Chargeback processing, explained.

A chargeback in plain words: the distributor paid list price, the hospital paid less, someone owes the gap

One box of guidewires, from signed contract to credit memo

1. The maker and the buying group signCedarline Purchasing Group agrees contract prices with Halvern Medical, by tier, for a set term. Halvern tells its six distributors.
2. The distributor stocks at list priceSummit Ridge Distribution buys guidewires from Halvern at $410 a box and holds them in its warehouse.
3. The hospital buys at contract priceGlenfarrow Medical Center orders 60 boxes and pays $318 each, as its contract says. Summit Ridge is now $92 a box short.
4. The claim is filed and checkedSummit Ridge claims $5,520 on an electronic file, with the invoice behind it. This is where a sale two days after the contract ended is caught, or paid.
5. The maker answers and creditsLine by line, the maker accepts or rejects with a reason code, issues a credit memo for what it accepts, and corrects its revenue estimate.

Step 4 is the subject of this guide. The people, companies and figures in this example are fictional.

Four things about chargebacks that surprise newcomers

Four terms on every chargeback: WAC, contract price, roster, credit memo

Fourteen terms are defined in the one-page summary.

Who pays whom in a chargeback: WAC, contract price and the gap the maker refills

In shortThe distributor buys at list price, sells at contract price and claims the difference from the maker. Four electronic messages, in a standard format called EDI (electronic data interchange), carry the arrangement: the contract price going out, the claim coming back, the answer, and sales tracing. The distributor's margin depends on being paid back, so the two sides are partners with opposite interests on every doubtful line.

The arithmetic on one box of guidewires

The chargeback is the list price minus the contract price, times the quantity. Under the published 2012 policy we read, the list price used is the one in force on the date of sale.3 In the example below the distributor is whole again after the credit, and the maker's net revenue on the box falls from $410 to $318.

Why distributors and makers both care about speed

A distributor is out of pocket from the day it sells at contract price until the credit arrives, and a rejected line ties up that cash longer.1 A maker that slows down can pay for it: the published 2012 policy we read let the wholesaler deduct chargebacks it had not been paid or credited on time from what it owed the maker.3

Distributors also carry weight by size. Three companies dominate drug distribution: Drug Channels Institute says they "distribute more than 96% of prescription pharmaceuticals in the U.S.", and projects their drug distribution revenues at $933 billion in 2026.11 HIDA's members, the medical product distributors, operate more than 500 distribution centers and serve more than 560,000 care settings.12

Chargebacks, rebates, GPO fees, 340B and Medicare refunds: five payments that are easy to confuse

In shortA chargeback goes to a distributor, on a sale it made at a contract price. A rebate goes to a payer or a buyer after the sale, on contract terms. A GPO administrative fee goes to the buying group. A 340B price is a federal discount that often reaches a hospital through the same distributor claim. A Medicare negotiated price refund goes to the pharmacy or other entity that dispensed the drug. They share contracts and data, and they reach different people.

The word also names a card payment dispute, and search results mix the two. Everything in this guide is about distributors and makers.

How GPO contract pricing and tiers shape a chargeback

GPOs negotiate for many hospitals at once, so a single product has many prices. A 2009 study quoted by the Government Accountability Office (GAO) put GPO contracts at about 73% of hospitals' nonlabor purchases, and the GPO trade association estimated that about 98% of US hospitals use GPOs, two to four each.17 The figures are old, and the 98% comes from the GPOs' own trade group. They are the best public ones we found.

GAO also reported the fees vendors paid GPOs in 2008: 1.22% to 2.25% on average, weighted by volume, with single fees from 0.09% to 10%.17

Within a contract, a hospital often holds a tier: a price level tied to how much it commits to buy. Tiers move. HIDA's recommended practices describe a hospital system that asked for tier four and told its distributor to use that price before the maker had approved the change. The distributor's claims were denied, and the guidance is that the maker's approval is the source of truth.18

No federal rule processes a chargeback: what governs the claim, and what changed by 2026

In shortYour contracts and your distributors' policies set the clocks and the reasons. Trade standards set the file formats. Federal rules reach the edges: drug price reporting, anti-kickback safe harbors, a ban on reselling hospital-bought drugs, 340B and financial reporting. Two 2026 changes matter: a revised claim file standard, and a contested 340B rebate pilot that would move some discounts off this path.

The 340B rebate pilot and what it could do to chargeback volume

The 340B program lets safety-net hospitals and clinics, called covered entities, buy outpatient drugs at steep discounts. HRSA, the agency that runs it, says covered entities reached about $100 billion of purchases at discounted prices in 2025.15

Today the discount comes up front, and HRSA's notice records commenters' view that this model has run for more than 30 years. Under the revised pilot, covered entities would order an approved maker's selected drugs at list price through 340B wholesaler accounts and claim the discount later as a rebate: list price less the ceiling price on the day of dispense.15

The pilot is small: drugs on the Medicare negotiation lists for 2026 and 2027, under 5.5% of total 340B sales on 2025 data, with plans effective January 1, 2027.15

A federal court in Maine blocked the first version, set for January 1, 2026, with an injunction on December 29, 2025, and vacated (canceled) it on February 10, 2026.15,31 Hospitals sued in the same court on October 9, 2026, and a House bill, H.R. 10770, would delay it through December 31, 2028.24,25 We found no ruling or vote as of November 3, 2026, and we did not read the docket.

For a chargeback team the practical point is narrower: HDA's 2026 revision of the 844 adds bill-to and sold-to entities beside the ship-to, so a claim can name both the covered entity and the contract pharmacy, the outside pharmacy that dispenses for it.8 If your systems read only the ship-to, plan for that change now.

The claim form, invoice, purchase order and delivery note behind one EDI 844 line

In shortAn electronic line tells you what the distributor says happened. The paper behind it, when a maker asks for it, shows what happened: the invoice, the hospital's purchase order and the delivery note. Four of the seven common errors need something the electronic line does not carry, so a line-matching check cannot see them.

Why a HIN or GLN mismatch can fail a chargeback line

Matching depends on naming the customer and the product the same way on both sides. The Health Industry Number (HIN) is a nine-character identifier for health care locations and departments, used across more than 3.5 million locations and services.32 A Global Location Number (GLN) does a similar job.

HIDA's guidance is to identify every owned and affiliated location, and it describes denials that came from a system asking for prices across affiliates that the maker covered only when owned, managed or controlled.18 A contract applied to the wrong entity of the same hospital system will fail even when nobody made a typing error.

Six questions a clean chargeback line must answer

Five chargeback calculations, worked: contract price, tier, unit of measure, running total, clean lines

In shortThe arithmetic is simple and the inputs are the difficulty. Each sum below uses the fictional example's numbers. Small changes in an input, such as a tier or a unit, move the claim a little or a lot, and the running total against a purchase order is the sum that manual teams rarely do.

How list price changes and bracket prices alter a chargeback

Two more inputs trip people up. When a list price changes, the chargeback depends on which list price applies, and the published 2012 policy we read used the WAC at the date of sale.3 The same policy let the wholesaler claim at the "highest bracket price" where a maker sells at several list price levels, called brackets, regardless of what it actually paid.3 A team that assumes every chargeback equals the distributor's actual cost minus the contract price will mark good claims as errors. Read your agreement first.

Why chargebacks get rejected: eligibility, pricing, dates, then the errors nobody counts

In shortThe one public survey of causes we found puts eligibility first, then pricing, then dates. Class of trade sits inside eligibility and has no shared definition. Duplicates, over-quantity claims and altered documents have no public count at all, because they surface only where someone keeps history and reads the paper.

The best public evidence is a Rutgers University study with the HDA Research Foundation, published in 2018. It surveyed 147 companies and heard from 32, including makers, distributors and GPOs, and interviewed 10 experts. Its focus was disruption after mergers and system conversions, so its numbers may describe stressed moments.1 Respondents ranked eligibility, pricing and date issues as the most common root causes of rejected chargebacks.1

The seven error patterns a chargeback review looks for

An expired contract is a business decision before it is an error

HIDA's recommended practice leans toward continuity: makers should tell distributors of extensions 45 days before expiry, and if the date passes with no new terms, the default should be to courtesy extend the existing agreement.18 A contract can say the opposite and require a written extension. Both happen. If your company grants courtesy extensions in practice, write them into the register that your check reads. Otherwise you reject sales your own account team promised to honor, and the distributor is left holding the difference.

What a chargeback mistake puts at risk: cash, reserves, Best Price and the audit opinion

In shortPay too much and the margin is gone, usually for good. Reject too much and a partner's cash stays tied up and rework grows. Estimate the reserve badly and reported revenue is wrong: public filings show restatements (corrected past results) and material weaknesses (control gaps that could let a serious error through) that started here. For drugs, a chargeback also feeds the prices reported to Medicaid.

The clock on a claim under one wholesaler's policy

Under that policy the maker also gave the wholesaler a chargeback advance to cover its credit exposure and carrying costs, set each quarter at no less than one month of billings. Audits covered only the 12 months before the audit began, with results not extrapolated to other periods.3 Read together, the terms mean an error you do not catch inside the window is hard to recover.

The published rejection and error rates, and what each one measures

Part of the gap has an explanation. A rejection rate counts what the maker refused, so an error the maker paid never appears in it. The seller figures count something else: the 4% is a share of chargebacks that "result in errors",4 and the 5–10% page calls its figure a share of value in one place and of volume in another, with no method.5 One seller also reports over 98% clean first-pass rates "with clean, accurate data", as another quoted it.2 Use all of these as prompts to measure your own.

When the reserve is wrong

A maker books its estimate of future chargebacks, the chargeback reserve, when it sells to the distributor, based on what sits in the channel: the stock distributors still hold. If that estimate is wrong, revenue is wrong.

This is a device issue as well as a drug issue. A medical device maker's 2025 annual report calls distributor chargebacks the majority of its variable consideration (revenue whose amount is uncertain at sale) and "subject to the greatest judgment", with the reserve moving on distributor sales and inventory.33 A medical supply company lists chargebacks among items that "may require significant judgment".34 An orthopedic device maker reduces revenue at the time of sale for estimated chargebacks.35

These filings are old, and they are the public record we could read. We name the filers only in the sources. The point is the mechanism: an error in chargeback data reached the financial statements.

Why a drug chargeback can reach Medicaid

For prescription drugs, the regulation counts chargebacks in the calculation of average manufacturer price (AMP), and Best Price, the lowest price to a wholesaler, retailer, provider, health plan or government entity in the period, is calculated net of them.19,20 A wrong contract price or a wrong customer class can therefore flow into the figures a maker reports to the government, the figures behind Medicaid drug rebates.40

Knowingly false information can draw a penalty of up to $100,000 per item in the statute, $236,451 at the current maximum, and late data up to $10,000 a day, $23,647 at the current maximum.40,41 Medical devices are outside the Medicaid drug rebate program, so a device maker carries the financial reporting risk and not this one.

How much money moves on chargebacks, and why contract data matters more each year

In shortWe found no published total for US chargebacks. Individual annual reports show they can run from about 4% to nearly 10% of a drug maker's gross product sales. The wider gross-to-net gap, the difference between list-price sales and what a maker keeps, reached $416 billion for brand drugs in 2025, and several 2026 changes add new prices, identifiers and channels to the data a chargeback depends on.

A device maker's numbers are harder to find. Device filings describe the reserve in words, and we found no device maker publishing the same share. Spend time on your own ratio instead: chargeback credits issued last year over gross sales through distributors.

Four changes that raise the stakes on contract data

What software already does on chargebacks, what AI reading adds and what must stay with a person

In shortLine matching against contract files has been software's job for years. New AI tools can read the documents behind a line and quote their source, which brings the running total and the paper into reach. The decision needs a fixed test with a named owner, because a reserve, an audit and a distributor all need the same answer tomorrow as today.

What an auditor will ask about an AI step

No regulator has written a chargeback-specific expectation for AI that we could find. The expectation comes from financial reporting: management must assess the controls behind reported revenue each year.22

An auditor testing a chargeback control will want to know who may change a rule, how a change was tested, whether the same input gives the same output, who approved each payment, and whether the original document survives untouched. A model that gives a verdict is hard to test that way. A model that reads, with a fixed test deciding and a person signing, fits the questions much better.

How chargeback teams work today: analysts, contracts, finance, and the exception queue

In shortClaims arrive daily and most match automatically. Analysts work the exception queue by hand, opening documents and looking up contracts, rosters and past claims. A contracts manager keeps the reference data right, a controller signs off on the numbers, and internal audit tests the record. The hours go into looking things up.

Reviewer to internal audit: who handles a stopped claim

Six steps between a claim file and a credit memo

Where a stopped claim's three and a half hours go

The Rutgers study found at least 10 different systems in use for contract and chargeback work, including homegrown and legacy ones. It also found that lead time and communication decide how well teams come through a change, and that a system conversion can send "different or new data fields" through EDI.1 HIDA's 2023 conference panel made the commercial point: savings from contract agreements "go away really quickly if we have a lot of transaction issues".12

Spreadsheets, a contract suite or an outside team

Summit Ridge's $7,480 claim: read in minutes, stopped at the contract date, half paid and signed

In shortThis is the example from the top of the page, inside the product. AI agents, each with one job, read each packet and cite each value to its page. Seven fixed rules give the verdict, with no model deciding. Dana Okafor, a chargeback reviewer, sees the failed rule with its comparison and its contract clause, rejects the held line with a reason code and her e-signature, and the clean line is credited.

1

The whole claim cycle on one line

Tuesday, October 6, 2026, 9:52 am. Dana's home screen is the packet line, laid out like a production line with four stations left to right: read by agents, rules decide, people approve, paid. This cycle holds 2,480 packets: 64 are stopped at a rule with $412,860 held, 18 are ready to approve for $286,430, 2,397 are decided, and one is still being read.

The stopped packets sit under the seven rule lights, each tagged with the rule it broke. The median time per agent is on the left, and the line log below records every step in order.

2

Agents read the 9:52 intake and cite each value

Dana presses Run the intake. Three packets ride onto the line: two from Summit Ridge and one from Crestline Health Distribution. The intake agent groups the files for each claim and sorts the pages. The extraction agent reads 6 lines and 72 fields, each cited to its page. The validation agent checks the documents against the contract and notes that invoice SRD-60458 is dated October 2. The history agent searches 1.2 million past lines.

3

One passes, one was already paid

Each packet goes through the seven rules in turn. CB-88249 passes all seven and moves to people approve. CB-88250 from Crestline stops at rule 4, not claimed before: its invoice line CHD-121577 was credited on memo CM-42794 on October 5. The panel shows what the agents read, the rule's test and sample clause 6.4, one claim per sale. $2,450 is held and $1,960 can be paid.

4

CB-88252 stops at rule 3, contract in force

Then the example's packet stops. The agents read four values, each with a source and a confidence: invoice date October 2, 2026; contract on the line CHP-2023-0144; contract end September 30, 2026, from section 3.1; no extension in the register. The rule compares two dates and fails: Oct 2, 2026 > Sep 30, 2026. The panel states who owns the rule and that no model decides.

The sample clause sits below: the contract ends September 30, unless extended in writing, and its prices apply only to products invoiced during the term. Glenfarrow signed a renewal that starts November 1, and line 2, invoiced September 28, is inside the term. The suggestion: reject the held line and pay the clean one. Held $5,520, payable $1,960, score 84.

5

Dana rejects the held line, with a reason code and her signature

The reject dialog has C-03, contract not in force on the invoice date, already ticked among the seven reason codes. Dana chooses the held line only, $5,520, over the whole claim at $7,480. The note to Summit Ridge is drafted from the evidence, and she can change it. Her e-signature carries its meaning: rejected after reviewing the evidence.

6

The clean line is credited, and the record keeps everything

The line log records it: Dana Okafor rejected the held line of CB-88252, C-03, e-signed, $1,960 credited. The response goes to Summit Ridge, the paid station moves to 2,398, and the panel moves on to the next stopped packet.

Opened from the claims list, the packet keeps the invoice with each cited value highlighted, the claim form, delivery note, price list and history. On the right: the rejection and its signature, the failed rule, the evidence, the clause and the comparison. If Glenfarrow's account team had agreed an extension, the place to record it is the contract register, and rule 3 would pass.

The contract clauses on the screens are the sample's own text. Real clauses vary, and section 07 explains why a courtesy extension may change the answer on this exact case.

Meridale's $70,070 claim: 80 balloon catheters over the order, and an invoice edited 28 days later

In shortMeridale Medical Supply claims 120 catheters against a purchase order with 40 left. No single month looks wrong. The agents add up the order, read the delivery note, which shows 20 received, and find text placed over the original figure. Two rules fail, $43,800 is held and five clean lines worth $26,270 can be paid. The product shows what changed, and a person decides what it means.

1

Two failed rules on one line

CB-88213 arrived October 5 at 7:42 am: 9 files, 6 lines, $70,070 claimed. Line 1 is 120 Dravenzo drug-coated balloon catheters at a contract price of $1,085 against a $1,450 list price. That is $365 each, $43,800 in all.

The verdict panel says hold, score 92, with rule 5 and rule 7 failed. The other five lines passed every rule, so $26,270 can be paid now.

2

360 + 120 = 480, against 400 ordered

Compare side by side puts the invoice next to the claim history for its purchase order. Earlier claims took 140 in July, 160 in August and 60 in September. This claim would take the total to 480, which is 80 over the 400 ordered.

3

The order and the delivery note disagree with the claim

The hospital's blanket purchase order says releases may not exceed the quantity ordered, with no change orders issued. The delivery note for this invoice shows 20 shipped and 20 received, with a receiving signature on file.

4

The file changed after it was created

The validation agent reads the file properties and the text layers. The invoice was created September 2 and modified September 30, 28 days later, in a desktop PDF editor. Text boxes reading "120" and "$130,200.00" sit over the original 20 units and $21,700, and the subtotal still says $21,700.

This is evidence, and rule 7 holds line 1. The suggestion is to reject with reasons Q-05 and A-07 and ask Meridale for the invoice from its billing system. The September report records that Meridale agreed to check releases against the order before it claims.

Duplicates, rosters, units and tiers on stopped claims, and the screens around them

In shortEach rule has its own side-by-side view. A partial verdict holds only the difference. Around the packets sit the claims list, the rule library with its test on last cycle, a dashboard by pattern and distributor, one integrity report a cycle signed by the controller, and settings for who approves what.

The same seven checks on four more stopped claims

Claims, rules and the test on last cycle

The dashboard, the integrity report and who approves what

Agents read and cite, seven rules decide, and a reviewer signs every chargeback payment

In shortEach agent has one job and limits it does not cross. None of them gives a verdict, approves a payment or changes a claim. The rules are fixed and versioned, so the same evidence gives the same verdict. A reviewer approves every payment and e-signs every rejection, and paying against a failed rule needs a second approver.

From intake to the rule check: five agents on each packet

The screens say "6 agents" on a packet: these five, plus a coordinator that runs them in order and tracks the reply clock. A separate reporting agent writes the monthly integrity report from decided claims, for the controller to sign.

Who approves, rejects, overrides and signs the integrity report

Approvers, limits, the hold score and the claim window are settings, recorded when changed. E-signatures and the audit trail support your financial reporting controls. They do not replace them.

The reviewers, controller and contracts manager on the screens

EDI 844 files and PDF packets in; verdicts, credit memos and EDI 849 responses out

In shortThe solution reads the files a chargeback cycle already runs on and leaves behind a verdict per claim with its rule and evidence, stopped packets for a person, credit memos and response files, and a monthly report. Your contract and finance systems remain the systems of record. The numbers below are labeled design targets and estimates.

The sample cycle's 64 stopped claims, by what they broke

Where a stopped claim's hours go, by hand and with the solution

Design targets and what they rest on

Systems a chargeback rollout connects to

For enterprise use: role-based access, maker-checker approvals (a second person checks the first), SOC 2 Type II audited operations and information security managed to ISO 27001, with an exportable audit trail of every step.

What wrong chargebacks cost you in overpayments, rework, tied-up cash and reserves

In shortStart with your own paid claims. We found no regulator or independent study that publishes how much chargeback money is paid in error, and the public figures measure different things. A few published facts frame the stakes: rejections are a small share of claims, the clock can be short, and for drug makers the data also feeds price reports.

Re-checking one paid claim cycle to find your own chargeback leakage

Five costs no calculator counts: audit findings, strained distributors, Best Price and more

What chargeback automation returns when you count dollars caught against hours and cost

In shortPut in your own numbers. The calculator separates money from time, and shows how many dollars of wrong payments must exist for the tool to pay for itself. Nearly every default is an assumption. The share paid in error decides the answer more than any other input, and you can only learn it by re-checking a paid cycle.

Duplicates, order totals, terms, tiers, hours: where a chargeback team gains

Dollars claimed, share paid in error and hours per stopped claim

A team with few distributors, simple contracts and no packets to read will see smaller numbers, and should say so before buying anything. A team with many tiers, blanket purchase orders and a long claim history has more to test.

How ready your contract administration team is: claims, packets, rosters, history and sign-off

In shortPick the answer closest to your team today. Eight questions score 0 to 2; the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

A low score means fixing rosters and contract dates first

What to ask a chargeback software vendor about packets, duplicate checks, rules and evidence

In shortAsk every vendor the same questions, us included. The answers show how the checks run: what the tool reads, what it remembers, what decides a verdict, what evidence you keep and how you would measure it on your own claims.

Pilot chargeback checks on claims you already paid, then decide on three numbers

In shortMeasure first, then test on your own paid claims with errors you planted, then run a live cycle in parallel, then decide with three numbers. None of these steps commits you to a vendor.

1. Time five stopped claims and pull last year's rejection rate, deductions and true-ups. That is your baseline.
2. Re-check one paid cycle for duplicates and date failures. The result tells you whether a business case exists.
3. Gather 500 paid claims with their packets, contracts and rosters, covering your two largest distributors. Copy them and plant known errors of all seven kinds.
4. Run in parallel on a live cycle, with your people deciding and your normal process still paying.
5. Decide with three numbers: catch rate by error type, false holds on clean copies, and hours per stopped claim.

Catch rate, false holds and hours per stopped claim

Pharma and medical device chargeback questions answered: EDI, rejections, duplicates and timing

In shortThe questions contract teams and buyers search for, each answered on its own.