Imagine you're a seller on Meesho. You shipped a bedsheet. A few days later, a return arrives at your door. You open it and find a bar of soap. What comes back isn't what was shipped. The seller is left with dead inventory, and to be compensated, they have to prove what went wrong themselves by submitting claims with supporting evidence.

This is not an isolated mistake. It is a pattern that quietly drains value from the system.

Somewhere between a product leaving a supplier and a return making its way back, value starts getting lost. Not through obvious failures like delays or damaged shipments, but through something far more systemic: wrong returns. A sellable product comes back as unusable inventory. At a small scale, it looks like an operational exception. Across millions of shipments moving through the network every month, it becomes a structural cost built into the business itself.

This is where claims and payments stop being a backend process and start influencing the health of the entire ecosystem.

The Real Cost of Wrong Returns

At its core, a claims system exists to correct what went wrong. But in high-scale commerce, the problem isn’t just resolution, it’s frequency.

Wrong returns directly hit supplier margins. The loss is twofold: the original product often becomes unsellable, while reverse logistics costs continue to apply. For suppliers operating on thin margins, even a small percentage of wrong returns can significantly impact profitability. Over the last few months alone, proactive fraud detection and compensation systems have already helped reduce claim rates by nearly 30% across key cohorts.

Suppliers respond in predictable ways:

  • Some increase prices to hedge against expected fraud losses
  • Some absorb the hit and see margins diminish.
  • Rest exit the platform entirely.

What follows is a chain of reaction. Higher prices reduce the demand for the product on the platform, whereas lower demand increases order volume. Supplier leaving reduces the variety of sellers on the platform. Over time, this makes it harder for Meesho to deliver affordable products at scale.

This is why wrong returns aren't a logistics issue. They're a growth constraint.

Why the Old Claims System Wasn't Working

Historically, the burden of proof sat with the supplier.

To raise a claim, suppliers had to:

  • Record an unboxing video upon receiving returns.
  • Upload evidence within a strict time window.
  • Navigate an approval process dependent on multiple signals.

Suppliers are required to submit claims within 7 days of receiving the return shipment.

Even when done correctly, outcomes weren’t always consistent. Most of the claims get approved; however, the process is still tedious, and money gets stuck in the supply chain for about ~15 days, impacting the working capital for the supplier. Many weren’t raised at all because the process itself was too Inefficient.

In effect, the system relied on reactive correction, waiting for a supplier to detect and then report fraud after the damage had already taken place.

How We're Catching Problems Before They Happen

The shift underway is fundamental: from reactive claims to proactive detection.

Instead of asking, “Can we verify this claim?”, the system now asks, “Can we detect the issue before it becomes a claim?”

This is where two structural changes come into play.

1. Using Images to Track Every Package

Every movement in the reverse journey from user pickup to hub transitions to supplier delivery is now captured through image data. Transparent packaging enables visibility, ensuring that what moves through the chain is consistently documented.

At the point of customer return, images serve as the source of truth. From there, multiple checkpoints reinforce traceability.

In parallel, dual scanning mechanisms map packet IDs to shipment labels at every stage. If a mismatch occurs, the system flags it immediately, preventing further movement.

The goal is simple: eliminate ambiguity.

2. Letting AI Make the Decisions Instead of People

Earlier, decision-making depended heavily on individuals, delivery personnel, hub managers, or manual reviewers. This created inconsistencies and, in some cases, opportunities for manipulation or error in the system.

Now, decision-making is being abstracted into systems. Image data is analyzed using models trained to detect mismatches. Patterns are identified across the supply chain. Risk is scored not after the dispute is raised, but at multiple stages of the journey.

In the forward journey, where speed is critical, predictive models identify high-risk orders with significant precision. Current predictive models are operating at nearly 80% precision in identifying risky orders during the forward leg itself.

Instead of slowing down all orders, only a small subset is subjected to deeper observation.

Operationally, these models are currently deployed more aggressively in the reverse leg, where speed sensitivity is lower, while forward leg interventions are limited only to high-risk orders.

This balanced precision without friction is what makes the system viable at scale.

Paying Sellers Without Making Them File Claims

The most important shift is not faster claims resolution. It’s removing the need for claims entirely.

Under the evolving compensation model, when a mismatch is detected:

  • The incorrect return is intercepted before reaching the supplier.
  • The responsible third party in the logistics chain is identified.
  • The supplier is compensated directly, covering both product value and shipping costs.

No claim needs to be raised. No evidence needs to be submitted. No waiting period exists.

For the supplier, the outcome is the same as getting paid for a completed sale. This goes beyond efficiency; this is how we redefine trust.

What's Working So Far

The downstream impact is already visible.

Claim rates have already reduced by nearly 90 basis points in a single cycle. At Meesho's scale, even a sub-1% reduction translates into millions of shipments with fewer disputes, lower supplier losses, and stronger operational trust across the ecosystem.

At scale, proactive debits and automated compensation mechanisms are already helping prevent a large volume of supplier claims before they enter the dispute resolution process.

Over the past few months, proactive fraud detection and compensation initiatives have already helped pass on multi-crore value back to suppliers by identifying and resolving wrong returns before they escalate into claims.

  • Significant volumes of claims are being avoided entirely through proactive detection.
  • Suppliers have received substantial compensation without initiating claims.
  • Fraud-related losses are being absorbed upstream instead of cascading downstream.

The proactive image analysis system driving these interventions has been operationally scaled over the past 7–8 months.

But the bigger shift is not the numbers. It is what those numbers represent.

When suppliers no longer need to price in uncertainty:

  • Product prices become more competitive.
  • Order volumes increase
  • Supplier retention improves

In other words, fixing wrong returns doesn't just reduce cost, it unlocks growth.

The Challenge: Fraud Keeps Evolving

There's a catch.

Fraud is not a static problem. Every system built to prevent it becomes a target for circumvention. Patterns evolve. Loopholes emerge. What works today becomes obsolete tomorrow. This means the solution cannot be static either.

The system must continuously:

  • Update detection models
  • Refine evidence collection
  • Adapt operational workflows

The baseline keeps shifting. Staying effective requires staying ahead consistently and deliberately all the time.

The Real Change: Taking Ownership

Perhaps the most critical shift is philosophical.

Earlier, responsibility for fraud detection was distributed:

  • Suppliers were expected to report issues.
  • Logistics partners were expected to validate integrity.
  • Internal teams acted as adjudicators.

Now, ownership is being centralized.

The platform itself takes responsibility for:

  • Detecting issues
  • Validating evidence
  • Compensating losses

This reduces dependency on external inputs and aligns incentives more tightly with outcomes.

When the system owns the problem, it also owns the solution.

Where This Is Headed: A Marketplace Without Claims

The direction is becoming clear.

A marketplace where:

  • Wrong returns are identified before completion.
  • Suppliers are compensated without intervention.
  • Claims workflows become obsolete.

This is not an incremental improvement in operational efficiency. It is a structural redesign of how trust and accountability are enforced within the ecosystem.

And it matters because trust at scale is not built through policies, it's built through predictability.

When suppliers know that losses will not be passed on to them, they stop optimizing for risk and start optimizing for growth.

That's when the marketplace begins to operate the way it was always intended to.

Final Takeaway

Claims systems were designed to fix mistakes. But at scale, fixing mistakes isn't enough. The real leverage lies in eliminating the conditions that create those mistakes in the first place.

By moving from reactive claims to proactive intelligence, the system doesn't just reduce fraud; it reshapes the economics of the marketplace.

And in doing so, it turns a cost center into a growth engine.