Courier allocation rules that cut freight and RTO

Allocation is a policy, not a preference

Most brands running three or four courier partners still allocate by habit. An executive opens the panel, sees who picked up on time last week, and pushes the day’s manifest there. That is not allocation, that is a mood. The mood costs money in two places: freight you did not need to pay, and returns to origin you could have predicted at the moment of order capture.

An allocation policy is a written, versioned rule set that decides, for every order, which carrier takes it and why. It runs before the label is generated, it is auditable, and when freight per order moves you can name the rule that moved it.

The five inputs that decide every routing call

A workable engine needs only five signals, and most order management systems already carry all five.

• Destination pincode, mapped to each partner’s zone table and to their serviceability file, refreshed weekly rather than at onboarding.
• Billable weight, which is the higher of dead weight and volumetric weight. Most Indian surface networks use a divisor of 5000, so a 30 x 25 x 12 cm carton bills at 1.8 kg even if it weighs 900 g.
• Payment mode. COD and prepaid behave like different products. COD carries a fee of roughly 1.5 to 2.5 percent of order value or Rs 35 to Rs 50, whichever is higher, and it carries most of your RTO risk.
• Order value, because a Rs 4,000 order and a Rs 400 order should not tolerate the same delivery risk.
• An address-level risk score built from your own history of NDR and RTO by pincode, by phone number, and by first-time versus repeat buyer.

Build the rule stack in the right order

Rules run in sequence, and the sequence matters more than the individual rules. Serviceability first, then constraint, then risk, then cost. Reverse that and you will route a fragile Rs 9,000 order to the cheapest partner in a pincode where that partner attempts delivery once.

A stack that survives contact with reality has four layers. Layer one removes carriers who do not service the pincode, do not accept COD there, or breach the weight and value caps written into their contract. Layer two forces specific carriers for special handling, which in most catalogs means fragile, oversized, liquid, or value above the insured limit. Layer three applies the risk rule. Layer four splits whatever remains by cost, with a load balancer so no partner falls below the monthly volume commitment you negotiated a discount against.

Layer three is where the money is. Prepaid RTO in most Indian D2C catalogs sits between 3 and 6 percent. COD RTO sits between 15 and 28 percent, and in low-ticket fashion it runs higher. An RTO does not cost you the forward freight alone. It costs forward freight, return freight, inward handling and QC at the warehouse, and cash locked for 20 to 35 days. As a planning number, price a returned COD order at 1.9 to 2.2 times the forward shipping line.

That multiplier changes what cheap means. Suppose partner A charges Rs 62 for a 500 g zone B shipment and completes 91 percent of COD deliveries on first attempt, while partner B charges Rs 54 and completes 84 percent. Partner B is the more expensive choice on any order above roughly Rs 700 once the RTO differential is priced in. Run that arithmetic per zone and per payment mode once a quarter and you will find at least two lanes where your default partner is the wrong one.

What to measure once the engine is live

Freight per order is a lagging summary, not a control. Four measures let you actually steer.

• First-attempt delivery rate by partner and zone. Anything below 85 percent on metro prepaid is a partner conversation, not a customer problem.
• RTO percentage split by payment mode, price band and partner, computed on the dispatch cohort rather than the calendar month, so returns land against the orders that caused them.
• Billable weight variance, the gap between what you declared and what the partner billed. A steady 4 to 8 percent gap is a packaging problem. A gap concentrated in one hub is a dispute.
• Rule hit distribution. If 80 percent of orders fall through to the cost layer, your risk layer is doing nothing and the thresholds need tightening.

The failure modes that quietly undo the gains

Three things break allocation engines after month two. Serviceability files go stale, so orders route to a partner who hands them to a franchise and adds three days. Rate cards get renegotiated but the cost table inside the engine does not, so the engine optimises against last year’s prices. And volume commitments get ignored, so the partner who priced at 12,000 shipments a month sees 4,000 and arrives at renewal with a repricing.

Set a calendar control against each one. Serviceability refresh weekly, cost table updated within 48 hours of any signed amendment, and a commitment tracker reviewed on the first Monday of every month with the person who signs the contracts in the room.

A 30-day sequence to get there

Week one, pull 90 days of order-level data and compute RTO and first-attempt rates by partner, zone, payment mode and price band. Week two, write the rule stack on one page and have finance and the warehouse lead sign it. Week three, run the engine in shadow mode, where it recommends but does not route, and compare its choices against what your team actually did. Week four, switch it on for one zone, hold seven days, then extend zone by zone.

Brands that do this properly usually take 6 to 11 percent out of freight per order and 1.5 to 3 points off RTO within a quarter. The second number is worth more than the first, and it is the one your allocation policy exists to protect.