How Locus Agentic TMS Improves Returns Experience in 2026

Returns experience is what a customer perceives between deciding to send something back and the refund arriving, and almost all of it is produced by transport rather than by the returns portal they interact with. The portal issues a label and shows a status; the collection window, whether anyone arrives, how long the item takes to reach a receiving site and when the refund clock can honestly start are all decided in the transportation layer. That is why returns programs run on a point solution tend to improve the interface without moving the experience. Locus, the world’s first Decision-Intelligent, Agentic TMS, plans collections in the same system and the same plan as deliveries, solving against more than 250 real-world operating constraints, which is where each of those four moments is actually decided.

Why Returns Experience Is a Transport Problem: The Business Case

Returns now influence the purchase rather than following it. The National Retail Federation’s 2025 Retail Returns Landscape, produced with Happy Returns, found 82% of consumers cite free returns as a major consideration when buying, up from 76% the previous year, against returns running at 15.8% of annual retail sales, worth $849.9 billion, and 19.3% of online sales. The experience is part of the product.

Delivering that experience has become more expensive at the same time. ATRI’s 2026 analysis puts the industry-average operating cost at $2.336 per mile in 2025, up 3.4% and the highest in the report’s history, so the dedicated trips that make a returns program responsive are also what make it unaffordable.

The way out is density, which is documented on the forward side. McKinsey finds that raising drops per stop from one to five cuts labor and vehicle cost by more than 50%. A collection made from a vehicle already working the street inherits that density rather than generating its own, which is the mechanism behind Locus data showing standalone return pickups at $4 to $6 per item against under $1.50 when the collection joins an existing route.

The carrier landscape underneath has also fragmented. The Pitney Bowes Parcel Shipping Index records non-major carriers doubling their share of US parcel revenue from 3.4% to 7.2% in a 23.1 billion parcel market, which adds capacity and adds reconciliation work at the same time.

Also Read: AI in Reverse Logistics: Turning Returns into a Competitive Advantage

The Five Moments of a Returns Experience

A customer experiences a return as a sequence of moments rather than a process. Each one is decided somewhere, and only the first is decided in returns software.

1 Being offered a collection window that the network can actually keep

The first moment is the customer choosing a slot. In most architectures the windows come from a calendar constrained by operating hours, because the returns platform has no view of whether that postcode will carry enough volume that day to justify a vehicle. The promise is made first and the economics discovered afterwards.

Locus computes collection feasibility against the network as it will actually run, in the same way its order management capability constrains delivery slotting to windows the network can fulfill rather than every window the calendar permits. Applied to the reverse leg, that means the slot a customer is offered reflects a route that exists rather than a hope that one will.

The customer-facing difference is small and compounding. A window offered from capacity is one the operation intends to keep, so the follow-up communication is a confirmation rather than an apology, and the customer’s next interaction with the brand is not a complaint. A window offered from a calendar is a coin toss that the retailer has already announced the result of.

2 Someone actually arriving, and the operation knowing if they did not

The second moment is the collection itself, and its failure mode is specific to returns: when nobody is home, the item stays with the customer and the refund clock keeps running. A failed delivery returns to a depot; a failed collection leaves the customer holding goods they have already mentally sold back.

Because Locus dispatched the collection, the attempt is an event it generated rather than one it is waiting to be told about. The Dispatch and Customer agents raise a failed attempt as an exception with a re-booking path attached, rather than leaving a silence that the customer discovers before the operation does.

That ordering matters more than it sounds. In most returns architectures the customer is the detection mechanism for a failed collection, because they are the only party who knows the driver did not come. An operation that learns about its own failures from the people affected by them is always one step behind the conversation, and every contact starts from a position of having to be told something it should have known.

3 The item moving, cheaply enough that the program survives

The third moment is invisible to the customer and decides whether the other four are affordable. Collections planned on their own have to generate their own density and mostly cannot. Collections planned into routes that already exist inherit it.

This is the core of what an agentic TMS does differently. Locus solves forward and reverse work in one plan, so a collection is evaluated as an insertion into a vehicle already serving that street, against remaining driver hours, vehicle capacity and the delivery windows already promised to the stops behind it. That is the difference between the published $1.50 and the published $5-plus. The comparison that matters for the decision is not against a perfectly planned route, which an insertion always degrades slightly, but against sending a vehicle for one collection, which it beats comfortably.

The experience consequence follows from the economics rather than sitting alongside them. A program paying dedicated-trip prices cannot afford to collect often, so it batches, and batching is what produces the long waits customers describe as the worst part of returning something. Cheap collection is what makes prompt collection sustainable.

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

4 Being able to answer where it is, and when the money comes

The fourth moment is the silence after handover, and it is where most returns experiences fail. A timeline assembled from carrier feeds can only show states every carrier emits, which in practice means the movement in the middle and neither end. The customer’s questions are about the ends.

Locus generates the timeline from its own execution record instead. The booking is a dispatch record, the assignment is a route decision, the attempt is a driver event, and the arrival is a receiving scan, so those states exist whether or not a carrier reports them, and a carrier feed contributes detail to a timeline that already exists rather than being the whole of it.

5 A refund promise the operation can actually keep

The fifth moment is the refund, and the number a retailer can advertise is set by variance rather than by average speed, because the promise has to cover the large majority of cases rather than the typical one. The two transport legs, booking to collection and collection to receipt, carry most of that variance.

Locus works on both. Collections placed on existing routes shorten the booking-to-collection leg without the cost of a dedicated trip, and because event time is recorded separately from report time, a refund clock measures when the item actually moved rather than when a file was uploaded.

The commercial value of that is the promise itself rather than the days. A retailer that knows its own distribution can state a refund window it will keep at a stated confidence, which is a different exercise from picking a number in marketing and asking operations to hit it. The first produces a promise slightly longer than the optimistic one and very rarely broken; the second produces the opposite, and customers experience a broken short promise far worse than an honest longer one.

6 The eight agents that run the lifecycle

Locus’s DiSCO framework assigns the returns lifecycle across eight named agents rather than one monolithic engine. Capacity models what the fleet and partner network can absorb. Dispatch places collections into routes and re-solves as conditions change. Carrier allocates across owned, contracted and on-demand capacity. Hub handles the receiving end. Customer owns notifications and the promise shown to the person waiting. Settlement closes the financial loop with partners. Orchestrator coordinates across them, and Mycroft, the AI co-pilot, gives operators a way to interrogate what happened and why.

The practical value of separating them is that a returns exception rarely belongs to one function. A missed collection is a dispatch event, a customer event and potentially a settlement event, and handling it well requires all three to see the same state. In a split architecture each of those three lives in a different system and the reconciliation happens in someone’s inbox, which is why missed collections tend to resolve slowly and expensively rather than badly.

The division also matters for what gets automated. A single engine has to be trusted or not trusted as a whole, which in practice means it is configured to the most cautious decision it makes. Separate agents can be governed separately, so the routine work runs at full speed while the decisions that warrant a human keep one.

7 Where the architecture is governed

An agentic system that makes decisions without oversight is an operational risk rather than an advantage. Locus applies six governance mechanisms across the framework: explainability so a decision can be interrogated, traceability so it can be reconstructed, evaluation so performance is measured against outcomes, autonomy levels so the degree of independence is a setting rather than an assumption, an execution sandbox for testing changes, and human review for the decisions that warrant it.

For returns specifically, the useful one is autonomy levels. Re-booking a failed collection is a low-stakes decision worth automating; issuing a refund against an unverified item is not, and the two should not run at the same level.

Traceability matters for a second reason in returns. Disputes about whether a collection was attempted are common, hard to resolve and expensive when they escalate, because the two accounts are a driver’s and a customer’s with no record between them. A dispatch-generated attempt event with a timestamp settles most of them before anyone has to adjudicate.

Agentic TMS vs Point-Solution Returns Platform: Key Differences

Dimension	Point-solution returns platform	Locus agentic TMS
Collection window offered from	A calendar	Network capacity and density
Collection planning	Standalone routes	Insertion into existing delivery routes
Published cost per pickup	Dedicated-trip economics	Under $1.50 when consolidated
Pre-collection states	Absent from carrier feeds	Generated as dispatch output
Failed attempt	Silence until someone notices	Exception with a re-booking path
Fleet	Separate from delivery capacity	Same fleet, same plan
Governance	Not applicable	Six mechanisms including autonomy levels

Also Read: Reverse Logistics Explained: Process, Types and Benefits in 2026

What to Look for in a Returns Experience Platform

Collections planned with deliveries, not after them. The cost difference between consolidated and standalone collection is roughly threefold, and it is only available to a system that holds both flows in one plan. Ask to see a collection added to a delivery route.

States generated rather than inherited. Booking, assignment and attempt are the events customers ask about and carriers rarely send. Confirm the platform produces them itself.

Feasibility checked before a window is offered. A slot the network cannot serve is a failure scheduled in advance. Ask whether the system can decline or reschedule a window it cannot keep.

Event time recorded separately from report time. A refund clock started on a carrier timestamp measures the paperwork. Confirm the platform distinguishes the two.

Autonomy that can be set per decision class. Re-booking and refunding should not run at the same level of independence. Ask how autonomy is configured and what sits under human review.

One fleet rather than a returns fleet. If collections are served by capacity held separately from deliveries, both pools carry their own buffer and neither reaches the consolidated cost. Ask whether the same vehicle can carry a delivery and a collection on the same route.

Returns Experience in Action: Real-World Results

A Canadian grocery brand delivering fresh and perishable orders across more than 30 cities ran its last mile through contracted third-party fleets, which meant the customer’s experience depended on partners the brand did not directly control, with collection exceptions generating support contacts at a far higher rate than deliveries. In the carrier orchestration deployment, moving allocation and tracking into a single orchestration layer produced 33% faster deliveries, 15% lower fulfillment cost, 25% less manual shipping time, and customer support resolution 10 to 20 times faster. That last figure is the returns-experience result stated in operational terms: contacts resolve quickly when the answer is already in the system rather than being assembled from several feeds.

A global FMCG manufacturer operating across 10 Asian countries ran distribution through more than 1,000 distributors and over 5,000 riders, with collection and delivery planned by different parties on different cycles. In the logistics automation deployment, consolidating planning into one decisioning layer saved more than 12,000 trips per month, optimized over $4B in orders, reached 1.8 million retail outlets and returned 3X ROI. Trips saved is the direct measure of collections and deliveries being planned together rather than separately.

Common Returns Experience Mistakes to Avoid

Improving the portal and expecting the experience to follow. The portal owns one of the five moments. The other four are transport outputs and will not move because the interface improved.

Offering collection windows the network has not agreed to. A slot issued from a calendar rather than from capacity is a broken promise scheduled in advance, and the customer still has the goods when it breaks.

Letting a failed collection stay silent. Unlike a failed delivery, the item remains with the customer and the refund clock keeps running, so the cost of the silence compounds.

Running one autonomy level across the returns lifecycle. Re-booking a missed collection and releasing a refund carry different risk, and treating them identically either slows the first or exposes the second.

Measuring the returns program on cost per pickup alone. Cost and experience are produced by the same decision, so a program optimized only on unit cost will batch collections and lengthen waits, and one optimized only on speed will dispatch dedicated vehicles it cannot afford.

Also Read: Top Reverse Logistics Companies and Software Solutions of 2026

Why Locus for Returns Experience

Locus, the world’s first Decision-Intelligent, Agentic TMS, runs the reverse leg on the same platform, the same fleet and the same plan as the forward network. The route planning system places collections into routes that already exist rather than building separate ones, which is what makes frequent collection affordable, and because Locus planned the work it generates the booking, assignment and attempt events rather than waiting for a carrier to report them. The eight DiSCO agents divide the lifecycle across capacity, dispatch, carrier allocation, hub handling, customer communication and settlement, coordinated by the Orchestrator, with six governance mechanisms including configurable autonomy levels so that low-risk decisions run unattended and high-risk ones do not. The platform operates at 1.5B+ deliveries for 360+ enterprise customers across 30+ countries at 99.99% uptime, against more than 250 real-world operating constraints.

Locus is recognized by Gartner for seven consecutive years, featured in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies, named a Leader in TMS by QKS Group in the SPARK Matrix, and ranked #1 in Route Planning on G2’s 2026 Best Software Awards. In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Locus continues to operate independently.

A returns experience is five moments, and four of them are decided in transport rather than in returns software: the window a customer is offered, whether a vehicle arrives, how cheaply the item moves, and whether the operation can answer where it is and when the refund lands. An agentic TMS changes each one because it holds the collection in the same plan as the delivery, which is what turns a dedicated trip costing $4 to $6 into a consolidated one under $1.50 and what allows the booking and attempt events to exist at all. Locus does that across eight named agents with governance applied per decision class, at enterprise scale. Request a Locus returns experience assessment to see which of the five moments your own program currently controls.

Frequently Asked Questions

What is an agentic TMS in the context of returns?

An agentic TMS plans and executes transportation decisions autonomously within configured guardrails, rather than presenting options for a planner to choose between. For returns that means the system decides whether a collection can be served, which vehicle takes it, when to re-book after a failure and what the customer is told, inside the same plan that runs deliveries.

How does Locus reduce the cost of a return pickup?

By placing the collection into a route that already exists rather than dispatching a vehicle for it. Locus data puts standalone return pickups at $4 to $6 per item against under $1.50 when the collection joins an existing route, because the consolidated collection inherits density from the forward network instead of having to create it.

Why can Locus show return states that carriers do not report?

Because it produced them. The collection booking is a dispatch record, the assignment is a route decision and the attempt is a driver event, all generated by the system that planned the work. A carrier has no commercial reason to emit those events, which is why timelines assembled purely from carrier feeds have gaps at both ends.

Which Locus agents are involved in a return?

Capacity models what the fleet and partner network can absorb, Dispatch places and re-solves the collection, Carrier allocates across owned, contracted and on-demand capacity, Hub handles receiving, Customer owns the notifications and the promise, and Settlement closes the financial loop with partners. The Orchestrator coordinates them, and Mycroft, the AI co-pilot, lets operators interrogate what happened.

How is autonomy controlled for returns decisions?

Through configurable autonomy levels, one of six governance mechanisms in the framework alongside explainability, traceability, evaluation, an execution sandbox and human review. In practice that means a low-risk decision such as re-booking a missed collection can run unattended while a higher-risk one such as releasing a refund on an unverified item sits under review.

Does Locus replace a returns management platform?

Not necessarily. Locus runs the transport layer of the returns lifecycle and integrates with the order management, warehouse and finance systems that own the rest. The distinction that matters is which system decides when the collection happens and what it costs, because those decisions produce most of what the customer experiences.