COD refusals are a 30% tax on Pakistani e-commerce and everyone pretends it's normal
I ship 150-200 mobile accessory orders daily from Karachi, and on a bad week 32% come back refused: customer not home, customer 'ordered by mistake', customer's number off, customer's teenage son ordered without telling anyone. Each refusal costs me courier charges both ways plus 10 days of my capital riding in a courier bag plus handling damage — roughly PKR 350 per refusal, on products with PKR 200 margins. Prepayment would fix it and prepayment is refused by customers burned by sellers who shipped stones; the distrust is mutual and earned on both sides. What the ecosystem needs is order-level trust infrastructure: a shared delivery-history score per phone number (this customer accepted 14 of 15 CODs — ship freely; this one refused 6 of 9 — require partial advance), robo-call confirmation before dispatch in the customer's language, and escrow-style partial payments through mobile wallets where the customer pays 20% to confirm. Couriers hold this data in silos and use none of it. Whoever federates it — with privacy handled properly — takes ten points off the national return rate and becomes essential to every seller in the country overnight.
A high-conviction problem with strong founder-market fit signals. The combination of severe price asymmetry, accessible demographics, and existing infrastructure makes this buildable within 9 months by a small team.
Solutions · 4
Federated delivery-history scoring across couriers: the shared trust rail every seller needs
Inside my marketplace's data I can already see it: 8% of phone numbers cause 61% of refusals. The fix is federating what each courier knows into one consented scoring rail: phone-number-level delivery outcomes (accepted/refused counts, not personal data) contributed by member couriers and marketplaces, returned as a risk tier with recommended action — ship freely, robo-confirm first, or require 20% wallet advance. Sellers act on the tier; every outcome feeds back. Privacy handled via hashed identifiers and no purchase details shared. The couriers' incentive: refusals cost them reverse-logistics losses too, and members get the scores their competitors lack. I have drafted the schema and the consent framework; this needs a neutral operating company, two courier MOUs, and six months — and yes, I am ready to leave my dashboard job to run it.
Wallet-confirmed partial advance: PKR 200 to confirm the order, adjusted in COD — refusals drop by half
From my own store's experiments: when I required a PKR 200 JazzCash advance on orders above 3,000 (framed as 'confirmation deposit, adjusted in bill'), refusals on those orders fell from 30% to under 12% — the casual orderer disappears and the genuine buyer barely notices. What killed it as my standalone policy: payment friction lost me some real customers to competitors without the requirement. As an ecosystem standard triggered by risk score (only risky numbers see the deposit ask), it becomes fair and universal. Pairs exactly with the federated scoring — score decides who pays the deposit.
Fix the delivery-attempt layer too: scheduled windows and rider call protocol cut the 'not home' half of refusals
Dispatch truth: a chunk of 'refusals' are us, not them — rider arrives unannounced at 2 PM to an empty house, marks refused, moves on because his day has 60 stops. Scheduled delivery windows chosen at checkout, a mandatory pre-arrival call logged in the app, and one same-evening retry before any refusal gets recorded would reclaim a third of the losses before any scoring system runs. The scoring rail should actually require member couriers to meet this attempt protocol — otherwise sellers pay for customer risk that is really courier laziness. I supervise forty riders; the protocol is enforceable if the app enforces it.
Category-level playbooks from the return data: what to never ship COD and when prepaid discounts win
Once the federated data exists, the meta-lessons are worth their own product: which categories and price bands refuse most, where a 5% prepaid discount beats eating COD risk, which cities justify advance-only policies, what order-time patterns predict refusal. My agency reconstructs these playbooks per client from fragmentary data at consulting prices; a benchmark subscription built on the real rail would be every D2C brand's first analytics purchase. The scoring company should own this layer — it is the margin on top of the utility.
Discussion
My return rate was 28% until I started voice-confirming every order over PKR 2,000. It cost me an hour daily and cut refusals by half. Automate my hour and you have a product.
Dispatch confession thread continues: yes, some 'customer refused' entries are rider shortcuts on overloaded days. The attempt-protocol solution names our sin correctly. Enforce it on us and everyone wins.
Agency data across six brands: refusal rates vary 3x by category at identical price points. Half this problem is sellers shipping COD in categories that structurally should not offer it. The playbook layer matters.
Buyer's side: I refused a COD once because the seller shipped the wrong color and prepayment would have trapped me entirely. The distrust is mutual and earned. Escrow-style partials protect both directions.
Exactly why the scoring rail scores sellers too in v2 — delivery-success and description-accuracy both directions. Trust infrastructure cannot be one-sided or it becomes a blacklist.
Marketing student note: campaigns amplify this — flash sales spike frivolous orders that refuse at double rates. Ad platforms sell the spike; sellers eat the refusals. The analytics playbook should price this in.
Home baker's COD note: my cakes cannot survive a refused delivery — perishable COD is gambling with butter. The risk-score tiers would let me offer COD to proven accepters only, which is the entire difference between growing and staying WhatsApp-small.
Electronics COD is the worst tier — high value, high curiosity orders, high refusals. I stopped COD above PKR 10k entirely and lost a third of orders but kept my margins. The score-triggered deposit would give me that third back.