A 30-day review of your COD losses

An RTO rate is easy to find. An RTO cost takes work, and knowing which single fix would pay back fastest takes more. This is the review we run to answer those questions, and you can run it yourself with a spreadsheet and one courier export.

Export 90 days of orders from your store and your courier or aggregator: order ID, tracking ID, payment mode, order value, pincode, courier, ship date, delivery or RTO date, final status, and every charge applied. Pull the remittance records from your bank for the same period. This is usually two exports and a bank statement, not a systems project.

While you are in there, split your definitions cleanly: cancellations before dispatch, RTO, and post-delivery returns. They are different problems and they need to stay separate.

Tag every order that came back undelivered. For each, cost the full stack: forward freight, return freight, packaging, COD and handling fees, damage or write-off, and a per-order allowance for blocked stock and the acquisition allocation. Where you do not know a component, use a conservative estimate and mark it as an estimate. A number you can defend beats a precise-looking number you cannot.

Then group the cost: by pincode cluster, by courier, by product, by order-value band, by NDR reason. Also match remittances against delivered orders for the same period; the shortfalls belong in the same picture.

Two outputs matter: the total (what failed COD cost you last quarter) and the ranking (which failure groups hold most of it).

Resist the urge to fix five things. Take the largest group and choose the one intervention that addresses it:

• High-RTO pincodes concentrated in slow lanes: courier reallocation or a confirmation gate for those pincodes.
• "Customer not contactable" and fake attempts: same-day WhatsApp contact on every NDR, with verified reattempts.
• Unconfirmed first-time orders: pre-dispatch confirmation, risk-tiered so repeat customers pass through.
• Remittance shortfalls and weight overcharges: the reconciliation process from this series, starting with the discrepancy tab you have never opened.

Write down the prediction before you start: this fix should move this metric by roughly this much. The prediction is what makes week 4 honest.

Run the fix for 30 days. Compare the target metric against the 90-day baseline, and compare the cohort that got the intervention against the rest where you can. Watch the side effects too: a confirmation gate that cuts RTO but also cuts genuine orders is not a win.

Then decide: keep, adjust, or revert and take the next item on the ranking.

It is not a tool purchase, a platform migration or a guarantee. It is one honest number (what failure costs), one ranking (where it concentrates), one fix, and one measurement. Brands that run this loop quarterly know more about their COD economics than most of their competitors, and each loop makes the next one faster.

If you want a second pair of eyes on the numbers, that is the work we do at Humalyze: a review of what failed COD orders actually cost your brand, ending in a ranked list of fixes priced against your own data.