A seller ships the correct product. A buyer claims they didn't receive what they ordered. 

Both tell their truth. In the vast supply chain, where millions of packages move unseen, lies the issue that Meesho's Trust & Safety team works to address.

The honest answer is that no marketplace can watch every leg of every journey. Products move from suppliers to field executives to buyers across thousands of pin codes, and the gaps in visibility are real. What you can do is build systems smart enough to close those gaps; one signal at a time, one pattern at a time, until the logistics workflow becomes more structured and reliable.

That's the journey this team went on. And based on historical records, the numbers show consistent progress.

The Problem: Two People, One Broken Transaction

Wrong Product Returns (WPRs) sound like a buyer issue. Someone ordered shoes, got a shirt, and raised a complaint. That's the surface of it.

But dig one layer deeper and the picture gets more complicated. Because when a buyer raises a WPR, the return journey begins and the product that comes back to the seller isn't always what the seller originally sent. A seller who did everything right ends up with a discrepancy on their doorstep, with the payout on hold and no established proof of what actually happened.

So WPR isn't a buyer problem or a seller problem. It's a standard platform challenge, and it sits right at the heart of what a marketplace is supposed to be. When it's unresolved, both sides of the transaction feel unsupported.

Over three years, Meesho significantly reduced Wrong Product Returns, as records indicate. That kind of reduction is the deliberate, routine effort of a team willing to sit with a problem that has no clean answer and build their way out of it, one system at a time.

The Context: A Marketplace Is Not a Warehouse

To understand why this problem was operationally significant, you have to understand how Meesho operates.

Most people imagine e-commerce as a warehouse: products stacked on shelves, scanned in and out, every movement logged. That model has its advantages: full visibility at every step, a clear chain of custody from shelf to doorstep.

Meesho works differently. Products move directly from suppliers to buyers, picked up by field executives and delivered without ever passing through a central facility. This is what lets Meesho serve millions of sellers across the country, small businesses, first-time entrepreneurs, and artisans, without requiring them to ship inventory.

But it means there's no central checkpoint. No moment where a third party looks inside a package and confirms this is what was ordered. A product leaves a supplier's hands, passes through one or more field executives, and arrives at a buyer's door and at every leg of that journey.

This approach is not a flaw in the design. It's a standard operational aspect of our decentralized system. But it creates a question that has to be answered somewhere: when a dispute arises, how do you know what actually happened?

Without evidence, every claim looks the same. From the outside, all of these are indistinguishable:

  • A buyer who genuinely received the wrong item
  • A seller who packed the wrong product by mistake
  • A field executive who swapped the product somewhere in transit
  • A buyer who received the right product but returned something else to keep it

Any of these is possible. All of them look identical without proof. And trying to resolve them fairly, within our operational volume, without any ground truth to work from, is a challenge.

This was the context the Trust & Safety team inherited. The problem was real, the volume was significant, and standard tools were required to solve it.

The Insight: You Can't Fix What You Can't See

The improvement wasn't a new algorithm or a smarter model. It was a shift in thinking: before you can address the problem, you have to be able to track it.

The team's first move was to collect data where none existed. Two operational practices were introduced.

Transparent packaging for non-boxed products like apparel, so that what's inside a package is visible at every leg of the supply chain without anyone having to open it. A garment that looks right when it leaves the supplier can be confirmed to look right when it's handed to the buyer.

Doorstep image capture: field executives began photographing the product at the moment of delivery. Timestamped, location-tagged, tied to that specific order. 

These two steps didn't close the gap immediately. But they did something more important: they gave the team something to reason from. Factual data. Verified records. A foundation.

And once you have a foundation, you can start refining processes.

The System That Fought Back

Meesho's response was not a single fix. It was a layered set of systems that routinely evolved consistent with standard security operations.

Face recognition for field executives. At random times during random deliveries, executives are required to submit a selfie. These images are checked against a database of previously terminated individuals. If a terminated executive attempts to join the system under a different identity, the flag gets raised. Shipments are stopped. The executive is removed again.

Catalogue verification at return pickup. When a reverse pickup is initiated, meaning someone is coming to collect the returned product from the buyer, the pickup executive now receives the supplier's catalogue image on their device. Their job is simple: only accept the return if what the buyer hands over matches what was originally delivered. If there is a mismatch and the executive picks it up anyway, that is flagged as a violation. Corrective action follows.

The Transaction Risk Management System - TRMS. This is the backbone. TRMS is a data science-driven system that reads signals at every touchpoint in an order's journey and predicts how risky that transaction is. It works in three phases:

  • At the time of order placement, it can flag or cancel orders that show early risk signals.
  • At the time of dispatch, it can trigger additional image capture requirements throughout the shipment journey.
  • At the time of delivery and return, it processes image match results and passes actionable flags to field executives, telling them, in real time, whether to proceed or hold.

TRMS does not replace human judgment. It makes human judgment faster, sharper, and more consistent across orders.

What Comes Next

The reduction has been substantial, which means more than half the problem has been addressed. The other half is what the team is working on now.

The next phase is about systematically identifying discrepancies before it even takes shape. More touchpoints in the supply chain will get image coverage. Evidence generation, which started as a routine step, is now being maintained as a standard layer across the entire order journey.

The ambition is not modest. Meesho continues to maintain structured processes to ensure transaction accuracy at pickup, in transit, at the doorstep and on the way back. Every leg of the journey becomes a checkpoint. Every data point becomes a signal.

Discrepancies will keep evolving. External challenges are not going to stop. But so will the systems built to catch it and right now, those systems are winning.