Chargeback Recovery Assessment
A Chargeback Recovery Assessment tells you how much of your retailer deduction book is realistically recoverable, and shows you the arithmetic that produced the number. You send a defined set of deduction records. We classify every deduction against the six canonical dispute categories, apply a published exclusion methodology, and return an estimate of the recoverable portion together with the analyst time that work consumes.
The estimate is reported net of every exclusion. It is an estimate, not a promise: actual recovery depends on your retailer mix, your reason-code mix, and the evidence available to prove each claim. For context, retailer deductions and chargebacks run about 2 to 5% of annual retail volume for wholesalers and manufacturers, and most are never disputed before the retailer’s window closes.
1. Send the data
A deduction detail export, the remittance or check detail, and your top accounts' retailer deduction reports.
2. We classify and exclude
Every line mapped to a category, then eight exclusions applied before any number is produced.
3. You get the estimate and the workings
The net figure, the exclusion waterfall, and the breakdowns behind both.
Want a rough number before sharing any data? Use the ROI calculator, which runs entirely on your own inputs. For the wider problem this sits inside, see chargeback recovery.
Exactly what you send us
The assessment needs three things: a deduction or chargeback detail export, the remittance or check detail showing what was actually withheld, and the retailer deduction reports for your largest accounts. Sample backup documents are optional and sharpen the classification. Nothing else is required to produce an estimate.
Required artifacts
Deduction or chargeback detail
From your ERP or AP system, one row per deduction.
Remittance advice or check detail
As a PDF or an export, showing what was withheld and against which invoices.
Retailer deduction or claim reports
For your highest-volume accounts.
Optional samples that sharpen the estimate
Send a small sample of backup documents for a handful of deductions, not a full document pull: purchase order or EDI 850, invoice, bill of lading, and proof of delivery. These are the same four record types Roy reads once you are live, and they are what make a classification specific rather than probable. Keep it to samples deliberately. A sample set is a two-hour ask; a complete document pull is a two-week ask, and that difference, not any question about the method, is what usually stalls an assessment.
Fields that matter
| Field | Why we need it |
|---|---|
| Retailer / customer | Groups the book by account and by that retailer's dispute practice |
| Deduction or reason code, exactly as the retailer wrote it | Maps the deduction into one of the six categories |
| Deduction amount | Sizes the line and the exclusion waterfall |
| Deduction or check date | Places the line relative to the retailer's dispute window |
| Invoice or PO reference | Ties the deduction to what was ordered and billed |
| Document, claim, or debit-memo number | The handle the retailer itself uses to identify the claim |
Send the reason code raw and unnormalised. The classification maps the retailer’s own code into the six categories, so a code that has already been cleaned, re-labelled, or collapsed into an internal bucket has lost the signal we classify on. To see what those codes resolve to, the deduction code library publishes the mapping.
Formats
An Excel or CSV export, PDF remittances, or a read-only view of your system. A file export is enough on its own. Where a connection is used it is read-only: there is no data migration, your records stay in the system of record you already run, nothing is replaced, nothing is written back, and there is no new software for your team to learn. The full posture is published at ERP integration.
Period
Twelve months where available, so that seasonality and each retailer’s dispute practice are both visible across the book. A shorter period still produces a usable estimate on a narrower base, so send what you can pull cleanly rather than waiting on a perfect year.
Handling
The transfer method is agreed with you before you send anything, and the data is used only to produce your estimate. See privacy.
Retailer-specific evidence expectations are published in the retailer guides, including Dollar General, Kohl’s, Burlington, Bealls, and Kroger.
What we do not count
The estimate is what remains after we remove everything we do not consider recoverable. Eight exclusions are applied before a number is produced, and the estimate is reported net of all eight. This is a filter, not a disclaimer. Each exclusion below is written so you can reproduce it on your own spreadsheet and check our arithmetic before you send us anything.
Outside the retailer's dispute window
Deductions that have passed the point where that retailer will still accept a dispute are removed from the estimate. Retailer windows differ by account, by programme, and by claim type, and we publish no window length for any retailer, because we hold no sourced figure for any of them. What matters for the estimate is that aged deductions are excluded rather than counted, which is also why the deduction date is a required field.
Valid deductions where you genuinely erred
Where the shipment really was short, the label really was wrong, or the routing instruction really was missed, the deduction stands and is excluded. This is a matter of principle, not tactics. Valid deductions are left alone, because knowing when not to dispute protects the retailer relationship and your team's time as surely as winning protects margin.
Duplicates, already recovered, already repaid
The same claim often appears twice in an export, or was reversed on a later remittance, or has already been made whole by the retailer. All three are removed so the estimate is not double counted. This is the single most common reason a raw deduction total overstates the real exposure.
No recoverable evidence trail
Where no document exists that can contradict the retailer's claim, the deduction is excluded. The governing principle is simple: a deduction is disputable when you hold a specific document that contradicts the retailer's claim, and not disputable when that same document confirms it. A claim we cannot prove against is not an opportunity, and counting it would make the estimate a wish rather than a forecast.
Below an economic-to-dispute floor
Very small deductions where the cost of assembling and filing exceeds the amount at stake are excluded. Some retailers formally dispose of small-dollar deductions under their own rules, which removes the option entirely. We publish no threshold, because the floor moves with the retailer, the reason code, and how much evidence a given claim takes to assemble.
Non-deduction dilution
Allowances, agreed markdowns and markdown money, co-op advertising, and contractual terms are not chargebacks, and they are excluded entirely. They are frequently the largest line in a deduction export, and including them is the fastest way to produce an impressive number that means nothing. They are a real cost and a different problem.
See revenue leakageAlready in dispute
Anything your team has already filed is excluded, so the estimate is incremental rather than a restatement of work already in flight. If a line is open with the retailer, that recovery is already being pursued and is not ours to forecast.
Out of scope by lane
Deductions belonging to a carrier, a 3PL, or a freight bill rather than to a retailer dispute are excluded. Roy works one lane: retailer deductions and chargeback disputes. Roy does not audit your pricing, run your allowances, chase your freight bills, or manage your returns, so nothing in those lanes enters the estimate.
Stated limits
The result is a modelled figure derived from your own data and reported net of the eight exclusions above. It is an estimate only: actual recovery depends on your retailer mix, your reason-code mix, and the evidence available to prove each claim. It is not a guarantee, and no part of it is a projection of an outcome. The categories behind the classification, and which of them tend to be disputable, are published in full at the deduction code library.
The same method, end to end
Here is the same method applied to one deduction book, from the file that arrives to the estimate that goes back. The vendor is a mid-market apparel wholesaler and every figure below is illustrative, invented for this walk-through. The figures represent no customer and are not a result produced by ROIAI One.
(a) What came in
Illustrative: a twelve-month deduction detail export of 4,000 lines totalling 2,000,000 in deductions, a set of PDF remittances covering the same period, and claim reports for the four largest accounts.
(b) Exclusions applied in order
| Step | Exclusion | Lines removed | Value removed | Running total (illustrative) |
|---|---|---|---|---|
| Start | Deductions received | n/a | n/a | 2,000,000 |
| 1 | Outside the dispute window | 600 | 300,000 | 1,700,000 |
| 2 | Valid deductions, vendor erred | 500 | 250,000 | 1,450,000 |
| 3 | Duplicates, already recovered or repaid | 300 | 150,000 | 1,300,000 |
| 4 | No recoverable evidence trail | 400 | 200,000 | 1,100,000 |
| 5 | Below the economic-to-dispute floor | 700 | 50,000 | 1,050,000 |
| 6 | Non-deduction dilution | 500 | 400,000 | 650,000 |
| 7 | Already in dispute | 200 | 100,000 | 550,000 |
| 8 | Out of scope by lane | 100 | 50,000 | 500,000 |
| Net | Net recoverable estimate | 700 | n/a | 500,000 |
(c) What remains
Illustrative: 700 lines and 500,000 of estimated recoverable deductions, split across the six categories and across the four largest accounts, with a disputable versus non-disputable line for each. That figure is an estimate of the recoverable portion of this illustrative book. It is not a recovery, not a win rate, and it never becomes an outcome.
The estimate, and the workings behind it
You get back the estimate, the classification behind it, and the exclusion workings, so you can audit the number instead of taking it on faith.
- The net recoverable estimate, with the full exclusion waterfall showing what was removed at each step and why.
- A breakdown by retailer and by the six categories: shortage, concealed shortage, compliance, pricing, incorrect UPC / vendor accuracy, and packing.
- The disputable versus non-disputable split, so you can see which part of the book has a document behind it.
- The analyst-time component, so you can see how much of the work is assembly and filing labour rather than judgment.
- A walk-through call, where we take you line by line through any exclusion you want to challenge.
The sequence is short and has no waiting room: we confirm the data is readable and complete enough to classify, we classify and apply the exclusions, then we walk you through the result and the workings together.
What happens next
If you go ahead, the estimate becomes the basis for the commercial conversation, and Roy starts on the live book: reading the evidence, assembling the packet, and filing with the retailer. See how Roy works and the published case studies. To size the opportunity yourself before any of this, the ROI calculator runs on your own inputs with no data shared.
Request your assessment
Request your Chargeback Recovery Assessment
Send three things to get started: your deduction detail export, your remittance or check detail, and the deduction reports for your largest accounts.
Questions buyers ask
What is a chargeback recovery assessment?
A Chargeback Recovery Assessment is a review of your own retailer deduction data that returns an estimate of how much of it is realistically recoverable. You send a deduction detail export, your remittance or check detail, and the deduction reports for your largest accounts. Every line is classified against six canonical dispute categories, eight exclusions are applied, and you receive the net estimate together with the workings behind it.
What data do I need to send?
Three things: a deduction or chargeback detail export with one row per deduction, the remittance advice or check detail showing what was actually withheld and against which invoices, and the retailer deduction or claim reports for your highest-volume accounts. Optionally, a small sample of backup documents for a handful of deductions: purchase order or EDI 850, invoice, bill of lading, and proof of delivery. Samples are enough; a full document pull is not required.
What file formats do you accept?
Excel or CSV exports, PDF remittances, or a read-only view of your system. A plain file export is enough on its own.
How far back should the data go?
Twelve months where available, so that seasonality and each retailer's dispute practice are visible across the whole book. A shorter period still produces a usable estimate on a narrower base.
Do you need access to my ERP?
No. A file export is enough. A read-only connection is an option, not a requirement, and either way there is no data migration and nothing for your team to replace or learn.
Do you write anything back into my ERP?
No. The connectors are read-only. Roy reads evidence out of your systems, and nothing is written back into them.
What deductions do you exclude from the estimate?
Eight categories are removed before any number is produced: deductions outside the retailer's dispute window; valid deductions where the vendor genuinely erred; duplicates, already-recovered and already-repaid items; deductions with no recoverable evidence trail; deductions below an economic-to-dispute floor; non-deduction dilution such as allowances, agreed markdowns and markdown money, co-op advertising and contractual terms; anything already in dispute; and anything out of scope by lane, such as carrier, 3PL or freight claims. The estimate is reported net of all eight.
Is the estimate a guarantee?
No. It is a modelled estimate derived from your own data and reported net of the eight exclusions. Actual recovery depends on your retailer mix, your reason-code mix, and the evidence available to prove each claim.
Does a human review the disputes?
ROI-AI's own analysts review the exceptions that need judgment. The review layer is ours, not yours: nothing lands in your team's queue and nothing waits on your approval.
What happens to my data?
The transfer method is agreed with you before you send anything, and the data is used only to produce your estimate. Our privacy page covers how we handle it.
How does ROIAI One charge?
On two components: a subscription sized to the estimated recoveries and to the analyst time it displaces, plus a performance fee on recoveries. The assessment is what produces the estimate the subscription is sized against, which is why it comes first. Specific terms are set with you.