Where do online returns go? The route has to earn its cost

The refund can happen without a physical return

Walmart’s seller guidance illustrates the choice: a Keep It rule can issue a full or partial refund while allowing the customer to keep the item. The seller configures departments, price and return-order limits; item-level overrides are possible. The policy can be a rational response to expensive reverse freight, even when the object still has resale value. Seller guidance.

Amazon’s 2025 filing reports a $1.6 billion return-allowance liability and a $1.2 billion asset for rights to recover products. Those balances recognize different sides of the return. They are not a cash recovery rate to calculate by dividing one by the other, nor a record of each item’s eventual destination. Return-allowance accounting.

The inventory owner carries a different exposure from the service platform. Amazon says third-party sellers own their inventory even when it provides fulfillment. A marketplace can arrange the customer experience and collect fees while the seller funds much of the merchandise loss. Vendor credits, carrier claims and platform protections can move that loss again. The refund screen does not reveal the final bearer. Inventory and seller-service policies.

An eligible route has to repay retrieval

Eligibility comes first: the product’s condition and commercial restrictions decide which channels it can enter, because a high headline resale price creates no retrieval value if the product cannot reach the channel that pays it. For each eligible route, estimate recovery after cleaning, testing, and storage. Then deduct selling fees, secondary transport, delay, and disposal. The best net route has to repay the added cost of retrieval.

Condition determines both price and eligibility. An unopened unit may return to ordinary stock. A used device may require refurbishment. An opened cosmetic may preserve physical material while losing an eligible resale route. A substituted laptop can have little legitimate recovery value but still warrant retrieval as evidence. Retrieval buys a disposition or an information benefit.

Safety, recalls, identity, abuse prevention, customer commitments, vendor terms and environmental duties can constrain the route. Those obligations belong in the decision before optimizing proceeds. The highest modeled bid is irrelevant when the route is unavailable.

Scroll horizontally to inspect the figure →

Inspection must earn its cost

Consolidation lowers some transport and receiving costs at the price of a later resale. Inspection buys information: a grade, a serial match, a functioning device or evidence that a seal has been broken. The cost earns its place when the finding improves expected recovery, changes eligibility or treatment, or prevents a loss. Verified condition can support a higher selling price within the same resale route. A change of destination is therefore one source of inspection value. Information that changes none of these outcomes has little economic value.

At the assumed labor rate, the $100 headphones support roughly 55 minutes of inspection before retrieval turns negative. The $20 household item cannot support even the two-minute base inspection under its weighted parcel route. Deeper inspection can pay when it opens a stronger resale channel.

Testing helps preserve $33.37 of expected headphone recovery. The cost bridge charges refund, product cost, outbound fulfillment, and freight. It also charges intake, fraud administration, secondary costs, and failed-unit risk, leaving post-return contribution at negative $31.24.

Inspection earns the value of a better decision, not the whole resale price. Consider a hypothetical with equal chances of high and low grade. One eligible resale route nets $40 for high grade and $0 for low grade; an eligible liquidation route nets $10 in either state. Before learning the grade, resale has expected proceeds of $20. A perfect test lets the operator choose $40 for the high grade and $10 for the low, averaging $25.

The perfect test adds at most $5 of information value before its cost. A $4 test earns $1; a $6 test loses $1. I’d measure real grading errors and changed disposition decisions, with transport and intake held out of both alternatives, before buying deeper inspection.

Scroll horizontally to inspect the figure →

The same price buys very different logistics

Holding the price at $20, $100, or $500 reveals how much the logistics depend on the object. Cheap parcel movement can preserve a large fraction of footwear value. Bulky furniture may need a local route even when a buyer still wants it. A large return tax can coexist with a positive retrieval decision because the comparison is incremental: the sale has already been lost; the operator is deciding how much of the loss it can now recover.

Return tax equals kept-sale contribution minus post-return contribution. Landed product cost, original fulfillment, and original payment expense appear on both sides and cancel from this incremental tax. They still determine the post-return loss and the share of kept-sale contribution consumed.

Bulky goods can retain value but fail parcel retrieval

The modeled $500 furniture item has about $3.60 of retail value per billable pound, versus more than $71 for the $500 laptop or phone. Its 48 × 20 × 20 inch package produces 138 pounds of dimensional weight before rounding. Weighted reverse freight of $98.90 exceeds the $85.16 economical maximum.

Centralized retrieval is negative under those assumptions, while store intake, short movement and local pickup remain positive. The local-routing proxy breaks even at about 48 miles. The consumer’s willingness to use the object has survived; a particular network cannot move it economically. A disposition system with another local buyer or collection design could reach a different answer.

At $20, beauty and household returns likewise can work through stores or consolidated drop-off while failing by parcel. Returnless refunds can reflect costly logistics for wanted products.

Scroll horizontally to inspect the figure →

Scroll horizontally to inspect the figure →

A liquidation bid leaves costs on both sides

Liquidation passes goods through wholesalers, local auctions and secondary sellers. A lot may mix customer returns with overstock or salvage, which changes what the buyer can recover.

The winning bidder still has to bring the goods in, inspect them and sell what survives. Freight, repair, fees and another customer’s return sit between the bid and the margin. The original retailer has its own seller-side costs.

The sample contains two completed Walmart/B-Stock lots. The microwave lot implies a bid near 14.6% of stated retail; the eBike lot, about 19.6%. Assumed buyer premiums and freight raise delivered cost to 26.7% and 49.4%; estimated retailer recovery after seller-side fees and preparation is roughly 10% to 12%. These lots do not estimate the market.

Scroll horizontally to inspect the figure →

Scroll horizontally to inspect the figure →

A fraud control has to pay for the harm it causes

Substituted or missing goods receive zero legitimate recovery through the condition probabilities. The administration allowance covers only added investigation and claims work, avoiding a second charge for the lost product.

The illustrative control adds inspection, false-positive customer-service cost, and refund delay. It loses money on $100 footwear and headphones but pays on $500 electronics and accessories, where substitution destroys more value.

Those fraud rates are sensitivity inputs. Calibration needs both the losses prevented and the legitimate customers incorrectly challenged, including added service work and delay. A procedure can identify more suspicious returns while making the business worse. The relevant outcome is the change in net loss and customer economics, with the control’s errors left in the record.

Scroll horizontally to inspect the figure →

A sale to the next owner leaves reuse unobserved

Sustainability claims need denominators and endpoints. Resale or donation counts can show scale without revealing their share of returns, third-party inventory, or all network intake. “Proper disposal” need not distinguish recycling, hazardous handling, incineration, energy recovery, and landfill.

Liquidation records a sale to the next owner. That buyer may resell, repair, export, harvest parts, recycle, or discard the goods. Without observing the next step, the transfer cannot be counted automatically as reuse.

The model tracks direct resale or reuse, downstream-unobserved liquidation, donation, recycling or material recovery, destruction or disposal, and fraudulent or missing units. These are scenario routes; retailer diversion rates remain unmeasured.

The policy creates part of the flow it must process

A retailer choosing its policy is solving a different problem from a warehouse routing an already-returned item. The policy can reduce the perceived risk of buying. It can also encourage returns or change which customers and products enter the flow. A cheaper return process is valuable, but a policy that minimizes returns by deterring profitable purchases may be expensive overall.

Janakiraman, Syrdal and Freling’s 2016 meta-analysis of 21 papers finds a larger overall effect of leniency on purchase proclivity than on return proclivity, with different effects across time, money, effort, scope and exchange. Those dimensions show why a generous policy has more than one switch. The evidence combines study settings and broader purchase measures; additional-purchase profitability and a current merchant’s optimal policy remain unmeasured. Research and limitations.

The commercial test is a customer-cohort contribution bridge: additional kept sales and later purchases, less induced return losses, service costs and abuse. That is outside the disposition workbook. Its nine product archetypes and seven condition states answer the narrower question of how to handle a return under declared assumptions. The two completed auction lots illustrate secondary transactions; they do not estimate a market or trace what later happened to the goods.

My judgment is that a good returns operation protects two assets at once: the recoverable object and the customer relationship that made the original purchase possible. Sometimes that means testing an expensive device carefully. Sometimes it means paying the refund and avoiding a useless journey. The relevant evidence is how the decision changes value.

The best route protects the item and the customer

I would retrieve the headphones because recovery pays for the work. For bulky or low-value goods, a local route or returnless refund can be the better choice. A good returns system earns value from decisions that preserve the recoverable object and the customer relationship, instead of maximizing the number of boxes brought back.