All articles
Marketplace Commission Overcharge Detection: A D2C Playbook
Indian D2C brands lose 2-3% of GMV to marketplace fee errors. Here's how to detect Amazon, Flipkart and quick-commerce overcharges in time to file claims.
Ishita Shah
May 28, 2026 · 12 min read
Indian D2C brands lose 2-3% of marketplace GMV every month to commission and fee errors that never get audited. On a ₹50 lakh run-rate that is ₹1-1.5 lakh of recoverable cash, every month, sitting inside settlement reports nobody opens. This guide explains the seven patterns that produce most of those errors, the dispute windows that close on you, and how to detect overcharges in time to actually claim them back. It is written for finance, ops, and founder readers at brands selling on Amazon, Flipkart, Meesho, Myntra, JioMart, Blinkit, Zepto and Instamart.
The short version: marketplace commission errors are systematic, not random. The platform calculates your settlement and you receive it. That asymmetry is the entire problem.
Why marketplace fee errors are systematic, not accidental
A marketplace settlement is the output of a rule engine running over your orders. Category codes, rate-card versions, weight slabs, return reasons, and TCS percentages all feed into a per-order deduction. Each of those inputs can be wrong, and when one is wrong it tends to be wrong on every order in that batch.
The Competition Commission of India's 2024 investigations into Amazon and Flipkart (1,027 and 1,696 pages respectively) found systematic anti-competitive practices in seller treatment. Independent of the antitrust angle, the practical takeaway for D2C brands is that marketplaces are not neutral arithmetic engines. A finance team that treats the settlement file as ground truth is choosing to leave money behind.
Industry estimates put commission and fee leakage at 2-3% of GMV across Amazon and Flipkart sellers (Source: ReconPe, Terra Insight, 2026; FireAI customer data). For a ₹5 crore monthly GMV brand that is ₹10-15 lakh a month before any quick-commerce exposure is added.
The seven commission overcharge patterns
Most of the errors FireAI flags fall into seven repeatable buckets. Run your last three months of settlement files past this list before you do anything else.
1. Wrong commission category
A kurta listed under "Western Wear" instead of "Ethnic Wear" gets charged at 18% instead of 12%. Six points on every order, every day, until somebody catches it. The fix is a one-line category code change in the listing; the recovery is a 60-day backlog of overcharges.
2. Weight-slab misclassification
Amazon Easy Ship and Flipkart Smart fulfilment both charge by a stepped weight slab (0-500g, 501-1000g, etc.). A 480g product that gets dimensional-weight rounded to 600g jumps a slab. On 5,000 orders a month that is ₹40,000-60,000 in wrong fulfilment charges. This is the single largest line item in our audit data.
3. Return commission clawbacks on cancelled returns
When a customer initiates a return then cancels it, the marketplace is supposed to reverse the return-related fee. About 8-12% of return cancellations do not get the reversal. The settlement shows the return fee charged once, the cancellation fee charged again, and the original commission never reversed. Three deductions for a sale that completed normally.
4. Rate-card version applied to the wrong date
Marketplaces revise rate cards quarterly. Older orders sometimes get reconciled against the newer rate card by mistake. If the old rate was 12% and the new rate is 15%, every order placed before the cutover loses three percentage points. This error is invisible without the dated rate card on file.
5. TCS computed on wrong base
GST TCS at 1% is supposed to apply to the net taxable value after returns and cancellations. Some settlement files compute it on gross sale value, including the lines that were later cancelled. The over-deducted TCS appears in your Form 26AS as a credit, so your finance team often does not flag it as cash leakage. It is still cash leakage; you carry it as a refund claim against income tax instead of marketplace.
6. Payment gateway fee mismatch on prepaid orders
Marketplaces deduct a payment collection fee (~2-3% on prepaid orders, ~1% on COD remittance shortfall). The fee is occasionally charged twice (once at order placement, once at settlement) on orders that switched payment mode between cart and checkout. The duplicate is small per order but consistent.
7. Quick-commerce stockout and ad-deduction penalties
This is new in the last 18 months. Blinkit, Zepto and Instamart charge stockout penalties when a dark-store SKU goes out of stock during a fast window. The penalty rate is not standardised, and it can be combined with an "involuntary cancellation" commission line that double-counts the lost order. Quick-commerce ad spend deductions also appear as commission-equivalent line items in some payout schemas, which makes reconciliation against rate cards meaningless without separating the two.
Quick-commerce: three new failure modes
Quick-commerce has become 15-25% of D2C revenue for personal care, food, beverage and grocery brands. The reconciliation tooling has not caught up.
Three failure modes are unique to this channel:
- Penalty stacking. A stockout that triggers an involuntary cancellation gets charged a stockout penalty and a cancellation commission. Both are legal under the contract; both being charged on the same order usually is not.
- Ad-spend deduction baked into commission. Blinkit's payout file does not always separate sponsored-product spend from referral commission. A brand spending ₹3 lakh on Blinkit ads can see ₹3 lakh disappear into a single "platform fee" line.
- Dark-store level rate variance. A 28% commission in one dark store can be 32% in another for the same SKU. This is not necessarily an error, but it is rarely visible to the brand without per-store reconciliation.
If quick-commerce is more than 10% of your revenue, run a separate audit pass for it. The Amazon/Flipkart rule library does not transfer cleanly.
The dispute window math: detection has a deadline
You can only recover what you find inside the dispute window. The windows are tight and they vary by platform.
| Platform | Dispute mechanism | Window | Notes |
|---|---|---|---|
| Amazon India | SAFE-T claim | 90 days from settlement | Filed via Seller Central → Reports → Payments → SAFE-T |
| Flipkart | Seller support ticket (fee dispute) | 60 days from payment cycle | Filed via Seller Hub → Help → Payments |
| Meesho | Payment dispute (supplier panel) | 45 days from payout | Limited rate-card audit; mostly missing-order disputes |
| Myntra | Partner Portal ticket | 60 days | Inconsistent enforcement; escalate via KAM |
| JioMart | Vendor support email | 30 days | No self-serve portal as of May 2026 |
| Blinkit / Zepto / Instamart | Account-manager raise | 30 days (informal) | No standardised public policy |
A brand that closes monthly books on day 15 of the next month, then runs quarterly reconciliation in finance, will miss roughly half of Flipkart's window every cycle. The only reliable path is to reconcile inside the settlement cycle, not at quarter-end.
Manual vs automated detection: the line-item math
A 500-order-a-day D2C brand generates 30,000-40,000 line items a month across settlement, MTR, MPR, returns, ads and payment files. Manual reconciliation, even by a sharp analyst, samples this rather than checks it.
| Manual (Excel) | Automated (rule engine) | |
|---|---|---|
| Line items checked per month | 200-300 (sample) | All |
| Time spent per month | 6-12 hours | <30 minutes review |
| Pattern recognition across months | Analyst memory | Versioned rule library |
| Recoveries documented for dispute | Ad-hoc | Auto-packaged with order ID, rate card, screenshot |
| Estimated leakage caught | 15-20% | 85-95% |
The 15-20% recovery rate for manual workflows is not because analysts are bad at the job. It is because the work is structurally unfit for human attention: thousands of small numerical errors scattered across files that change format every quarter. Code is better at this than people are.
How FireAI detects commission discrepancies
FireAI runs a four-step audit on every settlement file the moment it lands.
Step 1. Settlement ingestion. Amazon MTR, Flipkart MPR, Meesho payout, Myntra Partner export, JioMart vendor file, Blinkit/Zepto/Instamart settlement files are ingested via direct connector or scheduled file drop. Schemas are normalised to a common order-level model.
Step 2. Order cross-reference. Each settlement line is matched to its order in your order management system, your Shopify or Magento export, and your inventory file. Orphan settlement lines (charged but no matching order) are flagged immediately.
Step 3. Rate-card validation. Every order is re-priced against the rate card that was effective on the order date, in the category the SKU was listed under, with the actual measured weight from the inventory file. Mismatches surface as a delta in rupees.
Step 4. Discrepancy packaging. Confirmed overcharges are grouped into dispute-ready packets: order ID, settlement reference, rate card snapshot, calculated delta, and the exact field that diverged. The packet is what your accounts team uploads to the marketplace dispute portal.
The causal-chain view in FireAI also answers the next question your CFO will ask: why did the overcharge happen. A spike in Flipkart return commission claims this month traces back to a return-reason code change in the rate card that finance never received. That is the kind of pattern that determines whether the same error recurs next quarter.
Worked example: ₹3.8 lakh recovered in four months
A personal care D2C brand, ₹1.8 crore monthly GMV split 60% Amazon / 30% Flipkart / 10% Blinkit, ran FireAI against four months of settlement files.
What the audit surfaced:
- Weight-slab errors on a 425g face-wash SKU: ₹2.1 lakh across 8,400 Amazon orders. The product was being dimensionally-rounded to 600g; the listed dim weight was wrong in the catalogue.
- Category misclassification on a gift-set kit: ₹78,000 across 1,200 orders on Flipkart. The kit was charged under "Bath & Body" (16%) instead of "Personal Care - Multipack" (12%).
- Return-reversal misses: ₹62,000 across 340 cancelled returns on both platforms.
- Quick-commerce penalty stacking on Blinkit: ₹39,000 across 180 stockout events, of which 90 had both a stockout penalty and a cancellation commission.
Total recovered: ₹3.8 lakh. Filed inside the Amazon 90-day and Flipkart 60-day windows. Recovery rate from filed claims: 86%. Time spent by the brand's finance team: 4 hours total across the four months, all of it on dispute filing, none of it on detection.
The recurring weight-slab error was the one that mattered most. Without correction, the same error would have produced another ₹6.3 lakh of leakage over the following year.
GST and TCS treatment of recovered amounts
This is the section every guide skips. Recovered commission is not a windfall; it is a tax-touching transaction.
Commission credit notes. When a marketplace approves an overcharge claim, it issues a credit note against the original tax invoice that included the inflated commission. You reverse the input tax credit you originally claimed on the over-deducted GST in the commission. If the original commission invoice was claimed in March's GSTR-3B and the credit note is received in June, the reversal goes into June's return, not amended into March.
TCS over-deductions. TCS over-deducted by the marketplace shows up in your Form 26AS. You do not file a marketplace dispute; you adjust the excess against your income tax liability or claim a refund through the regular ITR process. This is slower (resolved at year-end) but is not a write-off.
Ad-spend deductions that come back. Quick-commerce ad-spend reversals are sometimes booked as commission credits rather than ad-spend credits. If your books separate ad spend from commission expense, the reversal entry needs to match the original deduction's GL line, not just net out the cash.
If you are unclear on any of this, route the question to your CA before adjusting older returns. The downside of getting GST adjustments wrong is larger than the upside of the recovered commission.
Filing a dispute: the playbook
Detection without filing is just an analysis exercise. The filing path differs by platform.
Amazon SAFE-T claim. Seller Central → Reports → Payments → SAFE-T claim. Attach the settlement screenshot, the order ID, and the rate-card evidence. Amazon's policy update in 2025 made some FBA reimbursements automatic, but commission and weight-slab disputes still require a manual SAFE-T raise.
Flipkart fee dispute. Seller Hub → Help → Payments → Raise a dispute. Flipkart requires the payment cycle reference and the order ID in the same ticket; tickets without both are auto-closed. Escalation path is via Key Account Manager if the first response is templated.
Meesho and Myntra. Both run on supplier-panel tickets. Meesho's payment-dispute resolution is faster than Flipkart's; Myntra's is slower and benefits from KAM escalation.
Quick-commerce. No self-serve dispute portal. The raise has to go through your account manager, ideally with a CSV of disputed orders rather than line-by-line emails. Build the AM relationship before you need it.
Across all platforms, the recoverable evidence is the same: settlement line, order ID, rate card, calculated delta. Brands that automate the packaging of this evidence file ~5x more claims than brands that prepare each one by hand.
What to do this week
Three actions that pay back inside a quarter:
- Pull the last 60 days of settlement files from Amazon, Flipkart and any quick-commerce platform you sell on. The Flipkart window is closing first, so start there.
- Run a category and weight-slab audit against your active SKUs. Mis-listed SKUs are the gift that keeps giving until they are fixed.
- Set a settlement-cycle reconciliation cadence, not a monthly or quarterly one. Anything slower than your shortest dispute window (currently 30 days for quick-commerce) loses you money structurally.
If you want FireAI to run the first audit pass against your data, book a demo and bring 60 days of settlement files. The first audit is free, and the recoverable cash typically pays for the first year of the platform.
Related reading from the FireAI library:
Sources cited: Competition Commission of India Amazon/Flipkart investigation reports (2024); ReconPe industry analysis (2026); Terra Insight marketplace fee audit research (2026); Amazon Seller Central SAFE-T documentation; Flipkart Seller Hub dispute policy. FireAI internal customer audit data, Apr 2025–Apr 2026.
FAQ
Keep reading
More from the blog
See FireAI on your data.
Book a walkthrough to see how FireAI turns your business data into AI-driven decisions.