Has this exact image been used before
Each submitted photo is fingerprinted and compared against every photo your store has received before.
The same image arrived with order 10432 in March.
Evidence forensics for post purchase claims
Damaged, missing, never arrived. Someone should check the evidence.
These are the claims you settle without ever seeing the product. What you have instead is whatever the customer sent you and whatever the carrier recorded. VerifyReturns reads those against each other and against the order itself, then hands your team a case file stating what the evidence supports and what it does not.
Built for brands shipping fragile or bulky goods, where return freight costs more than the item, and for consumables that are gone before anyone asks for them back.
The gap
Your returns platform scores the shopper. Nobody reads the evidence.
Most returns and risk tools flag suspicious customers from order history and device signals. That work is useful, and it stops at the person. The evidence attached to the claim goes past unexamined, whether that is a photograph nobody opened or a tracking history nobody read. Image generation now sits in every shopper's phone, and the evidence is still the thing your refund turns on.
9%
of retail returns are fraudulent, on a count that excludes gray area abuse
NRF, 2025 Retail Returns Landscape
1 in 4
refund dollars were abusive, in an analysis of more than a million refund claims
Riskified, 2024
$103B
lost to returns fraud and abuse in the United States in one year
Appriss Retail and Deloitte
What we check
Six questions asked of every claim before a refund leaves
The same six questions run whether the parcel arrived broken, arrived wrong, or never showed up at all. Each one returns a fact your team can verify and repeat back to the customer without accusing anyone of anything.
Is it your own catalog photograph
We match the evidence against your product imagery, because the easiest photo to send is the one already on your listing.
This is your product shot for SKU 4471, not a damaged unit.When was it taken
Capture time is read wherever the file still carries it, then compared against the carrier delivery timestamp.
Captured six days before the carrier delivered the parcel.Does the carrier agree
Tracking history and exception codes are read alongside the written claim. On a claim that the parcel never arrived, this is most of the answer.
No damage exception appears anywhere in transit.Does the claim match the order
Claimed items and quantities are checked against what shipped, and against your own policy window.
The claim lists a SKU that was not in this order.What does the pattern look like
Claim history is summarized as a plain count, never as a hidden score about a person.
Five refund requests across eight orders, photo only, none returned.
The case file
One case file per claim, written so it survives being forwarded
Your agent reads it once and decides. Every line is a fact with a source, so the file reads the same way to your customer as it does inside a chargeback response. This example is a damage claim. A never arrived claim reads the same way, with the carrier record doing most of the work.
Order 11884 · Claim received 3 days after delivery
Evidence insufficient to approve as submitted
Findings
- The submitted image matches an image received with order 10432 on March 4. Both were filed as a cracked base panel.
- Carrier tracking shows five scans and no damage exception between the origin facility and delivery.
- The photograph shows a two seat frame. This order contains the three seat configuration.
- This customer has filed five refund requests across eight orders. None involved a returned item.
Recommended next step
Ask for one new photograph showing the damage, the packaging, the shipping label, and a handwritten note with the code 4K7Q placed beside the item. A customer holding a genuinely damaged product can produce this in a minute.
Every input, finding and human decision is logged and timestamped.
Where we draw the line
Three things this product will never do
Call anyone a fraudster
We report what the evidence shows and what it fails to show. A photograph cannot prove intent, so we never assert it. Your team keeps the vocabulary that keeps you out of trouble.
Deny a claim on its own
Approving automatically is fine. Asking for more evidence automatically is fine. A denial always waits for a named human, because that is the decision someone will one day be asked to explain.
Show a score we cannot explain
A confidence number you cannot trace back to a specific fact is decoration. Every finding in the file points at a photograph, a timestamp, a tracking scan or a line in your own policy.
How it works
Three steps, and none of them ask you to move platforms
VerifyReturns sits beside the tools you already run. Your returns portal and your helpdesk stay where they are, and your policy stays exactly as written.
Connect and forward
Link your Shopify store and forward claims to one address. We read the order, the customer history, the carrier record and your written policy. You migrate nothing and rebuild nothing.
We examine the evidence
Photos, video and the carrier record are read against the facts of the order. Once you are running, a case file comes back the same working day.
Your team decides
Approve, or ask for the evidence that is missing. Every override is recorded with your reason, so the file follows your policy instead of a generic one.
Start here
Send your last 100 claims. We will tell you which orders the evidence does not support.
You install nothing and you sign nothing. Export the claims with their tracking numbers, attach whatever the customers sent in, and the report comes back naming the orders where the evidence holds, and the ones where it does not.
- A spreadsheet of claims with tracking numbers and a folder of images is enough
- Findings come back within three working days
- Your files stay in your folder, we never see customer emails, and copies are deleted after your report