The Ultimate Guide to E-Commerce Profit Leak Audits: Recovering Courier Overcharges & Hidden Fee Creep

Operational profit leaks share a structure: each one is too small to notice on a single order, entirely invisible in the metrics a founder actually looks at, and only material in aggregate. They are not caused by anyone doing their job badly. They are caused by nobody owning the reconciliation between two systems that were each working correctly.

This guide covers the three that recover the most money per hour of effort, what separates a real discrepancy from normal variance in each, and the export columns you need before you can audit any of them.

Leak 1: Volumetric Weight Overcharges

Couriers bill on the greater of dead weight and volumetric weight, where volumetric is (L × W × H) ÷ 5000 for standard courier lanes. Conveyor dimension scanners measure parcels at speed, and polybags bulge, cartons compress, and wrapping distorts — so billed weight drifts above catalogue weight in the courier's favour more often than against it.

The audit problem is separating drift from error. Kepler's weight-reconciler applies a two-gate test: a discrepancy is flagged only when it breaches both a ratio threshold and an absolute gram threshold — warning at ratio ≥ 1.2 with ≥ 100 g excess, critical at ratio ≥ 1.5 with ≥ 250 g excess. Anything else resolves as SAFE_IN_TOLERANCE.

Requiring both gates is what makes the output filable. A ratio test alone flags a 50 g sachet billed at 120 g — a 2.4× ratio worth 70 g of money. An absolute test alone flags 200 g of variance on a 5 kg carton, which is ordinary scanner error. Together they isolate discrepancies that are large both proportionally and in cash.

Deeper treatment, including the dispute-filing sequence: How to Audit Courier Weight Discrepancies.

Leak 2: Dead Freight on Cancelled and Ghost Shipments

Two related failures, one root cause — the order management system and the warehouse disagree about whether a shipment exists.

Dead freight is a freight charge on an AWB whose shipment status is CANCELLED_BEFORE_PICKUP. Nothing moved; the charge should not exist. These are the cleanest disputes available and almost nobody files them, because a cancelled AWB does not appear in the delivered-orders report anyone reviews.

Ghost fulfilment is the physical version: the customer cancels, the cancellation webhook is delayed or dropped, and the warehouse dispatches anyway. Now you have paid forward freight on an order with no revenue, and you will pay return freight to get it back.

The control for the second is a dispatch lock — the warehouse re-checks order status against the OMS at the point of pickup rather than trusting the pick list printed hours earlier. The audit for the first is a simple anti-join: every AWB carrying freight, minus every AWB with a delivery or RTO scan.

Leak 3: Payment Settlement Discrepancies

Gateway settlements are the leak founders most assume is someone else's problem, and they are reconciled least often because the files are awkward: the order sits in your commerce platform, the capture in the gateway, and the money in a bank statement three days later with a batch reference that matches neither.

The findings worth building alerts for are less about fee rates than about money that silently goes missing:

• Double-deducted refunds — a refund processed twice, or deducted from settlement after already being deducted at capture. Kepler surfaces these as PAYMENT_DOUBLE_REFUND_DEDUCTED_CRITICAL. They are rare per-transaction and expensive when they persist, because nothing in your dashboards shows a refund twice.
• Dispute evidence windows expiring — chargebacks where the evidence deadline is approaching and nobody has filed. Surfaced as PAYMENT_DISPUTE_EVIDENCE_EXPIRING_URGENT. An unfiled dispute is an automatic loss, so this is the highest-value alert in the category: it costs nothing to act on and the full transaction value is at stake.
• Settlement gaps — captured payments with no corresponding settlement line after the expected T+n window.

Note what this list does not include: a claim about typical MDR overcharge rates. Effective rates vary enormously by instrument mix, and any figure quoted without your own instrument mix behind it is noise. Reconcile against your contracted rate card.

A Worked Example

The Export Columns You Need

Every audit above stalls at the same place: a missing column. Before starting, confirm you can export:

1. Catalogue — SKU, packed dead weight (g), packed L/W/H (cm). Measure the packed parcel, not the product.
2. Courier invoice — AWB, order number, SKU, billed weight, billed freight, surcharges, zone, invoice date, shipment status. Status must distinguish cancelled-before-pickup from RTO.
3. Orders — order ID, AWB, status and status timestamps, payment mode.
4. Settlement — transaction ID, order ID, gross, fee, tax, net, settlement date, refund and chargeback references.

If a column is missing, request it before building anything. Most aggregators expose all of these; they are simply not in the default export, and an audit built on inferred values produces findings you cannot defend when the courier pushes back.

Sequencing

Run these in order of evidence quality, not size. Dead freight first — it is indisputable and it establishes that you are checking. Weight second, in batches, once your catalogue dimensions are re-measured. Settlement last, because it needs the most data plumbing and the findings are the hardest to explain to a third party.

The compounding benefit is not the first recovery. It is that a courier who knows you reconcile bills differently from one who knows you do not.

Related Guides

• How to Audit Courier Weight Discrepancies: A Step-by-Step Practical Guide
• Courier SLA Breach & Late Delivery Guide: How to Claim 100% Freight Refunds on Delayed Shipments
• Configuring Shiprocket Multi-Courier Routing & Shipping Rate Rules: Operator Tutorial