysis.
2. Is it a cohort or a snapshot? Period payments over period charges is not a collection rate, whatever the report calls it.
3. What is the median age of what is still open? Sixty days is a pipeline. Six hundred is a problem. A single percentage cannot tell them apart.
A number without a measurement date and a defined cohort behind it is not a number. It is a snapshot of a system mid-thought — and it is usually shown to you at the least informed moment it will ever exist.
Figures are one practice’s own unaudited system output, published from a no-charge pilot engagement with the practice’s written permission. The practice is not identified by name or location, and no patient-level information appears. The full analysis these figures come from is published as Profit Audit Findings, and the two-pull comparison as The Second Pull.