TL;DR

Your payment processor captures every authorization, decline, void, refund, and chargeback at the terminal, BIN, and cashier level. Most retailers pull only a monthly summary and miss the patterns that expose theft. The fix is to request 90 days of transaction-level data, segment refunds by employee ID, audit voids in the last minutes of each batch window, and review any cashier whose refund-to-sales ratio runs well above store average, for example more than 1.5 standard deviations. The data is already paid for and already sits in your gateway portal.

What this actually is

Loss prevention in retail historically meant cameras, security tags, mirrors at the back of the store, and bag checks at the door. Those tools catch external shoplifting. They do not catch the larger problem: internal shrink and refund fraud at the register.

The National Retail Federation's security surveys track employee theft and process failures as their own sources of shrink, alongside external theft. Much of that loss touches the point-of-sale system. Cash drawer manipulation, fake refunds processed to a personal card, voiding a paid ticket and pocketing the cash, ringing comp transactions for friends and family, and gift card balance manipulation all leave records in payment processor data.

Every card swipe, dip, or tap creates a row. The processor logs the auth code, the BIN, the last four digits, the device ID, the cashier or operator ID if the POS forwards it, the timestamp to the second, and the response from the issuer. Refunds carry the same metadata plus a reference to the original sale, or a flag indicating no original reference exists. Voids carry the user ID and the elapsed time between auth and cancellation. Chargebacks land later with reason codes defined by each card network.

The Visa interchange schedules and Mastercard interchange materials define what data each transaction must carry to qualify for each rate category.

This is not analytics in the marketing sense. It is forensic accounting using data your processor already collected.

Retail loss prevention through processor data is the practice of analyzing authorization, refund, void, and chargeback records from your payment processor to detect employee theft, refund fraud, and external card abuse.

How it works under the hood

The processor sits between your point-of-sale terminal and the card networks. Every retail transaction follows a defined flow, and at each step the processor writes a record you can pull later.

  1. Authorization request. The POS sends card data, transaction amount, merchant ID, terminal ID, and (if configured) the cashier or operator ID. The processor forwards the request to the network, which routes it to the issuing bank.
  2. Authorization response. The issuer returns an approval code or a decline reason code. Both are logged. Approved auths sit in a pending state until the batch closes that day.
  3. Capture and settlement. At day-end batch, approved auths are submitted for settlement. The processor packages them with a batch ID, batch timestamp, and cumulative total.
  4. Void. If a transaction is canceled before settlement closes, the void is recorded with the user ID that triggered it, the original auth code, the original timestamp, and the elapsed time between auth and cancellation.
  5. Refund. A refund credits a card. The processor logs the cashier ID, the destination card BIN and last four, the amount, and a reference to the original sale, or a flag indicating the refund was entered without an original transaction reference, the so-called manual refund path.
  6. Chargeback. The issuer can claw back funds, for most dispute reasons up to 120 calendar days after processing, under Visa or Mastercard reason codes. The processor delivers the notice with the reason code, the disputed amount, the original transaction reference, and a response deadline your processor sets, which is shorter than the network's own window.

The data lives inside the gateway portal your processor activated at signup. Most retailers never open it past the daily settlement screen. The CSV export is usually free; how much history it keeps varies by provider. Larger merchants can often request an SFTP feed or a reporting API with the same fields delivered nightly.

Inside the export, each row carries dozens of fields. The ones that matter for loss prevention: auth code, transaction type (sale, void, refund, chargeback), card BIN, last four, amount, currency, terminal ID, operator ID, original transaction ID for refunds and voids, settlement batch ID, settlement timestamp, issuer response code, AVS result, CVV result, and the entry mode (chip, swipe, contactless, keyed, e-commerce).

Operator noteIf your POS does not stamp the cashier or operator ID on each transaction, your processor data is half-blind. Check the terminal configuration. Most retail-grade POS systems support employee IDs. Many are installed with the field disabled because the integrator never asked.

Where it goes wrong for operators

Five patterns dominate processor data when fraud or internal theft is moving through the register.

1. Refund-to-sales ratio by employee. Suppose a specialty store's refund rate sits around 2 percent of sales. One cashier running at 4.5 percent is either processing cash returns and keeping the cash, or running fake refunds to a personal card. At $1,500 daily sales per cashier, a 2.5 point gap is $37.50 per day, or $13,700 per year, from a single register. At a 25-cashier chain, four such outliers would be about $54,800 a year.

2. Voids in the closing window. A single void a minute after a sale is normal. It is a customer changing their mind. A pattern of voids in the final 5 minutes before batch close, across multiple drawers or one drawer repeatedly, is not. This usually flags an employee voiding a paid cash ticket so the drawer balances while the cash walks out the door. A repeated weekly pattern of closing-window voids at one terminal is worth a look.

3. Same-card refund magnet. The same destination card, matched by BIN and last four, receiving refunds initiated by multiple cashiers in the same week is almost always an employee's personal card, a spouse's card, or a card belonging to a regular accomplice. At a $50 average refund and 3 hits per week, that is $7,800 per year per ring.

4. Card-present chargeback clusters. Visa fraud codes 10.1 to 10.3 (card-present fraud) and 13.1 (merchandise or services not received) sometimes cluster by zip code, BIN range, or SKU. For example, a single zip producing 15 percent of card-present chargebacks against 2 percent of sales volume can signal a local fraud ring using counterfeit or stolen cards at your terminals. The pattern is visible only when you cross-reference chargebacks against sales by zip.

5. Card testing and rapid declines. Sub-dollar auth attempts (often $0.01 or $1.00), repeated declines followed by an approval using the same card or device fingerprint, or sequential BINs from the same issuer in rapid succession all mean a stolen card is being validated at your terminal before a larger purchase. The processor logs every attempt, including the declines that the cashier and POS may not surface on the receipt printer.

Watch outThe refund-without-receipt rate is one of the most useful metrics for cashier-level fraud. If one cashier's share of unreferenced refunds is far above the store average, look closer before you assume another explanation.

In the example below, one cashier ringing $720 a day in refunds is about a $74,650 a year leak. The data to find that cashier already sits in your gateway.

Worked example with real numbers

Hypothetical profile: Five-location specialty retailer in the U.S. Northeast. $2 million monthly card volume. $45 average ticket. Currently on tiered pricing at a 2.4 percent effective rate. Card-present transactions, mix of Visa, Mastercard, and American Express.

The loss prevention manager pulls 90 days of transaction-level data from the gateway. CSV export, about 133,000 sale rows plus refunds and voids. Imports into a spreadsheet, filters for refunds, buckets the result by cashier ID and store ID.

Store-level refund rates after the pull: Store 1 at 1.7 percent, Store 2 at 2.1 percent, Store 3 at 1.9 percent, Store 4 at 2.0 percent, Store 5 at 3.9 percent. Store 5 is the outlier by a wide margin.

Drilling inside Store 5 to cashier-level refund volume, one closer accounts for 62 percent of total refunds, with $720 in average daily refund value. The other six cashiers at that store run between 4 and 12 percent of refund volume each. The closer also works the only shift where no manager is on the floor for the final two hours.

Pulling the original sale references on the closer's refunds: 48 percent are flagged as refund without receipt, entered manually with no reference to an original transaction. That is far above the store's other cashiers.

The math: $720 daily refund volume x 30 days x 12 months = $259,200 annualized refund flow originating with one cashier. The unreferenced 48 percent of that is $124,416. If 60 percent of the unreferenced portion is fraudulent (the cashier processing a fake return to a personal card, or a complicit shopper returning items pulled from the sales floor), the leakage is approximately $74,650 per year from one register.

That is only the internal slice. Suppose the same data pull also surfaced three card-present chargeback clusters tied to one BIN range, totaling $11,400 across 90 days, all under Visa dispute condition 10.3 (other fraud, card-present environment) and filed by three different issuing banks within a 6-week window. That pattern points to stolen or counterfeit cards being used at the store's own terminals, so the next step is to check entry modes and which cashiers rang those sales.

Hypothetical exampleTotal leakage exposed by one 90-day data pull at this hypothetical retailer: roughly $75,000 per year in internal refund fraud plus $11,400 over 90 days in external card-present chargebacks. Cost to pull the data: zero. Time spent: one afternoon by the LP manager. The processor had been delivering the same data every month inside the portal nobody opened.

The processor sends the same dataset every month. The difference between catching $75,000 in fraud and missing it is whether anyone in operations opens the file.

Operator playbook

  1. Pull a 90-day transaction-level extract from your gateway. CSV format, free at most processors. Request these fields at minimum: auth code, transaction type, card BIN, last 4, amount, terminal ID, employee or operator ID, original transaction reference, void flag, batch ID, settlement timestamp accurate to the second, entry mode, and response code.
  2. Calculate refund-to-sales ratio by employee and by store. Flag any employee running more than 1.5 standard deviations above store mean. Ask your processor whether it offers vertical peer benchmarks; the request itself signals that you are watching.
  3. Audit voids in the final minutes of each batch window. Pull every void where the elapsed time between original auth and void is under 10 minutes. Cross-reference against shift logs and security camera time-stamps. Set your own action threshold from the store's normal void pattern.
  4. Match refund destination cards to employee on-file cards. Pull BIN and last four for the credit card receiving each refund. Compare to the card-on-file list HR or payroll holds for direct deposit and for tipped-staff prepaid cards. Some processors block this comparison by policy. Ask in writing.
  5. Subscribe to chargeback alert networks. Verifi CDRN and Ethoca Alerts are the two main networks; both price by quote. They let you refund the disputed transaction before the chargeback files, which avoids the chargeback fee ($15 at Stripe and Helcim, $20 at PayPal, prices checked Oct 2026), the time cost of writing a response, and the hit to your chargeback ratio with the card networks.
  6. Configure manager approval for refunds over a threshold. Set the floor at $50 to $100 depending on average ticket. Most modern POS systems include the setting. It cuts unreferenced-refund fraud because the cashier can no longer act alone.
  7. Pull chargeback reason codes monthly. Map results by zip code, BIN, and SKU. If a single SKU drives more than 15 percent of fraud chargebacks, tighten CVV or AVS rules for that SKU, or restrict the sales channel until the pattern clears.
  8. Move from tiered pricing to interchange-plus. Tiered statements hide the actual interchange category for each transaction. Interchange-plus statements show every category code, which is the data you need both for forensics and for verifying that the processor is passing through real interchange rates instead of marking up the qualified bucket.