Leakage and Fraud at Karigar Connect
Karigar Connect is a Lahore-run online marketplace for Pakistani handicrafts. It lists 3,100 sellers, mostly individual artisans, and processes PKR 240 million in gross merchandise value a year, taking an 8% commission. Payment is cash on delivery, with Karigar paying sellers weekly after courier remittance. Analytics show that 34% of buyers who place a second order contact the seller directly on WhatsApp and complete the next purchase off-platform. Disputes - wrong item, damage, non-delivery - run at 6.4% of orders, and the trust team resolves most by refunding the buyer and deducting from the seller. A 0.7% share of orders come from fraudulent sellers who vanish after two weeks. A pilot that held seller payouts in escrow for 14 days after delivery triggered a revolt: 240 sellers paused listings and signed an open letter. The board wants both leakage and fraud fixed within one quarter without raising the 8% commission, and the CFO has PKR 12 million to spend.
1. Diagnose the 34% repeat-buyer leakage. What intervention actually addresses the cause rather than the symptom?
Read the worked answer
Cause: the platform's value to a buyer is discovery and trust on a first purchase. Once the first parcel arrives, the buyer has the seller's WhatsApp number and the seller has every incentive to offer a 5% discount in exchange for skipping the 8% commission. The commission is invisible to the buyer, so there is no buyer-side loyalty pulling the second order back.
What not to do: policing WhatsApp, hiding seller contact details, or inserting fees into off-platform deals. All three are unenforceable at this scale, and the 240-seller revolt already shows how much tolerance exists for restrictive rules.
What addresses the cause - make on-platform worth more than off-platform:
- Buyer-side protection that cannot follow them off-platform. A written guarantee - full refund on non-delivery or damage, resolved in 48 hours - is worth real money to a buyer dealing with an unknown artisan, and it expires the moment they message on WhatsApp.
- Saved address, order history, and wallet credit so a repeat checkout takes two taps.
- Seller-side services artisans cannot self-provide: negotiated courier rates, branded packaging, catalogue photography. These are the reason to stay that a discount cannot replicate.
- Re-price the repeat order. Drop commission to 5% from the second order with the same buyer, so the seller's incentive to defect falls from 8 points to 3 - cheaper than losing the transaction entirely.
Measure on-platform share of repeat GMV, monthly, not total GMV.
2. Should Karigar adopt escrow? Analyse who bears the risk today and who would bear it under escrow, and propose an alternative.
Read the worked answer
Today the platform bears the fraud risk and the seller bears the delay. The trust team refunds the buyer and deducts from the seller, which means the platform absorbs disputes it cannot cheaply prove (6.4% of orders) and eats outright losses on the 0.7% who vanish - 0.7% of PKR 240M = roughly PKR 1.68M a year.
Escrow moves the pain to the artisan's working capital. A 14-day post-delivery hold on a cash-on-delivery business means a small producer waits three to four weeks for money they have already spent on materials. For an artisan with no credit line, that is not a policy change - it is a solvency event, which is exactly why 240 sellers paused listings. Escrow is a rational design aimed at the wrong population.
Alternative - risk-based payout tiers:
- New sellers and any seller above a dispute threshold: 14-day hold, plus ID verification and a small refundable reserve.
- Established sellers (say 30+ completed orders, dispute rate under 3%): full-speed weekly payout, with a clawback right if a dispute later lands.
- A buyer-protection fund sized at roughly PKR 2–3M - about one to two years of expected fraud loss - so refunds are paid from the fund while investigation proceeds, instead of being deducted from a seller who may be innocent.
This gets most of the fraud protection while holding payout delays to the small minority who generate the risk. Announce it as a fast track sellers can earn, not a penalty they must escape.
3. Allocate the PKR 12 million across leakage, disputes and fraud for one quarter. Show your prioritisation logic.
Read the worked answer
Prioritise by expected loss avoided per rupee, not by which problem sounds most alarming.
Sizing the problems: seller fraud is about PKR 1.68M a year of direct loss. Disputes at 6.4% of orders are largely a cost of doing business that also destroys repeat purchase - the economically larger number. Leakage is the largest of all: 34% of repeat GMV is a permanent commission loss that compounds every year a buyer stays.
Suggested allocation (one quarter):
| Item | PKR | Why |
|---|---|---|
| Buyer-protection fund and 48-hour dispute SLA | 3.0M | Funds refunds without deducting from innocent sellers; the guarantee is the anti-leakage asset |
| Seller ID verification and risk-based payout tiers | 1.5M | Directly targets the 0.7% vanish rate at the point where it originates |
| Courier rate negotiation and packaging for sellers | 2.5M | Seller-side value that WhatsApp cannot match - the durable leakage fix |
| Buyer wallet credit and two-tap reorder | 2.5M | Pulls repeat orders back on-platform |
| Data and instrumentation | 1.0M | Cohort-level repeat-GMV tracking; without it none of the above can be judged |
| Contingency | 1.5M | Room to raise the 5% repeat commission experiment if leakage data supports it |
Explicitly not funded: blanket escrow, and any discount-led buyer promotion. Both buy a quarter of quiet at the cost of the seller base or the margin. Report leakage and fraud as separate KPIs so a win on one cannot mask a loss on the other.