The AR workdown operating system.

Executive summaryAged AR melts when it is run like a factory.

Most AR follow-up teams run a queue sorted by days in AR and dial from the top. The result is predictable: callers burn hours on balances that will never pay while collectible dollars age past their filing deadlines. The workdown operating system replaces the sort with a score, the flat queue with payer-tiered queues, the anecdote with caller math, and the monthly aging report with a weekly melt reading.

The frameInventory, not backlog.

A factory never asks a machinist to wander the floor looking for the oldest part. It sequences work by value and due date, stations it by skill, and measures throughput per hour. Aged AR deserves the same discipline, because it behaves the same way.

Call the aged book what it is: finished-goods inventory that has not shipped. Every account over 90 days is a unit sitting on the floor with a face value, a probability of ever converting to cash, and a hard expiry date called timely filing. Some units are live cash wearing a thin blocker, a missing medical record, an eligibility span that needs a corrected member ID, a claim the payer says it never received. Some units are dead stock, denied final on medical necessity after two appeals, or billed against a payer where the rendering provider's enrollment was inactive on the date of service. Treating both kinds identically, which is exactly what a days-in-AR sort does, is how teams post heroic call volumes and flat cash.

The operating system runs on five interlocking parts, and this paper walks each one: an inventory score that ranks every account by expected recoverable cash, payer-tiered work queues that route each account to the channel where that payer actually responds, caller productivity math that turns headcount into a predictable daily touch capacity, a status-code discipline where every payer response maps to exactly one next action with a date, and a weekly melt cadence that reads the output gauge and re-scores the floor. None of the parts is exotic. The compounding effect of running all five at once is the whole point.

The distribution above is what the ASP-RCM AR Workflow Engine observes per 100 accounts entering triage on the active book. The 16 upstream accounts matter most for caller economics: no volume of follow-up calls collects a credentialing denial, so the engine routes them out before a caller ever dials. Honest visibility into the 9 write-off candidates is equally part of the system. A workdown program that never recommends a write-off is hiding dead stock inside the ratio.

Part 1 · The scoreOne formula sets the queue position.

The score is deliberately simple enough to audit by hand: expected recoverable dollars, sharpened by filing urgency. Every account gets re-scored weekly, and the score alone decides what gets worked next.

Walk the arithmetic. Account A: $4,800 × 0.62 × 1.5 = 4,464. Account B: $12,500 × 0.08 × 1.0 = 1,000. Account C: $950 × 0.85 × 1.2 = 969. Account D: $6,200 × 0.45 × 0 = 0. The $4,800 account outranks the $12,500 account by more than four to one, because the score measures expected recoverable cash, not face value, and because its filing clock is nearly out. This is the arithmetic version of a rule ASP-RCM publishes on its AR Workflow Engine page: a 45-day balance with a known, fixable issue beats a 120-day balance that has been worked twice and denied for medical necessity. Oldest does not equal most collectible.

Two design notes keep the score honest. First, C is an estimate and should be treated like one: start with coarse bands (0.9 for confirmed payer error, 0.6 for documentation blockers, 0.3 for first-level denials with appeal rights, 0.1 for exhausted appeals) and let actual outcomes tighten the bands monthly. In the production engine this estimation is model-driven, scored from every prior call against every payer; a spreadsheet version with honest bands still beats a days-in-AR sort. Second, F is a multiplier and not a term in a sum, so an expired deadline zeroes the score no matter how large the balance. That is deliberate. A caller minute spent on an unappealable expired claim is inventory shrinkage disguised as effort.

Part 3 · The caller mathA shift is a throughput budget.

Factories know their units per hour. AR floors usually do not. The math below turns one specialist's day into a predictable touch budget, and the exhibit shows where the hours actually go.

Now the arithmetic. Industry-published AR follow-up touches sit at 12 to 14 per FTE-hour. Across the seven productive hours above, that is 84 to 98 touches per specialist per day. ASP-RCM specialists on the active book average 30 to 34 touches per FTE-hour, which is 210 to 238 touches per day, the published 2.4x lift. The lift is not callers dialing faster. It comes from three removals: prefetched account context (claim, EOB, last note, payer rule, denial-code translation loaded before the account opens), a predicted call outcome and recommended script surfaced before the dial, and queue sequencing that eliminates the half-minute search and the minute of scrolling that compound across an eight-hour shift.

The second half of caller math is conversion, because a touch that ends in "call back next week" is not throughput. On the active book, 71 percent of calls end in either resolution or a documented payer commitment with a callback date, against an industry-published first-call resolution baseline of 38 percent. Every commitment is calendared: a promise-to-pay or a promise-to-process gets a follow-up date, and a commitment that lapses re-enters the queue with its collectability estimate cut. Tracking the promise-kept rate per payer is one of the fastest ways to sharpen the C factor in the score, because payers keep or break commitments in stable, measurable patterns.

Part 4 · The next actionEvery status code has exactly one next move.

The factory analogue is the routing sheet: when a station finishes a unit, the sheet says where it goes next. In AR, the payer's status response is the station output, and ambiguity about the next move is where days quietly leak.

Two rows deserve emphasis. The in-process row is the discipline most floors lack: repeated status calls on a claim inside the payer's stated adjudication window consume touch budget and produce nothing, so the system forbids them by setting a recheck date and hiding the account until then. The credentialing row is the other silent killer. On the active book, 16 of every 100 worklist accounts are root-caused upstream, credentialing inactive, eligibility expired, or authorization missing, and the engine routes them out of follow-up entirely because the fix lives in the upstream queue, not in a phone call.

Part 5 · EscalationFour rungs from phone call to formal appeal.

When a status response says "denied" and the denial is fightable, the account leaves the call queue and climbs a ladder. Each rung has an entry condition, a work product, and a deadline discipline of its own.

The ladder is also a costing device. Each rung costs more specialist minutes than the one below it, so the priority score gates entry: a $180 balance does not climb to rung 3 no matter how winnable it is, and a patterned $60,000 underpayment across 40 accounts climbs to rung 4 as one case, not forty. Rung assignments follow the same senior-routing logic the engine uses in production, where medical-necessity appeals, underpayment recoveries, and second-attempt failures route to senior specialists automatically and the router learns from every hand-back.

Part 6 · The expiry clockTimely filing is the countdown on every unit.

Inventory on a factory floor does not evaporate. AR does. Every account carries a filing or appeal deadline, and once it passes, the balance converts from asset to argument. The F multiplier in the score is driven by this grid.

The grid earns its keep in the danger band. Every Monday re-score recomputes days-to-deadline for the whole book, and everything that crossed into the 30-day band is surfaced to the supervisor as a named list with dollars attached. The goal of the whole system, stated bluntly, is that timely-filing write-offs trend toward zero, because every such write-off is a dollar the team once had a legal right to collect and simply ran out of clock on. Appeal windows get the same treatment as filing windows: a denial that is fightable on rung 2 has its own countdown, and the ladder's SLA exists so that appeals do not die of age the way claims do.

Part 7 · The output gaugeThe melt waterfall: four exits, one shrinking bucket.

A bucket only shrinks through four exits: cash collected, appeal recoveries, rebills after an upstream fix, and documented write-offs. The waterfall walks the archetype's over-90 bucket from $2.40M to $1.08M over twelve weeks, exit by exit.

Check the arithmetic: $0.64M collected plus $0.28M in appeal recoveries plus $0.19M rebilled and paid after upstream fixes plus $0.21M in documented write-offs equals $1.32M out of the bucket, and $2.40M minus $1.32M leaves $1.08M, which is 9 percent of the $12M book. Note what the exits imply. Eighty-four percent of the melt ($1.11M) is recovered cash in some form; sixteen percent is adjudicated truth. Both are wins. The write-off slice is the one most teams get wrong in both directions: hiding dead stock inflates the AR asset and wastes caller hours, while casual write-offs without a documented dollar value and reason turn the aging ratio into fiction. Every figure here is an illustrative archetype, not a client result.

Part 8 · The cadenceThe week is the production cycle.

A factory runs on shift rhythms. The workdown runs on a weekly loop: re-score Monday, execute Tuesday through Thursday, read the melt Friday. The loop is short enough that a stall is caught in days, not in next month's aging report.

The Friday melt meeting reads three numbers in a fixed order. First, gross melt: how many dollars left the over-90 bucket and through which of the four exits. Second, net melt: gross melt minus the dollars that newly aged into the bucket this week, because a workdown that ignores inflow is bailing a boat with the drain open; a persistent inflow problem is a front-end problem and gets routed to denial prevention, not to more callers. Third, the forward forecast: what the re-scored inventory says next week should produce. Because the forecast comes from the inventory itself, a miss is diagnostic. If forecast melt was $120K and actual was $70K, the gap points at a specific queue, payer, or specialist, the same week it happened.

Part 9 · The gaugesThe scorecard the whole factory answers to.

Seven gauges, one page. The first three are the industry-standard output measures; the rest are the operational instruments that explain why the output moved. Definitions follow the HFMA MAP Keys so the numbers survive an auditor.

A note on reading the scorecard honestly. Net days in A/R and the aged-bucket shares can both be flattered by aggressive write-offs, which is why the scorecard pairs them with the timely-filing write-off line and the documented-adjudication discipline from the melt meeting. A team that hits 9 percent over-90 while its filing write-offs climb has not melted anything; it has renamed the loss. The MAP Keys exist precisely so that definitions cannot drift to fit the story, and the weekly melt meeting exists so the story cannot drift ahead of the definitions.

How ASP-RCM runs itThe operating system, with an engine inside.

Everything in this paper can be run on a spreadsheet and a disciplined floor, and a determined team should start that way tomorrow. ASP-RCM runs the same operating system with the AR Workflow Engine underneath, which is where the published multipliers come from.

The engine also supplies the dashboard the cadence runs on, in three live views: the specialist sees the active worklist and next-best account, the supervisor sees touches per hour and SLA breaches in real time, and the client director sees cash month-to-date versus forecast with A/R over 90 trending. Upstream, Credential OS flags accounts where the rendering provider's enrollment was inactive on the date of service and the authorization tracker flags exhausted auth balances, so both classes of dead-on-arrival follow-up leave the call queues before a caller touches them. The full capability set lives on the AR Workflow Engine page, and the pattern proof lives in our case studies.