Voids, refunds and discounts: loss prevention controls at the POS

Distinguish events by their effects

A POS label alone does not establish what happened. Check whether a void stopped an operation before completion or whether an already completed payment required a refund. Available actions depend on the provider, payment method and operation state.

A recorded refund may still be pending with the processor. A discount changes the sale’s commercial terms and should be evaluated against the relevant campaign or authorization. Do not group all these events into a single cash difference.

Reconstruct the sale and subsequent changes

Link each event to its ticket, line, amount, currency, promotions and payment references. For partial refunds, compare cumulative amounts with the original operation while accounting for legitimate exchanges or adjustments.

Verify money returned and merchandise returned separately when both apply. An inventory record does not prove a bank credit; a refund does not prove that a product returned to inventory.

Three modifications that require different controls

For a void, check when the sale ceased to be valid and whether payment completed. For a refund, link the amount to the original sale, earlier refunds and merchandise receipt where applicable. For a discount, verify the calculation base, current policy and authorization rights. One “unusual amount” rule cannot resolve all three problems.

Illustrative example: a COP 300,000 sale has a previous COP 100,000 refund. A new COP 250,000 request exceeds the remaining sale value by COP 50,000. The control must aggregate valid refunds linked to that operation; combining similar receipts or ignoring refunds from another day leads to an incorrect conclusion.

Distinguish refund request, approval and execution. A rejected request must not reduce the balance as though paid; pending execution needs follow-up to its outcome. Retain currency, original and refund references, and each event date. File receipt time alone cannot establish the business sequence.

For discounts, compare eligible operations under the same policy and campaign. A valid promotion may explain an unusual amount. For voids after payment, verify the provider reversal or till effect according to the payment method. Closure evidence must correspond to the effect being controlled.

Start with verifiable controls

Review events without recoverable links, inconsistent cumulative amounts, expired discounts or missing authorizations. Each signal should include the originating records and the conditions under which the operation would be valid.

Exclude technical retries, ticket changes, authorized campaigns and sales outside the extraction period before concluding that an irregularity exists. Operator-level concentration needs context; it does not turn a signal into proof of fraud.

Confirm the money trail

For refunds, retain the provider reference and status. POS acceptance is not equivalent to a completed refund. When reviewing settlement, separate gross amounts, charges and net credits using the available contractual information.

An issuer-mediated chargeback differs from a merchant-initiated refund. When both affect the same sale, check their states and financial effects to avoid counting exposure twice or returning money twice.

Resolve cases and improve controls

The case should identify its owner, missing evidence, decision and outcome. If an incomplete integration caused the signal, fix it and reevaluate the affected population. For authorized exceptions, preserve who approved them and their scope.

Measure how many cases could be resolved with evidence and how long that took. Do not classify every refund as a loss or every void as fraud. The objective is to identify operations needing action and support that action with facts.

• Identify modification type
• Check balance and authorization
• Verify final effect