Flat cost, falling odds
A claim that has not paid costs about the same to work in its fifth week as in its fifteenth. Someone opens it, reads the history, works out what the payer did and why, decides what to do, does it, and records the outcome. Almost none of that depends on the claim's age. Where age matters at all it works against you, because there is more history to read.
What does change with age is the chance that the work turns into money.
Two lines that move differently
Cost per touch is flat. It is set by how much reading the claim needs and how much judgement the decision needs, not by the calendar.
Expected recovery per touch falls. A payer that has already pended a claim four times is not more likely to pay it on the fifth attempt because time has passed. Some categories recover as well at ninety days as at forty; most do not. And the fall is not smooth — at the filing deadline it goes to zero and stays there.
Those two lines are the whole economics of follow-up. Everything else — worklist order, staffing, which claims get a second touch — is a consequence of the gap between them.
What that does to triage
The rule most operations end up with is: work the large balances first. It is not stupid. Balance is the one number always present, always accurate, and requiring no interpretation, and if cost per touch is constant then a larger balance returns more for the same spend.
But balance is a proxy for expected value, and a poor one, because it carries no information about probability. Expected value is roughly the probability of payment multiplied by the balance, less the cost of the touch. Sorting by balance sets the probability term to one for every claim. That is only correct if all claims are equally likely to pay, which nobody believes.
The probability term is not unknowable. Most of it sits in the claim's own history: which payer, what the last remittance said, whether the claim has been touched before, and whether the last touch produced any movement. A claim that has changed state since it was last worked is behaving differently from one that has sat still for three weeks, and it deserves a different priority.
The arithmetic you can do yourself
A claim worth forty dollars and one worth four hundred cost the same to work. Sort by balance and the forty-dollar claim is never reached — not by decision, but because the queue never gets that far. Sort by expected value and it is reached only when the cost of working it has fallen far enough to make reaching it worthwhile.
That gives you two levers, and only one of them is yours.
You cannot make a payer more likely to pay a stale claim. The probability term is set by the payer, by the denial reason, and by how long you took. It is largely fixed before the claim reaches a worklist.
You can change the cost of a touch. Every minute spent finding the claim, reading the remittance, hunting for the last note and working out what a reason code means for this payer is cost, and none of it is judgement. Cut it and the threshold moves: claims that were not worth working become worth working, without a single thing changing about the payer.
Where the flat cost is hiding
If cost per touch is the side you can move, it is worth knowing what it consists of. Ask someone to note the moment they stop gathering and start deciding. The gathering half is retrieval — it can be done ahead of time, done once for several claims at once, or not repeated at all if the last person wrote down what they found. The deciding half is the work you are actually paying for.
The same split applies across claims rather than within one. If a payer returns the same denial reason on forty claims, the first should cost you the analysis and the other thirty-nine should not.
What to measure
Cost per touch appears on no standard report, and neither does the thing it should be compared against. Two numbers are worth building:
- Touches per resolution, by payer and by denial reason. This is the closest thing you have to the cost side, and it identifies the categories that are expensive to work rather than merely numerous.
- Recovery per touch, in the same cut. This is the other line. Where it falls fastest is where the deadline is closest, and where a triage rule is quietly converting work into write-offs.
Neither is hard to produce if touches are recorded with their outcomes. If they are not, the flat cost stays invisible and the only variable left to manage is the balance — which is how a worklist ends up sorted by the one number that says least about whether the work will pay.
Our page on claim-status follow-up sets out the same arithmetic against the aging buckets it produces.