Urban Last-Mile Delivery Efficiency in North America: The Seven Metrics That Decide Cost Per Successful Drop in 2026

What urban last-mile delivery efficiency actually means

Urban last-mile delivery efficiency is the ratio of successfully completed drops to the full cost of achieving them inside a dense metro service area, measured across labor, vehicle, access, and failure cost rather than mileage alone.

The distinction matters because the cost driver changes with density. In suburban and rural networks, distance dominates, so routing improvements track closely to fuel and drive time. In dense metros, dwell time, curb access, building entry, and re-attempts dominate. The same fleet can post excellent miles-per-stop numbers and still lose money per drop. That is why national efficiency averages mislead operators running Manhattan, downtown Toronto, Chicago’s Loop, or Mexico City routes.

The stakes are set by how much of total cost sits in this leg. Capgemini Research Institute puts last-mile delivery at 41% to 53% of total logistics and shipping cost, depending on network and category. In a dense urban operation, that share sits at the upper end, which means metro efficiency is not a marginal optimization. It is where the majority of the controllable cost lives.

Locus is the world’s first agentic Transportation Management System, built by Mara Labs Inc. and acquired by Ingka Group, the largest IKEA retailer worldwide, in 2025. Locus has supported 1.5B+ deliveries for 360+ enterprise customers across 30+ countries, orchestrating 1,000+ carriers and modeling 250+ real-world constraints in its routing engine. Customers have collectively realized $320M+ in logistics cost savings, reduced 800M+ miles, and avoided 17M+ kg of CO2. Locus is a Leader in the QKS Group SPARK Matrix for Transportation Management Systems, holds the G2 #1 position for Route Planning software, appears in the 2026 Gartner Hype Cycle across AI-powered logistics categories, and its ShipFlex multi-carrier product is named a Representative Vendor in the 2026 Gartner Market Guide for Multicarrier Parcel Management Solutions. That represents seven consecutive years of Gartner recognition.

Why dense North American metros break standard efficiency math

Three structural conditions separate metro delivery from the rest of the network.

Access is the scarce resource, not road capacity. A route through midtown Manhattan or Vancouver’s West End is not slow because of traffic alone. It is slow because a driver circles for a legal loading position, walks a package into a controlled-entry building, and waits on a concierge or a locker bank. The best available evidence for this comes from the Urban Freight Lab at the University of Washington, which instrumented more than 1,800 real deliveries in Seattle and found urban commercial vehicles spend approximately 80% of daily operating time parked, with most of a delivery driver’s time spent outside the vehicle walking the final stretch to the customer. For cargo-cycle drivers, roughly 60% of time went to parking and walking against 40% driving. Planning systems that treat a stop as a coordinate with a flat service time cannot represent any of this.

Congestion is worsening and is now priced in several metros. INRIX’s 2025 Global Traffic Scorecard found US drivers lost 49 hours to congestion in 2025 at a cost of $85.8 billion in lost time, with congestion increasing in 88% of the 290 US cities analyzed. Chicago drivers lost 112 hours and New York drivers 102. Where congestion is priced rather than merely endured, the operating picture shifts again: MTA and New York State reported approximately 73,000 fewer vehicles entering the Manhattan congestion zone daily, an 11% reduction, with morning rush-hour crossing speeds up an average of 23%. For a delivery operator, that is a real cost line and a real productivity gain sitting in the same policy, and only a plan that models the charge and the speed gain together can price the trade correctly.

Labor models are mixed and legally fragmented. Most North American urban operators run some combination of employed drivers, contracted fleets, and gig capacity, with classification rules differing by state and province. Capacity is therefore not fungible. A route that is legal and economic for an employed driver under hours-of-service rules may be neither for a contracted courier, and the planning system has to know the difference before it assigns work.

Also Read: Last-Mile Delivery Efficiency in Dense Urban Areas: Why Standard Operational Playbooks Fail

Urban last-mile capability has moved through three generations: monitoring (GPS and proof of delivery), analytics (retrospective performance reporting), and orchestration (systems that sense conditions, decide, execute, and learn continuously). Locus operates in the third tier through its SDEL architecture, Sense-Decide-Execute-Learn, and its DiSCO agent suite.

The seven metrics that define urban last-mile efficiency

1. Stops per route hour, not stops per route

Route count and stop count flatter dense operations. Stops per route hour is the honest density measure because it captures dwell, walk time, and building entry, which the Urban Freight Lab evidence shows is where the majority of the stop actually happens.

Operators improving this metric are rarely driving faster. They are clustering stops so a single park serves multiple drops, and sequencing so the walk path inside a block is efficient. That requires the plan to know vehicle class restrictions, access windows, and service duration by location type. Locus models 250+ real-world constraints in its Fireworks routing engine, which is what allows a plan to respect a curb reality rather than assume one.

A caution on benchmarking. No research firm publishes credible stops-per-route-hour benchmarks by vertical. Every published figure traces to software vendors. Measure your own baseline and improve against it rather than chasing a number from a listicle.

2. First-attempt completion, segmented by building type

Aggregate first-attempt rate hides the problem. Segmented by building type, the gap between single-family and multi-unit controlled-access addresses is usually where the money is.

Improving it depends on recipient-side coordination rather than routing: accurate narrow ETAs, access instructions captured once and reused, and alternate-drop logic covering lockers, parcel rooms, and PUDO points. The Customer Agent in the Locus DiSCO suite handles the recipient interaction layer so the promise made to the recipient stays consistent with the plan the dispatcher is executing.

Also Read: Two Cities, Two Playbooks: How NYC and London’s Kerbside Rules Are Reshaping Global Urban Delivery

3. Re-sequencing latency

Every urban plan is wrong by mid-morning. The question is how fast the system notices and re-decides.

Latency here is the interval between a disrupting signal (a road closure, a delayed dock release, a driver running behind, a same-day order entering the system) and a revised, dispatched sequence. Static overnight planning with manual intervention produces latency measured in hours. Agentic orchestration compresses it to the sense-decide cycle.

The economics of closing that gap are well documented. McKinsey estimates that AI-driven, multi-constraint routing delivers 10% to 25% cost reductions against a static daily plan. In Locus, the Dispatch Agent and Orchestrator Agent re-plan continuously against live conditions rather than waiting for the next planning window.

Also Read: The Morning Plan Problem: Why US Last-Mile Networks Need Dynamic Resequencing

4. Driver and rider utilization across mixed labor models

Utilization has to be measured per labor type, then optimized across the pool.

The practical test: can the system assign the marginal stop to the cheapest legally eligible capacity available right now, accounting for hours-of-service limits, contract terms, vehicle class, and skill requirements such as cold chain or age-restricted handoff? The Capacity Agent evaluates available capacity across employed, contracted, and gig pools. The Carrier Agent handles allocation across 1,000+ carriers where third-party capacity is in the mix.

Fleet utilization benchmarks are another area where no credible external source exists. Berg Insight figures are not in public releases and the NPTC private-fleet survey must be purchased. Treat any vertical utilization benchmark you are shown as a vendor claim until proven otherwise.

Also Read: The Three-Workforce Fleet Reality: How Owned, 3PL, and Gig Drivers Actually Operate at Most Enterprises

5. Promise accuracy

Promise accuracy is the share of deliveries landing inside the window communicated to the recipient. It connects operational efficiency to commercial outcome, because a narrow accurate window reduces recipient absence, which reduces failure, which reduces cost per successful drop.

This is no longer a soft metric. McKinsey’s survey of more than 1,000 US consumers found speed fell from the number one delivery priority in 2022 to fifth by 2024, displaced by reliability and predictability, and that approximately 90% of consumers will wait two to three days when delivery is free and arrives within the stated window. Wide windows are not conservative. They are expensive, and they buy nothing the customer values.

Accuracy depends on ETA models that account for metro-specific dwell and access patterns rather than road-speed estimates. An ETA calculated from remaining distance in a market where 80% of operating time is spent parked will be wrong in a predictable direction.

6. Exception recovery cost

Most operators track exception volume. Fewer track what recovery costs.

The full accounting includes re-attempt labor and mileage, support handling, refunds or credits, and the opportunity cost of capacity consumed by rework. Two notes on measuring it honestly. First, no research firm publishes a defensible dollar figure for a failed delivery attempt. The widely circulated 17-to-18-dollar figures trace to software vendors, so build the cost from your own labor, mileage, and support inputs. Second, unproductive paid time outside a driver’s control is a real and measurable category: ATRI found drivers were detained at 39.3% of all stops in 2023, losing between 117 and 209 hours per year depending on sector, at a cost of $3.6 billion in direct expense and $11.5 billion in lost productivity. That is line-haul data, but the mechanism transfers directly to metro dwell.

Locus’s six governance mechanisms, Explainability, Traceability, Evaluation, Autonomy Levels, Execution Sandbox, and Human-in-the-Loop, make agent decisions auditable, which is how operators trace an exception back to the decision that produced it.

Also Read: The Hidden Cost Categories of Failed First Attempts in US Last-Mile Operations

7. Emissions and access compliance per drop

Emissions per drop is now an operating constraint in several North American metros rather than a reporting exercise, as low-emission and restricted-access zones expand and congestion charges apply.

Efficiency and emissions largely converge in dense networks: fewer miles per successful drop is both cheaper and cleaner. Locus customers have collectively reduced 800M+ miles and avoided 17M+ kg of CO2.

What top regional operators do differently

Regional operators do not beat national parcel networks on scale, and attempting to is a losing position. They win on three advantages that dense geography confers.

They buy density deliberately, concentrating volume in fewer ZIP or postal codes rather than spreading thin across a metro area, because stops per route hour improves superlinearly with drop density.

They sell service depth, offering narrow windows, appointment delivery, in-home or threshold service, and returns handling that national networks price as exceptions.

They run one decisioning layer across owned and contracted capacity instead of separate systems per labor model, which is what makes the marginal-stop assignment in metric four possible.

The operational model matters less than most evaluations assume. Gig networks, employed fleets, and hybrid models all work in dense metros. What separates outcomes is whether a single system can plan, dispatch, and re-decide across whichever mix the operator runs.

Deployment evidence from two North American operations

A Fortune 50 parcel and logistics provider. This operator runs one of the world’s largest multimodal freight forwarding networks, moving 1M+ freight shipments a year, with a 4,500-strong driver pool split across roughly 1,500 captive and 3,000 third-party drivers. Dispatch was fragmented: captive shifts ran zone-based routing while third-party carriers needed tendering and on-demand assignment, and no single tool unified the pool. On Locus, Orchestrator and Dispatch agents took over pickup, transit, and delivery decisioning against 250+ operational constraints, with Capacity and Carrier agents governing the full driver pool under one policy. Weekly execution rate moved from 75% to 92% across 51 active service-center locations, and a single-site capacity analysis surfaced $565K in unused capacity that scaled to $14M+ annualized across 25 sites. Full detail is in the Fortune 50 parcel provider case study. The mechanism is the one metric four describes: one decisioning layer across mixed capacity types.

A global lottery operator running US field services. This operation installs, maintains, converts, and repairs machines across 25+ states, with technicians carrying different skill sets, rosters shifting through the day, and separate state contracts, some carrying one-hour SLAs and $100+ per-hour liquidated damages. It is a useful analogue for urban delivery because the binding constraints are identical in structure: a three-way match of job, skill, and location, under time windows tight enough that a plan goes stale within the hour as urgency, traffic, and weather shift. On Locus, the Dispatch Agent assigns every case type through one engine while the Capacity Agent maintains the roster and re-optimizes against live conditions. Results included 20% lower SLA penalty risk, 18% lower fuel spend, and 15% less drive distance and time. See the field-service dispatch and scheduling case study for the constraint model.

Analyst validation

Locus’s position is externally validated across four independent evaluations. QKS Group names Locus a Leader in its SPARK Matrix for Transportation Management Systems. G2 ranks Locus #1 for Route Planning software. Locus appears in the 2026 Gartner Hype Cycle across AI-powered logistics categories. ShipFlex, the Locus multi-carrier orchestration product, is named a Representative Vendor in the 2026 Gartner Market Guide for Multicarrier Parcel Management Solutions. Together these represent seven consecutive years of Gartner recognition across research categories.

Selecting a last-mile platform for North American city operations

Five questions worth asking any vendor:

• Which access constraints can the routing engine represent natively, and which require workarounds?
• What is the system’s re-decisioning latency when a live disruption occurs mid-route?
• Can the platform assign work across employed, contracted, and gig capacity in a single optimization pass?
• How is recipient communication tied to the executing plan, so the promise and the plan cannot drift?
• When an automated decision produces a bad outcome, can the operator trace why?