The return reaches you nine days after delivery. The flyer is sealed, the label is correct, the weight feels right. Inside is a torn, washed-out t-shirt — three sizes wrong, a brand you've never stocked. Your ₹899 kurta is gone, the refund already left your settlement, and the only question that matters now is brutal: can you prove what came out of that parcel? Without return fraud video proof, the honest answer is no.
Most sellers meet return fraud the way you meet a pothole — suddenly, and with the damage already done. The sellers who stop losing money to it treat it as a system: know the shapes fraud takes, know what evidence wins claims, and run an opening workflow that produces that evidence without anyone having to remember to be careful.
The five shapes of return fraud
Fraud is not one behaviour; it is a small taxonomy, and each entry leaves a different evidence trail.
- Wrong item returned. The most common shape: a cheaper or older product comes back in place of yours. Sometimes deliberate, sometimes a genuine mix-up by a buyer juggling returns from three apps. The claim doesn't care about motive — only about whether you can show what came out of the flyer.
- Used or worn item. The “one-function wardrobe” — bought Thursday, worn Saturday, returned Monday with perfume in the fabric and the tag re-pinned. The hardest to prove without close-up video of the garment's condition.
- Part swap. Your genuine item returns, but the original branded components are gone — the dupatta missing from a set, original buttons replaced. Subtle enough to slip past a casual inspection.
- Missing parts. The set comes back as a single piece; the 3-piece combo is now two. Easy to miss if your team restocks without checking against the packing slip of the original order.
- The empty parcel. Packaging, paper filler, sometimes a brick's worth of cardboard for weight. Rare, but a total loss per incident — and the variant where evidence-at-opening is the only defence that exists.
Why bother classifying? Because the taxonomy decides the response. An empty parcel with an intact outer seal points at a problem before courier pickup; a re-taped flyer points at transit; a worn garment points squarely at the buyer. File the wrong story and the claim dies on the mismatch between what you allege and what your own photos show. Classify first, then claim.
Scale matters here too. At two returns a week, fraud is an anecdote — annoying, absorbed, forgotten. At fifty returns a month, even a modest fraud share inside them is a recurring four-figure leak, and it concentrates: the same reason codes, sometimes the same pin codes, occasionally the same buyer across multiple orders. Sellers who log every incident — order, type, value, outcome — start seeing the repeats within a quarter. The log costs thirty seconds per incident and turns “bad luck” into a pattern you can actually screen against.
The video proof standard marketplaces actually accept
Claim adjudication is a volume business. The person — or system — reviewing your claim has seconds, not minutes, and a queue behind you. Written narratives lose to objective artefacts every time. The evidence standard that wins, across claim types, looks like this:
One unbroken video that starts on the unopened parcel — return label and seal condition clearly visible — and ends on the revealed contents, with no cut in between. The moment a video has an edit, it has a doubt. The order identity in frame: the AWB or return ID on the label, readable. A video of an anonymous parcel proves nothing. Condition close-ups after the reveal: the wrong brand label, the worn collar, the missing dupatta. And the paper trail — your original packing slip and invoice showing what was dispatched, so the gap between shipped and returned is undeniable.
Notice what is absent from that list: anger, adjectives, and honesty. The marketplace cannot see your character. It can see a video.
It also helps to understand the incentive landscape you are filing into. Marketplaces carry their own fraud cost on the buyer side — serial returners, refund abuse — and they do act on patterns; your evidence feeds that machinery even when an individual claim is small. A seller who files clean, well-documented claims consistently is building a track record that makes each subsequent claim easier to approve, the same way a seller who files noisy, evidence-free disputes trains the system to discount them. Think of every filing as two transactions: this claim's rupees, and your credibility balance for the next one.
The scan-to-record workflow
Here is the uncomfortable truth about evidence: every seller agrees with it in April and abandons it by the second week of a sale season. Holding a phone, finding the order, naming the file, uploading it somewhere findable — that is five minutes per return, and at twenty returns a week it quietly becomes the task everyone skips. Evidence that depends on willpower is not a system; it is a mood.
The workflow that survives real life inverts the cost. Scan, open, done. Scan the barcode on the return label — the camera starts recording, already knowing which order this is. Open the parcel in frame. Stop. The video attaches itself to the order, next to the packing slip, the invoice, and the original dispatch record. No file naming, no WhatsApp-group archaeology three weeks later when the claim needs it. This is exactly what Robnu's video proof module does: the scan is the trigger, the order ID is the index, and the evidence is already filed where the claim engine will look for it.
Two practical details make or break the setup. First, the camera position: fixed mount over the opening table, wide enough to keep the parcel and your hands in frame, close enough that label text resolves. A phone on a ₹300 gooseneck stand clears the bar; nobody needs studio equipment. Second, the dispatch side: recording at packing time costs a few seconds per order and gives every claim its “what was actually shipped” anchor. When a buyer alleges you shipped the wrong item — fraud's mirror image, and it happens — the packing clip is the only defence that exists. Evidence cuts both ways, and the sellers who record both ends stop fearing either conversation.
Where claims actually die: the funnel
Talk to enough sellers about fraud and a pattern emerges: the money is rarely lost at the adjudication stage. It is lost earlier, in the boring stages nobody graphs.
Per hundred fraudulent returns (illustrative, but the shape holds): a chunk are never even detected, restocked unopened on a busy day. Of those detected, many have no usable evidence because recording was skipped. Of those with evidence, a painful share miss the claim window — the video exists, the deadline passed, the money is gone anyway. Only then does evidence quality decide the remainder. The funnel says something useful: the highest-return fixes are workflow fixes — inspect everything, record everything, file same-day — not better arguing.
Run your own funnel once and the priorities pick themselves. Count last month's returns, the share you actually inspected against the original order, the share with video, the share of problem cases filed in time, and the share of filings that paid out. Most sellers discover the same uncomfortable shape: the adjudication stage they complain about loudest is the one stage already working, and the silent stages they never thought about — the unopened pile, the unrecorded Thursday — are where the rupees actually leave. Fix the funnel from the top, because every leak upstream multiplies downstream: a return that is never inspected can never be recorded, filed, or won.
Where Robnu fits
Robnu is the agentic OMS for AJIO and Meesho sellers, and fraud defence is one loop inside it: the return arrives already classified against its order, the scan starts the recording, the evidence files itself, and the claim goes out through claims filed for you with everything attached — autonomous filing is rolling out now, with a rare one-click human approval where a marketplace requires it. The outcome lands back on the order, so your numbers show recovery rate as a real metric, not a feeling.
The fraud log lives in the same place: every incident typed, valued, and resolved, so the quarterly question “is this getting worse?” has an answer with a number on it. And because the evidence captures itself, the habit survives the weeks that usually kill it — sale season, a packed returns table, a new person at the bench who has never been told the rules.
Robnu is free for everyone right now — every feature, every order, no card, no timer. When paid pricing eventually launches, sellers under 25 orders/day stay free forever. Start before the next empty flyer arrives: the worst time to build an evidence habit is the day after you needed it.