This post in 30 seconds.

• Around 9% of every return you process is fraudulent (NRF 2025), and the worst of it never touches your return portal.
• You'll get the fraud types worth fighting, the signals that flag a fake before you refund, the policy and tooling levers in order, and the one channel almost every brand leaves wide open.
• Written for founders, COOs, and Heads of CX at $10M-$100M Shopify brands with a CS team and a phone number on the site.

Most brands fight return fraud at the return portal. They tighten the form, add a photo upload, score the order, and call it handled. Then a refund goes out anyway, because the fraud never went near the portal. It came in on the phone, from a caller who told a tired rep the box showed up empty, and the rep refunded it to close the ticket and move on.

I run phone support for a bunch of Shopify brands, which means I read a lot of real return and refund calls. The pattern that surprised me wasn't the portal abuse everyone writes about. It was how often a refund that would never survive the portal sails through on a phone call instead. That gap is the subject of this post.

If you're the founder or Head of CX at a $10M-$100M brand watching your refund line creep up while your CS team gets buried in the same "where's my order" and "I want a refund" calls, the fix isn't only better software. It's consistent enforcement on every channel a customer can reach you on, the phone included. Book a 30-min call and we'll talk through where your returns are actually leaking.

What fraudulent returns actually cost you

Returns are a normal cost of selling online. Fraud is the slice that shouldn't be leaving.

The National Retail Federation projects $849.9 billion in returns for 2025, about 15.8% of annual retail sales. Online runs hotter: 19.3% of online sales come back. And 9% of all those returns are fraudulent. On a $30M brand running a 20% online return rate, that math gets uncomfortable fast.

The refund is rarely the real cost. The un-resellable inventory and the rep time behind every flagged return are. A wardrobed dress comes back worn and you write it off. An empty-box claim ties up a rep for twenty minutes and a warehouse check. A serial returner cycles five orders through your CS team before anyone notices the pattern. NRF found that 85% of retailers now use AI to fight return fraud, but only 45% find it effective, which tells you the tooling alone isn't closing the gap.

There's a second cost nobody puts on the P&L: precedent. Refund one bogus claim to make a caller go away and you've trained that caller, and the loose networks they sometimes belong to, that your phone line is the soft door. That's the cost that compounds.

The fraud types you're actually fighting

You can't write one rule that catches all of it, because "fraudulent return" covers five pretty different behaviors. Each one has a tell, and each one tends to arrive on a specific channel.

NRF's tracked-incident data backs the ranking: 71% of retailers who track it saw overstated return quantities, 65% saw empty-box or "box of rocks" returns, and 64% saw decoy returns with counterfeit items. And this isn't a tiny fringe of bad actors. Close to two-thirds of consumers admit to at least one costly returns behavior, and 45% think bending the truth on a return is fine.

The point of naming them is simple. Wardrobing is a policy problem. Empty-box is a carrier-and-inspection problem. Item-not-received is a phone problem. If you treat all three the same, you over-block your good customers and still miss the calls.

The signals that flag a fake before you refund

The hard part of return fraud is that it looks legitimate on the surface. The fix is history plus a few signals that rarely fire on a real customer.

• Time-to-return under 48 hours on a use-once category. A formal dress back the Monday after a Saturday wedding is the textbook wardrobing pattern.
• More than three returns in 90 days from one account. That's the threshold most teams use to kick a customer into manual review instead of auto-approve.
• Address, payment, or device clustering. One person running several "accounts" to dodge per-customer return limits.
• An item-not-received claim after a delivered scan. The carrier says it arrived. That doesn't make every claim false, but it earns a second look.
• A weight mismatch on the inbound parcel. A 200-gram package that should weigh two kilos is the empty-box tell before anyone opens it.
• Photo-request abandonment. Ask for a quick photo of the damage and a real customer sends it. A low-effort fraudster goes quiet. That one ask alone clears a big slice of fake damage claims.

None of these are proof on their own. Two or three firing together is your cue to inspect before you refund, not to ban on the spot.

The policy levers, before you buy anything

Most of the lift here is free. You change words and thresholds, not your tech stack.

• Write the policy in plain language and put it everywhere. Product page, checkout, confirmation email. A clear return window, condition rules, and proof requirements remove the "I didn't know" defense. If you don't have one yet, our return policy generator gives you a starting draft.
• Tighten the window by category. Use-once, seasonal, and high-AOV items don't need a 90-day window. A shorter clock on those kills most event-driven wardrobing.
• Add a restocking fee on high-risk SKUs. Not everywhere. Just the categories that get abused.
• Set account-tier return limits. A return-rate cap that quietly routes a customer to manual review before the next auto-approve.
• Require structured return reasons and proof. Forcing a reason on the record, plus a photo on damage claims, raises the effort enough to deter the lazy fraud.
• Keep a repeat-abuser watchlist, and document the reason before you ban. A ban you can explain holds up. A ban you can't turns into a chargeback and a one-star review.

Here's the objection that stops most teams: "I don't want to punish good customers to catch a few bad ones." That's the right instinct. Return-fraud research generally puts the false-positive tolerance around 5% before you start damaging customer experience. So the goal isn't zero fraud. It's catching the obvious patterns without friction on the 95% who are fine. A good returns policy does most of that quietly.

If your refund line is climbing and you want a second set of eyes on where the leaks are, book a 30-min call and we'll look at it with you.

The tooling layer, and why 45% say it doesn't work

Once policy is tight, software helps. It just isn't the whole answer, and the NRF number proves it.

Returns-management platforms automate the boring parts: approvals, condition checks, restocking, and refund timing. Returns management software plus a Shopify returns app handles the volume your team shouldn't touch by hand. Risk-scoring tools go further, linking transactional and account data to flag high-risk returns at intake. Predictive return models have cut return rates by up to 13% for the retailers using them.

On the back end, chargeback tools matter too, because a fair number of fraudulent returns end in a chargeback after the refund clears. Fighting those is its own discipline, and chargeback prevention belongs in the same plan.

But remember the NRF finding: 85% of retailers run AI against return fraud and fewer than half say it works. That's not because the models are bad. It's because they guard the portal. They score orders and returns that come through the form. They don't pick up the phone. Which is the channel almost nobody is watching.

The channel everyone forgets: the phone line

Here's what reading thousands of real return and refund calls taught me. A lot of refund fraud isn't a technical exploit at all. It's social engineering aimed at your CS reps.

Fraud researchers describe a whole underground economy built on this. Riskified documents refund-fraud outfits that take a 15-40% cut and "manipulate the systems responsible for processing returns and refunds, usually the merchant's customer service representatives." The move is simple. They fabricate a complaint, invent a customer-service failure, and lean on your escalation procedure until someone issues a refund on a legitimate purchase. The portal never sees it. A person did.

And the thing about a person is that a person bends. At hour nine of a stacked queue, your rep grants the exception "just this once" to close the ticket. A real customer would never know your exact return window. A fraudster has it memorized and uses your rep's fatigue against them. That inconsistency, the gap between what your policy says and what your reps actually do at 6pm on a Friday, is the door.

This is where Ringly.io fits, and it's a different job than a fraud-scoring tool. Ringly is AI phone support for Shopify brands. The AI answers your inbound calls 24/7, and on a return or refund call it applies the same rules every single time: the same return window, the same proof-of-purchase ask, the same escalation path. It doesn't get tired, it doesn't grant the "just this once" exception, and it logs the full transcript of every call so you have a record instead of a he-said-she-said. The genuinely hard calls, a real defect, a grieving customer, a legitimate exception worth making, escalate cleanly to your team by rule, not by mood.

"My customers also feel like it's a normal person. They feel like they can communicate if they have questions."
Claudia Droge, TechCraft Studio

It pulls your policy and order data from your knowledge base and your Shopify store, so it can check the order status before it talks return windows, and run your return logic through custom actions. Across 50+ brands, the AI resolves 73% of inbound calls autonomously at roughly $0.42 per resolved call. WashCo, a Shopify brand we launched, recovered $22,664 in attributed revenue in its first 7 days on the phone, because consistent handling does more than block fraud, it also rescues the legitimate sales that were getting lost in the WISMO pile.

There's a cost story underneath this too. Take a brand running a 6-rep CS team:

That's roughly 70% of repeatable calls, the order-status, refund-status, and return-window questions you hear all day, handled by the AI with the policy applied the same way every time. The other 30%, the calls that actually need judgment, go to your team. Exact pricing gets set on a call, but those are the savings shapes we see across 50+ Shopify brands. If your reps are the ones caving on refunds, book a 30-min call and we'll map what your phone line is actually leaking.

Frequently asked questions

Is return fraud illegal? Yes. Deliberately deceiving a retailer to obtain a refund (claiming an item never arrived when it did, returning an empty box, wearing and returning) is fraud, and at scale it can be prosecuted. Most brands handle it commercially first with policy and bans, and escalate to law enforcement only for organized, high-value cases.

What's the difference between return fraud and return abuse? Return fraud involves clear deception, like an empty-box claim or a fake non-delivery. Return abuse is exploiting a generous policy without outright lying, like serial bracketing or wardrobing on the edge of the rules. The line matters because fraud justifies a ban, while abuse usually calls for a policy change instead.

How do I stop wardrobing specifically? Tighten the return window on use-once and event categories, require the original tags and condition for a full refund, and watch for sub-48-hour returns on those SKUs. A restocking fee on high-AOV items removes most of the incentive without touching your everyday customers.

Can I just ban serial returners? You can, but document the reason first and communicate it clearly, or you'll trade a small loss for a chargeback and a public complaint. A better first step is an account-level return-rate cap that routes the customer to manual review rather than an outright ban.

Will tightening my return policy hurt sales? A clear, fair policy rarely hurts conversion, and an unclear one quietly invites abuse. The risk is over-blocking. Keep false positives under about 5% so the 95% of honest customers never feel the friction.

How does an AI phone agent help with return fraud? Most refund fraud over the phone works because human reps apply the policy inconsistently under pressure. An AI phone agent applies the same return window, proof rule, and escalation on every call, logs the transcript, and hands the genuinely hard cases to your team, so the exceptions are decisions you made on purpose.

Which return fraud signals should I watch first? Start with return rate per account (more than three in 90 days), time-to-return on use-once categories, and item-not-received claims after a delivered scan. Those three catch most of the volume with almost no false positives on real customers.

Talk to us

If you run a $10M-$100M Shopify brand and your refund line is climbing while your reps quietly cave on the phone, a 30-min call is the fastest way to see which channel is actually leaking and how consistent enforcement plugs it.

The 3-layer guarantee.

1. Live in 14 days or it's free until launched.
2. 65% resolution in 90 days or we refund the last 3 months of subscription fees.
3. We keep working free until we hit 65%.

Ruben (Ringly co-founder) takes these calls personally.

Book a 30-min call →