You bought a $400 jacket in January, decided it didn't fit, returned it in February. The card's statement nets to zero. A year later, the bank closes the account for inactivity. Did the refund cancel the activity event? Or did the original purchase keep the card alive even though, balance-wise, you spent nothing?
The answer matters more than people expect, because returns are how a lot of cardholders accidentally test the inactivity rule. They think they bought something, so the card must be fine. The bank, looking at the same statement, sees something different.
The short answer
The original purchase counts. The refund doesn't. A $400 charge that you fully return on day 31 still pushes your inactivity clock forward to the date the original charge posted. The merchant credit is filed separately and is invisible to the inactivity-review engine. The exception is a successful chargeback, which reverses the original transaction rather than crediting beside it — and that can erase the activity event entirely.
Why refunds and inactivity get tangled up
The intuition is reasonable. If a $50 charge and a $50 refund cancel out on the statement, you'd expect the issuer's review system to treat them as a wash. It doesn't. The two events live in different parts of the transaction model, and the inactivity engine looks at one and ignores the other.
Your card has two streams of transactions from the issuer's point of view: outbound (merchant purchases you authorise) and inbound (payments, refunds, statement credits, rewards postings). The inactivity review tracks the most recent outbound posting only. Anything inbound — payment, refund, $25 statement credit for using a rotating category — is invisible to the timer.
The mechanic in plain terms
Take that $400 jacket. Charge posts April 1. Return posts May 15. The bank's transaction ledger now contains two rows: a $400 outbound purchase dated April 1, and a $400 inbound credit dated May 15. The inactivity engine queries the outbound table, finds April 1, and records that as the most recent activity date.
If your issuer's inactivity threshold is 12 months, the card is now safe until April 1 of the following year — not May 15. The refund didn't push the date forward, and it didn't undo the date either. The card simply doesn't care that the money came back.
The reason isn't a quirk; it's interchange. Banks earn money on outbound charges through merchant fees, not on balance-management traffic. Inactivity reviews exist to find accounts that aren't generating any interchange revenue, so the only signal that matters is "did a merchant charge post on this card?" Refunds, payments, and statement credits don't carry interchange, so they don't appear on the radar. This is also why a $0-balance card with perfect autopay history gets closed for inactivity.
The chargeback exception — when activity gets unwound
There is one situation where a refund-shaped event does erase the original activity: a chargeback or fraud reversal that re-classifies the charge itself.
A merchant refund is a separate transaction filed on top of an existing posted purchase. The purchase stays on the ledger; a credit lands beside it. The inactivity engine still sees the purchase.
A chargeback reverses the original transaction. When the issuer's dispute team rules in your favour, the original $400 isn't sitting there with a $400 credit alongside — it is voided. In Chase's and Citi's reporting systems, the transaction status flips to "disputed and reversed," and the inactivity-review query excludes reversed transactions. The activity event disappears with the disputed charge.
Practical effect: if the only "activity" on your card during the review window was a fraudulent charge that you successfully disputed, you may be back at zero from the issuer's clock perspective. Rare in practice, but worth knowing if your only recent transaction is a dispute.
Partial refunds, store credit, and other variations
Three sub-cases come up constantly:
Partial refunds. You bought $400 of goods, returned $250 worth. The original $400 posted charge is intact; you got a $250 inbound credit. Activity event is unaffected.
Store credit instead of a card refund. Some merchants — H&M, Sephora, many small retailers — refund to store credit rather than crediting the card. No inbound credit ever appears on the card statement. The card sees only the original purchase. Counts.
Refund to a different card. If you returned an item bought on Card A but received the credit on Card B (this happens after a fraud-replacement reissue, or with a merchant who can only credit your "current card on file"), Card A sees only the outbound charge — a clean activity event. Card B sees an inbound credit it can't explain. Activity timer on Card A is reset; Card B's is unchanged.
The "test the timing" trap. Some people try to game the rule by buying something on day 363 of an inactivity window and returning it on day 365 to reset the clock and stay flat. It works mechanically, but it's fragile — see the issuer wrinkles below.
Issuer-specific wrinkles
The general rule holds at all six major US issuers, but the edge cases differ:
American Express. A charge that posts and then refunds within the same statement cycle is sometimes batch-reconciled, and the inactivity engine updates the timer with the cycle-close date rather than the original post date. For most cardholders this is a non-event; it matters only if you are cutting the timer extremely close. Amex's window is 12–18 months, so there's usually plenty of room.
Chase and Citi. Both explicitly exclude reversed transactions (chargebacks, merchant voids before settlement) from the inactivity query. Refunds that follow a fully-posted purchase still leave the original posting on the books and counting.
Capital One. The most permissive interpretation in the industry: even authorization-only transactions that never fully post (a pay-at-the-pump pre-auth that later releases, for example) sometimes ping the activity timer. Undocumented and shouldn't be relied on.
Discover, Bank of America. Textbook: most recent outbound posted charge, refunds invisible, reversals exclude the original. No edge cases worth noting.
The free-trial trap that catches everyone
The most common way people misjudge the refund rule isn't through returns — it's through cancelled free trials. A typical signup posts an authorization-only $0 or $1 hold on day 1, then the real charge posts when the trial ends on day 14 or 30. If you cancel before the trial closes, the authorization releases and no charge ever posts.
People often assume that "I signed up for that subscription" means the card got pinged. It usually didn't. Only charges that fully post into a statement count. A cancelled free trial is, from the issuer's review-engine perspective, identical to never signing up at all.
If you're trying to read your statement and figure out whether anything recent counts, the rule is: look only at lines with a posted date, not a pending or authorized date. Five warning signs your card is on the closure track covers what to check in the issuer's app.
How to use the rule without relying on it
The takeaway most cardholders should walk away with is defensive, not offensive. The refund rule is real and consistent enough to know about. It is not reliable enough to build an activity strategy on.
The rule tells you that returning the one thing you bought on a card this year is probably fine — the original purchase still counts. It does not tell you that buying-then-returning is a substitute for actually using the card. A merchant void before settlement, a successful chargeback, an Amex batch-reconciliation timing edge — any of these can quietly remove the activity event you thought you'd banked.
If you want a reliable activity strategy, run a small charge that you don't reverse. A $1/month posted via a recurring service is the cleanest version. Subscription anchors work too, as long as you let the charge settle and don't cancel within the refund window. The inactivity review only ever asks one question — "did a charge post recently?" — and answering yes is cheaper than answering "yes, with footnotes."