US Returns Hit $850 Billion in 2025: Why US Retailers Are Restructuring Reverse Logistics in 2026

Heading into 2026, US retail and e-commerce logistics leaders are navigating one of the most concentrated reverse logistics restructuring cycles in recent retail history.

The latest annual benchmark — the 2025 Retail Returns Landscape data from the National Retail Federation and Appriss Retail — quantifies the scale of the problem retailers are now working through: consumers were expected to return nearly $850 billion in merchandise bought in 2025, with an average return rate of 15.8% across all retail.

The number that matters for 2026 operations is not only the total value of merchandise coming back. It is the structural relationship between e-commerce growth, return volume, transport cost, fraud exposure, and customer retention.

US retailers are trying to do two difficult things at once: grow the digital channels that produce the highest return intensity, while reducing the cost, speed, and margin impact of those same returns.

For Locus, this is where reverse logistics becomes a routing and dispatch problem. Returns are no longer simply about policy, labels, and refunds. They are about deciding the right path for every returned item: customer pickup, store drop-off, locker, Return Bar, parcel carrier, 3PL hub, processing center, refurbisher, liquidator, recycler, donation partner, or store shelf — while protecting delivery SLAs, controlling cost-to-serve, and reducing avoidable miles.

What is AI returns routing in the US?
AI returns routing in the US uses machine learning, optimization algorithms, and real-time operational data to decide the lowest-cost, fastest, most capacity-aware, and most risk-appropriate path for returned goods across home pickups, stores, lockers, Return Bars, 3PL hubs, parcel carriers, processing centers, and disposition partners.

For a deeper look at the broader reverse logistics strategy, see AI in reverse logistics for retail returns optimization.

What AI Returns Routing Actually Does

AI returns routing is not a single model or dashboard. It is an operational decision layer that connects return intent, inventory condition, transport capacity, facility constraints, fraud risk, customer policy, and disposition economics.

In practical terms, AI returns routing answers six questions for every return:

1. Should the item be collected, dropped off, shipped, exchanged, kept by the customer, or routed through a partner network?
2. Where should the item go next: store, DC, return center, refurbisher, liquidator, recycler, donation partner, or vendor?
3. Which mode should move it: owned fleet, parcel carrier, 3PL, gig fleet, store transfer, courier pickup, or linehaul?
4. When should the return move, given customer SLAs, facility capacity, labor availability, and route density?
5. What level of verification is required, based on fraud risk, SKU value, customer history, and return reason?
6. How can the return be combined with forward delivery, store replenishment, or consolidation flows to reduce marginal cost?

This is why AI returns routing is different from a basic returns portal. A portal captures the request. AI routing determines the economically and operationally correct path.

---

The 2025 Returns Reality Setting 2026 Operating Conditions

Aggregate scale: US consumers were expected to return nearly $850 billion in merchandise bought in 2025. That translates to an average return rate of 15.8% across all retail and 17.9% for online-only sales.

For every $1 billion in sales, the average retailer incurs $158 million in merchandise returns.

The 2026 operating implication is clear: returns are not a marginal cost. They are a first-order P&L line item, especially for retailers with high e-commerce exposure.

That pressure is not easing. The US Census Bureau reported that retail e-commerce sales reached $298.9 billion in Q1 2026, up 9.7% from Q1 2025. E-commerce also accounted for 16.2% of total retail sales in Q1 2026, up from 15.3% in Q1 2025.

For operations teams, the practical cost drivers are not abstract. They include:

• miles per return pickup;
• cost per return shipment;
• failed pickup attempts;
• driver hours consumed by low-density reverse stops;
• dwell time at stores, lockers, and consolidation points;
• processing cycle time at returns centers;
• refund and exchange SLA adherence;
• fraud investigation cost;
• inventory recovery value lost when items reach the wrong destination too late.

A returns strategy that does not address routing, dispatch automation, and capacity-aware orchestration will struggle to change the underlying cost curve.

---

Customer Experience Economics: Returns Are Now Part of the Purchase Decision

Returns policy directly affects conversion, loyalty, and repeat purchase behavior.

Salsify’s 2025 consumer research found that 74% of global shoppers say free returns influence their decision to purchase online. The same research found that 72% expect clear, easy-to-understand return policies when shopping digitally.

Price sensitivity also matters. 66% of shoppers have cut back on non-essential purchases and 52% have shifted to store-brand products, which increases scrutiny of delivery fees, return fees, restocking fees, and refund delays.

The 2026 implication: retailers that operate returns as a back-office function while investing heavily in front-office customer experience are misallocating attention. For a meaningful share of customers, the return is the post-purchase experience.

That does not mean every return should be free, instant, or collected from the doorstep. It means return options should be routed and priced intelligently.

A high-value customer returning a high-margin item in a dense metro zone may justify a different return pathway from a low-margin, high-risk SKU in a low-density rural route. AI returns routing makes those trade-offs operational rather than manual.

---

How AI Returns Routing Works

AI returns routing works by combining predictive scoring, optimization, and dispatch execution.

1. Data ingestion

The system ingests data from the returns portal, OMS, WMS, TMS, carrier systems, store systems, fraud tools, inventory platforms, customer data platforms, and last-mile dispatch systems.

Useful inputs include:

• SKU and category;
• item value and margin;
• return reason code;
• customer location;
• customer return history;
• fraud score;
• item condition where known;
• carrier rate cards;
• store and facility capacity;
• inventory demand by location;
• driver availability;
• route plans;
• service-level commitments;
• refund and exchange rules.

2. Scoring

AI models score the return across multiple dimensions:

• probability of fraud;
• likely item condition;
• resale value;
• best disposition path;
• transport cost;
• processing cost;
• customer value;
• SLA urgency;
• facility congestion risk.

3. Routing decision

The routing engine evaluates the viable paths: customer pickup, store drop-off, locker, Return Bar, parcel return, 3PL hub, DC, refurbisher, outlet, donation, recycling, liquidation, or keep-the-item.

The objective is not simply “nearest location.” It is the best economic and operational outcome.

4. Dispatch execution

The decision must then be executed through auto-dispatch logistics software, carrier assignment, driver route planning, store workflows, customer notifications, and facility intake.

5. Feedback loop

Actual outcomes — pickup success, transit time, dwell time, fraud result, resale value, processing time, and refund SLA adherence — should feed back into the model.

That feedback loop is what turns returns routing from static rules into a continuously improving operating system.

---

Retailer Response Patterns Now in Execution

US merchants entered 2026 with returns restructuring high on the operational agenda. The drivers are straightforward: rising return volume, higher labor cost, carrier cost pressure, fraud exposure, and increased customer expectations.

Retailers are responding in three broad ways.

1. Policy redesign

Retailers are narrowing return windows, charging for some returns, segmenting return policies by item type, restricting serial returners, and offering exchanges or store credit where appropriate.

Policy redesign can reduce abuse, but it does not solve the transportation and processing cost of legitimate returns.

2. Network redesign

More retailers are using stores, lockers, return bars, parcel carriers, 3PLs, and consolidation networks to reduce fragmented shipments and create more flexible return pathways.

This is where BOPIS and omnichannel fulfillment strategy intersects with returns. Stores are no longer only forward fulfillment points. They can also become return intake, exchange, resale, and consolidation nodes.

3. Execution-layer automation

Retailers are investing in dispatch automation, AI routing, fraud scoring, capacity-aware facility routing, and returns visibility.

For logistics leaders, the operational questions become:

• Which returns should go to store, DC, refurbisher, resale, recycler, donation, or liquidation?
• Which returns should move by parcel, courier, 3PL linehaul, gig fleet, or store transfer?
• Which return flows can be consolidated without breaching refund or exchange SLAs?
• Which facilities have the capacity to process specific categories today?
• Which customer segments require a faster or more controlled return experience?
• Which routes can combine forward deliveries and reverse pickups?

This is where dispatch automation and route optimization become material to the returns P&L.

---

Fraud and Behavior: Why AI Routing Must Consider Risk

Returns fraud is now too material to be handled only after the item arrives.

NRF and Appriss Retail report that the average retailer says 10.3% of returns are fraudulent. Pindrop Labs estimates that e-commerce return fraud may cost US retailers $23 billion annually, with 14% of e-commerce returns estimated to be fraudulent.

Costly behavior is not limited to organized fraud. Pindrop Labs found that 39% of US consumers report that they always or often buy multiple sizes or variants online intending to return at least one. The same research reports that 27% say they have wardrobed at least once in the past year.

The next step is connecting fraud intelligence to routing decisions.

A high-risk return should not always follow the cheapest route. It may require:

• in-person item verification at drop-off;
• driver image capture during pickup;
• routing to a facility equipped for inspection;
• delayed refund until item-level validation;
• exception handling before restock or resale;
• customer-specific return options based on risk and lifetime value.

AI returns routing should therefore optimize for risk-adjusted cost, not just distance.

---

AI-Based Returns Routing vs Rule-Based Returns Workflows

Traditional returns workflows rely on static rules: send apparel to one DC, electronics to another, store returns to store inventory, parcel returns to the original warehouse, and high-value items to inspection.

Those rules are easy to understand, but they break down when volume, capacity, fraud risk, and customer expectations change in real time.

Dimension	Rule-based returns workflow	AI-based returns routing
Decision logic	Static rules based on SKU, channel, or return reason	Dynamic optimization using cost, capacity, risk, SLA, inventory, and route data
Facility selection	Often defaults to original DC or nearest facility	Routes to the best destination based on available capacity and recovery value
Fraud handling	Separate post-return review process	Fraud score can influence pickup, drop-off, refund, and inspection path
Transport mode	Predefined carrier or channel	Selects owned fleet, parcel, 3PL, gig, store transfer, or consolidation dynamically
Forward/reverse coordination	Usually planned separately	Can combine reverse pickups with forward routes
Scalability	Works for simple flows; struggles with exceptions	Better suited for high-volume, multi-node, omnichannel networks
KPI impact	Limited optimization beyond compliance with rules	Targets cost per return, dwell time, SLA adherence, recovery value, and route utilization

For a deeper comparison of dynamic optimization and static planning, see the AI vs rule-based route optimization benchmark.

---

Consolidation Networks and the Return Drop-Off Shift

Third-party consolidation networks are becoming more important because they reduce the cost and complexity of fragmented parcel returns.

A visible US example is Happy Returns, a UPS company, which operates a nationwide Return Bar network where shoppers can drop off eligible returns without packaging or labels. Returns are consolidated, sorted, verified, and bulk-shipped through the network.

Operational benefits can include:

• reduced shipping cost through bulk consolidation;
• item-level verification at drop-off;
• lower packaging friction for shoppers;
• improved visibility before returns reach a DC;
• better alignment with consumer expectations for convenient returns;
• fewer one-off parcel shipments.

For retailers planning 2026 returns infrastructure, the strategic question is not simply whether consolidation networks are useful. The question is whether the operating model fits the retailer’s category mix, density, service promise, inventory strategy, and fraud exposure.

A consolidation network without intelligent routing can still create bottlenecks. Stores, lockers, Return Bars, and 3PL hubs need dispatch logic that accounts for:

• pickup frequency;
• parcel and tote volume;
• facility cut-off times;
• route density;
• carrier capacity;
• refund and exchange SLA requirements;
• congestion at processing centers;
• recovery value by destination;
• labor and dock constraints.

In Locus terms, the returns network needs one orchestration layer across forward and reverse flows — not separate planning tools for store transfers, parcel returns, home pickups, and 3PL movements.

---

The Six AI Returns Routing Mechanisms That Reduce Cost

AI-powered routing and dispatching affects returns cost across six operational mechanisms. For the optimization logic behind these decisions, see how AI route optimization works.

AI routing mechanism	Operational use case	Cost lever	Customer or SLA impact	KPI to track
Reverse pickup routing optimization	Home pickup returns across dense and semi-dense zones	Reduces miles per pickup, driver hours, and failed attempts	Better pickup reliability and tighter time-window adherence	Miles per pickup, pickups per route, on-time pickup rate
Drop-off network routing	Movement from stores, lockers, Return Bars, and 3PL hubs to processing centers	Improves consolidation and lowers linehaul cost	Faster refund triggers and lower dwell time	Dwell time, consolidation rate, cost per return
Returns-aware forward routing	Outbound routing that considers return probability	Positions high-return-risk shipments closer to reverse infrastructure	Faster exchange and return cycle time	Return cycle time, exchange SLA adherence
Multi-modal returns dispatch	Orchestration across parcel, courier, owned fleet, 3PL, gig, and store operations	Selects the right mode by density, urgency, and cost-to-serve	More consistent service options by geography	Mode cost, SLA adherence, exception rate
Capacity-aware processing dispatch	Routing returns to DCs, stores, refurbishers, liquidators, recyclers, or donation partners	Reduces backlog, storage cost, and processing delay	Faster resale, refund, or exchange completion	Processing cycle time, facility utilization, recovery value
Round-trip optimization	Batching returns pickups with forward deliveries	Reduces marginal cost of reverse stops	Expands return convenience without proportional cost increase	Reverse stops per forward route, route utilization, cost per stop

Reverse pickup routing optimization

When retailers offer pickup-from-customer returns, route consolidation across nearby pickups reduces miles per pickup and driver hours per route.

The key is not simply adding return stops to a route. The routing engine must account for:

• customer time windows;
• driver shift constraints;
• vehicle capacity;
• forward delivery commitments;
• pickup item size and handling needs;
• traffic and service time;
• missed-pickup risk;
• route profitability.

AI engines optimizing across both forward delivery and reverse pickup produce different cost profiles from systems treating them as separate problems.

Drop-off network routing

For consolidation models, routing between drop-off points — Return Bars, store-based returns, locker networks, and 3PL hubs — and processing facilities affects consolidation efficiency.

Capacity-aware routing across the network reduces dwell time and storage cost. It also helps retailers prevent one facility from becoming a bottleneck while another has available labor, dock capacity, or category-specific processing capability.

The operational question is not “Where is the nearest facility?” It is “Where should this item go today, given cost, capacity, demand, disposition options, fraud risk, and SLA requirements?”

Returns-aware forward routing

Routing systems that incorporate return probability into forward delivery decisions can position high-return-risk shipments closer to consolidation infrastructure or pre-position returns capacity.

This is relevant for categories such as apparel, footwear, electronics, and home goods, where return probability, margin, item condition, and resale timing materially affect recovery value.

A returns-aware forward routing model can consider:

• SKU-level return history;
• customer-level return behavior;
• geography and pickup density;
• exchange likelihood;
• store inventory and demand;
• parcel zone cost;
• delivery promise and forward SLA.

The objective is not to slow down delivery. It is to avoid designing forward routes that create expensive reverse flows later.

Multi-modal returns dispatch

US returns flow across courier pickup, drop-off consolidation, locker networks, parcel carriers, store-based returns, owned fleets, 3PL linehaul, and gig fleets.

Coordinating dispatch across these modes — rather than treating each as a separate stream — captures cost efficiencies through shared capacity and modal flexibility.

For example:

• dense urban returns may be batched into courier routes;
• low-density returns may move through parcel or scheduled store transfers;
• high-value electronics may require controlled pickup or verified drop-off;
• low-value goods may be routed to local liquidation, donation, or recycling channels;
• exchange-led returns may be prioritized for fast processing.

This requires a routing layer that can make mode decisions dynamically, not a static rule table that assigns all returns of a certain type to one carrier.

Capacity-aware returns processing dispatch

Processing facility capacity varies by day, season, category, labor availability, and inbound volume. AI dispatch routing returns to facilities with available capacity reduces dwell time, storage cost, and processing latency.

The disposition decision matters as much as the transport decision. A returned item may need to go to:

• a store for resale;
• a DC for restock;
• a refurbisher;
• a recommerce partner;
• an outlet;
• a liquidator;
• a recycler;
• donation;
• destruction where required.

AI routing should evaluate cost, time, recovery value, condition, demand, and risk before selecting the path. Routing everything back to the originating DC can be operationally simple and financially wrong.

For facility constraint planning, see capacity planning for omnichannel retailers.

Round-trip optimization

The structurally largest lever is batching returns pickup with forward delivery routes.

A driver completing forward deliveries in a neighborhood with a nearby returns pickup can execute both on the same route. This reduces the marginal cost of returns pickup compared with sending a separate vehicle, driver, or carrier movement.

Magnitude varies by operational density, customer base, and route structure. Dense metros, high online penetration, predictable return windows, and owned or controlled fleet capacity create stronger opportunities. Low-density rural lanes may require a different combination of parcel, scheduled pickup, and consolidation.

The honest framing for VPs of Operations: AI-powered routing affects returns cost through specific mechanisms, and the magnitude varies by operation. The 2026 evaluation question is which mechanisms apply to your operating model — and whether your routing platform can orchestrate forward and reverse flows together.

Benefits of AI Returns Routing for US Retailers

AI returns routing is valuable because it connects customer-facing policy with operational execution. The benefits are measurable across cost, speed, risk, and inventory recovery.

Lower cost per return

AI routing reduces unnecessary transportation, fragmented parcel moves, low-density pickups, and misrouted returns. It can consolidate pickups, choose lower-cost viable modes, and avoid sending items to facilities that cannot process them efficiently.

Better route utilization

When reverse pickups are added intelligently to forward delivery routes, retailers can increase route density without creating a separate reverse logistics fleet.

Faster refund and exchange cycles

Routing decisions that account for verification requirements, facility capacity, and destination readiness can reduce dwell time and shorten return cycle time.

Higher inventory recovery value

Returned inventory loses value when it sits in the wrong facility or arrives too late for resale. AI routing can send items to the location where they are most likely to be restocked, exchanged, refurbished, resold, or liquidated at the best recoverable value.

Stronger fraud controls

Fraud scores can influence route, refund timing, verification level, and facility destination. That turns fraud detection from a back-office review into an operational control embedded in the return path.

Better customer experience

Convenient returns do not need to mean uncontrolled cost. AI routing helps retailers offer differentiated return options by geography, customer value, item type, risk profile, and operational capacity.

---

Key Capabilities Retailers Should Look For in an AI Returns Routing Platform

Retailers should evaluate AI returns routing platforms against execution capability, not AI claims.

The platform should support:

• forward and reverse route optimization in one environment;
• owned fleet, 3PL, parcel carrier, gig fleet, store, locker, and consolidation network orchestration;
• capacity-aware routing across stores, DCs, return centers, refurbishers, and partner facilities;
• fraud-aware routing and refund decision support;
• dynamic mode selection based on cost, density, SLA, item value, and risk;
• customer time-window management for home pickups;
• driver constraints, vehicle capacity, service territories, and shift rules;
• store drop-off and pickup workflows;
• real-time exception management;
• integration with OMS, WMS, TMS, carrier systems, returns portals, fraud engines, and inventory platforms;
• KPI measurement across cost per return, dwell time, recovery value, pickup success, route utilization, fraud rate, and SLA adherence.

The core principle is simple: AI must be embedded where routing and dispatch decisions are executed. Otherwise, retailers end up with better analysis but the same fragmented operating model.

---

The 2026 Evaluation Framework for US Retail Logistics Leaders

Six questions for US retail VP Operations and E-Commerce leaders evaluating reverse logistics restructuring in 2026:

1. Where does our returns rate sit against NRF benchmarks — 15.8% overall and 17.9% for online-only sales — and is variance explained by category mix, customer segment, policy design, or operational underperformance?
2. Have we built our returns experience around current consumer expectation data, including the finding that 74% of global shoppers say free returns influence online purchase decisions, or are we operating on legacy assumptions?
3. Are we deploying fraud-aware returns routing given that NRF and Appriss Retail report 10.3% of returns are fraudulent? If not, what is our defensible alternative?
4. Are we evaluating third-party consolidation networks, store drop-off flows, lockers, parcel carriers, and 3PL hubs as part of one reverse logistics network?
5. Have we mapped the six AI routing mechanisms — reverse pickup optimization, drop-off network routing, returns-aware forward routing, multi-modal dispatch, capacity-aware processing, and round-trip optimization — against our operating model to identify which produce material cost impact?
6. Are we measuring returns experience as a customer retention and conversion lever, or only as an operational cost?

For operators, those questions should translate into a measurable KPI baseline.

KPI	Why it matters
Cost per return	Core cost-to-serve metric across transport, handling, processing, and exceptions
Miles per return pickup	Measures route efficiency for home collection models
On-time pickup rate	Indicates customer experience and dispatch reliability
Dwell time at drop-off or consolidation point	Identifies network bottlenecks and delayed refund triggers
Processing cycle time	Measures how quickly returned inventory can be refunded, exchanged, restocked, refurbished, or liquidated
Recovery value	Tracks how much value is retained from returned inventory
Fraud rate	Measures exposure and the effectiveness of risk controls
Consolidation rate	Shows whether returns are moving in efficient batches rather than fragmented shipments
SLA adherence	Connects reverse logistics execution to customer promise and refund/exchange timelines
Exception rate	Identifies avoidable manual interventions, failed pickups, missing scans, and routing errors

A practical implementation path should start with measurement, not model selection.

Retailers need order, return, inventory, carrier, facility, customer, and fraud data connected into one decision environment. That typically means integrating the returns portal, OMS, WMS, TMS, carrier systems, store systems, fraud tools, and last-mile dispatch platform.

A phased rollout can look like this:

• Phase 1: Baseline. Measure cost per return, miles per return, return cycle time, recovery value, fraud rate, dwell time, and SLA adherence.
• Phase 2: Pilot. Select one region, one return channel, one category, or one carrier mix where density and data quality are sufficient.
• Phase 3: Orchestrate. Connect forward delivery routes with reverse pickup and drop-off flows.
• Phase 4: Scale. Extend across owned fleets, 3PLs, parcel carriers, stores, lockers, and consolidation networks.
• Phase 5: Optimize continuously. Use route, capacity, fraud, and disposition outcomes to improve routing logic over time.

---

Governance, Compliance, and Customer Fairness

AI returns routing must be governed carefully because it can affect customer experience, refund timing, fraud review, and access to return options.

Retailers should define clear controls around:

• explainability: dispatchers and customer service teams should understand why a return was routed a certain way;
• human override: exceptions should be reviewable by authorized teams;
• bias monitoring: customer segmentation and fraud scoring should not create unfair or non-compliant treatment;
• privacy: customer behavior data should be handled according to applicable data privacy requirements;
• auditability: return decisions, route assignments, fraud flags, and refund actions should be logged;
• model monitoring: performance should be reviewed against cost, SLA, fraud, and customer complaint metrics.

The goal is not to let algorithms replace returns policy. The goal is to make policy operationally executable, measurable, and adaptable.

---

Why Choose Locus for AI Returns Routing

Locus is built for logistics execution, not only analytics.

For retailers, 3PLs, and omnichannel operators, the value of AI returns routing depends on whether decisions can be executed across real routes, real drivers, real stores, real carriers, and real facility constraints.

Locus helps operations teams connect forward and reverse logistics through:

• AI-powered route optimization;
• dispatch automation;
• dynamic carrier and fleet orchestration;
• pickup and delivery sequencing;
• capacity-aware planning;
• real-time visibility;
• exception management;
• SLA tracking;
• multi-network orchestration across owned fleets and partners.

This matters because returns optimization is not complete when a model recommends a destination. It is complete when the return is picked up, routed, verified, processed, and recovered at the right cost and within the promised timeline.

The Real Question for 2026

The 2026 returns landscape is not a marginal optimization problem.

It is an active restructuring window shaped by nearly $850 billion in merchandise returns, rising e-commerce penetration, customer intolerance for return friction, and material fraud exposure.

The strategic question is:

Given that returns are simultaneously a major operational cost burden and a customer retention lever — and given that AI-powered routing represents one of the primary cost-reduction levers available — are we restructuring reverse logistics infrastructure to capture both, or are we executing returns policy changes without the routing layer that determines whether those changes deliver economic outcomes?

For US retailers, 3PLs, and omnichannel operators, AI returns routing is not about replacing policy with algorithms. It is about making every return decision operationally executable: the right channel, destination, transport mode, timing, verification level, and facility — at the lowest sustainable cost-to-serve while protecting SLA adherence and customer lifetime value.

---

Sources Referenced

• National Retail Federation and Appriss Retail, 2025 Retail Returns Landscape data
• U.S. Census Bureau, Quarterly Retail E-Commerce Sales, May 2026
• Salsify, 2025 Consumer Research Report
• Pindrop Labs, The $23B Fraud Problem in Retail
• U.S. Chamber of Commerce, AI and retail returns
• Global Finance, AI return on investment