From Static Route Plans to Continuous Re-Optimisation: A European Last-Mile Efficiency Benchmark

Why a European route plan starts decaying the moment it leaves the depot

Most European last-mile operations still plan once. Orders cut off, the optimiser runs overnight or early morning, routes are released, and vehicles depart against a sequence that represents the best available answer to a question asked before anything happened.

That plan is a forecast. It assumes the travel times used to build it, the vehicles available when it was built, the customers reachable at the hour it predicted, and the road network as it existed at the moment of computation. Every one of those assumptions begins expiring at the depot gate.

The operational consequence is not that the plan is wrong. It is that the plan is progressively wrong, and nobody measures the rate. Operations teams track on-time delivery and cost per drop as end-of-day outcomes, which tells them the plan degraded without telling them when, where, or how fast. By the time the numbers arrive, the day is over and the diagnosis is unavailable.

This is why efficiency programmes built on better planning so often disappoint. The optimiser is usually not the weak link. An operation can improve its morning plan by several percentage points on paper and see no movement in cost per drop, because the gains were computed against conditions that lasted ninety minutes. Investment goes into the quality of a forecast when the constraint is the absence of any mechanism to revise it.

A field-service deployment puts the pattern plainly. In one global operation running across more than 25 jurisdictions, the internal assessment was that even a well-built plan went stale within the hour as urgency, traffic, and weather shifted. That is the honest description of static planning at scale, and it holds in dense European urban delivery for the same reasons.

What makes European plan decay faster than the model assumes

Three forces compound, and all three are structurally stronger in European cities than the planning assumptions most operations inherited.

Congestion volatility, not congestion levels. The relevant variable is not how congested a city is but how unpredictably it varies. The INRIX 2025 Global Traffic Scorecard recorded London drivers losing 91 hours to traffic, with UK drivers losing 59 hours on average at a cost of £822 per driver and roughly £11 billion nationwide. Bristol recorded 64 hours, Manchester 62, Leeds 59, Birmingham 57. Notably, London delays fell 10% year on year while 62% of urban areas worldwide saw congestion increase and only 26% improved. The UK figures are the most granularly published in Europe, and the pattern they show is the point rather than the specific numbers: direction of travel differs by city. A planning model calibrated on last year’s averages is calibrated on a distribution that is still moving, in different directions, in each city you serve.

A fragmented and tightening access map. European urban access regulation is not one rule set but hundreds, each with its own boundary, vehicle standard, charging basis, and timetable. The European Commission’s urban vehicle access regulations framework covers a landscape in which 73% of UVARs are low or zero emission zones. The Clean Cities Campaign counted active low emission zones rising from 228 in 2019 to 320 in 2022, a 40% increase, with 27 cities set to expand or tighten existing zones and at least 35 planning zero-emission zones by 2030.

This matters for decay specifically because vehicle eligibility is not a cost to reconcile later. It is a hard constraint on which asset can serve which street. When a vehicle substitution happens mid-day, a breakdown, a driver absence, a capacity reallocation, the replacement vehicle may be ineligible for part of the route it inherits. A static plan has no mechanism to notice.

Narrow time windows with asymmetric penalties. A window missed early is usually recoverable. A window missed late cascades, because every subsequent stop inherits the delay. European operations increasingly commit to tight windows, and Deloitte research put first-attempt home delivery failure at 10% to 15% across markets including Spain, Germany and the UK. Each failure is both a direct cost and a decay event, because a failed attempt returns capacity and volume to a plan with no slot for either.

The asymmetry is what static planning handles worst. A plan built in the morning distributes its slack evenly, because at build time every stop is equally uncertain. By mid-afternoon the uncertainty is concentrated at the end of each route, where the slack has already been consumed and the remaining windows are the ones with the least room. An operation that cannot redistribute slack during the day is holding its buffer in the wrong place for most of the working day.

Also Read: Killing the Empty Mile: How Dynamic Routing is Decarbonizing European Supply Chains

The benchmark: measure decay, not plan quality

Most efficiency benchmarking compares plans. It should compare decay, because two operations with identical plan quality and different decay rates will post materially different cost per drop. Four measures do the work, and none requires a new system to start collecting.

Two notes on using these. Measure plan adherence at a fixed hour rather than end of day, because end-of-day adherence blends decay with recovery and hides both. And use cost per successful drop rather than cost per drop, since dividing by attempts rewards an operation for making more of them.

The distinguishing measure is re-optimisation latency, and it is where the Gartner finding bites. When 95% of supply chains must react quickly to change but only 7% can execute decisions in real time, the gap is not awareness. Most operations know a route is slipping. They cannot re-sequence the fleet and get the answer to a driver before the window closes.

Also Read: Route Optimization Software With Real-Time Dynamic Re-Routing: A 2026 Buyer’s Guide

What counts as a re-optimisation trigger

Not every event deserves a re-plan. Re-optimising on noise erodes driver trust, because a sequence that changes three times an hour stops being an instruction and becomes a suggestion. The discipline is deciding which events cross the threshold, and setting that threshold per service tier rather than globally. Four classes cover most of the operational surface.

Capacity triggers. A vehicle breakdown, a driver absence, a late depot departure, or a vehicle substitution. These are the most consequential in European operations, because a substitution can change zone eligibility. A capacity event silently becomes a compliance event, so any substitution should re-check the inherited route against the replacement vehicle’s access rights before the driver leaves the yard.

Demand triggers. An urgent order arriving after cut-off, a cancellation, a customer-initiated reschedule, or a same-day insertion. These change the shape of the problem rather than the conditions around it, and each is worth solving across the fleet rather than pushing to whichever driver happens to be nearest.

Condition triggers. Traffic incidents, weather, road closures, and access restrictions activating on their own schedule. These are the triggers most operations already watch and least often act on, because watching requires a data feed while acting requires a solver that can be called mid-day.

Execution triggers. A failed attempt, an unexpectedly long dwell at a stop, or cumulative drift against the planned sequence. Cumulative drift is the most useful and least monitored of the four. Three consecutive stops each running eight minutes late is not three small problems; it is a route that will miss its final window, and it is detectable well before it does.

The design decision is the threshold rather than the trigger list. A premium one-hour window justifies re-optimising on a five-minute variance. A next-day economy route does not. Applying one global threshold guarantees over-reacting on economy work and under-reacting exactly where the penalty sits.

Also Read: What is an Agentic TMS? A Practical Guide for Enterprise Logistics Leaders in 2026

Static planning and continuous re-optimisation compared

The row that decides the others is scope. Reassigning a single late stop to a single driver is a local repair that usually moves cost onto a neighbouring route, which is why operations that re-plan driver by driver often see no improvement in aggregate cost per drop. Fleet-level re-optimisation solves for the network, and it is the reason this is an architectural distinction rather than a feature comparison. A platform built to produce one plan and validate it afterwards cannot be configured into one that decides continuously.

Also Read: Cubic Meters, Not Parcels: Why European Furniture Retailers Need Volume-Constrained Routing Under CSRD

How Locus closes the plan decay gap

Locus, the world’s first Decision-Intelligent, Agentic TMS, is built around continuous decisioning rather than a daily planning run. Within DiSCO, the Dispatch agent plans and re-sequences against more than 250 real-world constraints per computation, including vehicle-to-zone eligibility, time windows, and driver hours, so a mid-day substitution is checked against the same rules that governed the original plan. The Capacity agent maintains the roster and its remaining-hours state, and the Orchestrator agent keeps the two aligned so a change in one does not silently invalidate the other. The cycle is Sense, Decide, Execute, Learn, which means the outcome of today’s decay informs tomorrow’s plan rather than only being reported. Six governance mechanisms bound autonomous action, including explainability, traceability, and human-in-the-loop override, so a re-sequencing decision can be explained to the operation that lives with it.

Locus has been 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 (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.

Two deployments show continuous re-optimisation against fragmented rule sets, which is the structural analogue of the European access map.

A global lottery operator runs a field-service operation installing, maintaining, converting and repairing machines across more than 25 US states. The rule fragmentation is severe: contracts, labour laws and revenue terms differ by state, some carrying one-hour SLAs with liquidated damages above $100 per hour, while six distinct job types each demand different technician skills, so every assignment is a three-way match of case, skill and location. Zones, schedule types, staffing models and standby time kept changing, and the operation’s own assessment was that even a well-built plan went stale within the hour as urgency, traffic and weather shifted. Locus models each state’s contracts, labour laws, SLA windows, zones and skills as live constraints rather than post-plan checks, routes cases to eligible technicians automatically with the tightest SLA windows protected first, and re-optimises against live traffic, weather and urgency with the Capacity and Dispatch agents coordinated by the Orchestrator. The operator reported SLA penalty risk down 20%, fuel spend down 18%, and drive distance and time down 15%, with more than 25 states running on one autonomous dispatch engine.

A Fortune 50 parcel and logistics leader demonstrates the same mechanism at fleet scale. Its 4,500-strong driver pool spanned captive fleets on zone-based routing and third-party carriers requiring tendering and on-demand assignment, with no single tool unifying them, so decay in one pool could not be absorbed by capacity in another. Locus ran pickup, transit and delivery decisioning as one chain across the network against 250-plus operational constraints per computation. Weekly execution across 51 active service-centre locations moved from 75% to 92%, and a single-site analysis surfaced $565,000 in unused capacity that scaled to more than $14 million annualised across 25 sites. The capacity was not created by the platform. It was already there, hidden inside plans nobody was re-deciding.

Also Read: 5 European Logistics Innovations Reshaping 2026: From Agentic TMS to GenAI Customer Service

Start by measuring how fast your plan dies

The useful shift for a European logistics leader is to stop asking whether the morning plan is good and start asking how long it stays good. That reframing changes what gets instrumented, what gets bought, and what gets held accountable.

Run the four benchmarks against one ordinary week, not a peak week. Capture plan adherence at hour four, time a real trigger event from occurrence to updated driver sequence, count how many at-risk stops were actually rescued, and recalculate cost per successful drop rather than cost per attempt. Most operations discover that adherence is lower than assumed by mid-afternoon, that latency is measured in tens of minutes rather than single digits, and that recovery is happening informally over the phone rather than systematically in software.

None of those findings is a planning problem. They are all decay-response problems, and they are invisible to a benchmark that only compares plans.

Request a Locus European last-mile assessment to baseline plan adherence, re-optimisation latency, recovery rate and cost per successful drop against comparable urban operations.