How to Evaluate a European TMS in 2026: Lane Design, Cost Models, Optimisation Engines and Human Oversight

Evaluating a European TMS is the process of testing whether the platform treats statutory rules as constraints inside its optimiser, treats cost as a decision variable and not a report, and can absorb new regulation by configuration, with explainability and human oversight built in. Most platforms look alike on the surface, so the differences that matter sit under the hood: how lanes are designed before drivers are contracted, how cost enters the model, what the engine does when a hard rule meets a saving, and how control divides between headquarters and the regions. Locus, the world’s first Decision-Intelligent, Agentic TMS, is built around those three questions.

Why the Evaluation Matters: The Business Case

Cost inputs move. The average price of diesel in France was €2.408 per litre on 21 September 2026, up 45.8% in a year, and French law revises road-transport prices with energy costs, so a static fuel figure goes stale.

Hard limits are specific. Regulation (EC) No 561/2006 sets nine hours of daily driving, a 45-minute break after 4.5 hours and 11 hours of daily rest, and a breach invites a fine, so the engine has to treat them as limits and not as costs.

Oversight is becoming law. Article 14 of the EU AI Act requires high-risk systems to be designed so that natural persons can oversee them, including understanding their limits, overriding or reversing outputs and stopping the system, and the obligations for Annex III systems apply from 2 December 2027.

Driver allocation may be in scope. Annex III covers AI intended to allocate tasks or monitor work performance, so whether a dispatch or allocation system is high-risk is a scoping question each operator must answer.

What is Different About Evaluating a TMS for Europe

Outsourcing is common and governed by each country. A business rarely holds all the resources it needs, and subcontractor rules differ by country, so the model has to carry them.

Contracted workers come first. Workers contracted for set durations must be given work ahead of ad-hoc capacity, which makes the contracting and lane decisions linked.

Industry shapes the model. Retail needs differ from courier, express and parcel needs, and a delivery to a hypermarket takes longer with order volume, parking space, opening hours and vehicle type all acting as inputs.

Planning is often decentralised. When every region plans for itself, leakage happens in places nobody measures and nobody owns the total.

Constraints change faster than software. Geography, union terms and restrictions change, so an engine that needs a development cycle for each change is already behind.

How to Evaluate a European TMS in Seven Steps

1. Ask where the rules live

Give each vendor a rule from your operation, such as a union agreement or a zone, and ask whether it is applied inside the optimiser or checked after it. This separates a plan you can run from one your planners rebuild by hand.

2. Design lanes before you contract

Redesign lanes against operational and business constraints, then run them as static routes alongside dynamic optimisation, with traffic calculated rather than assumed, road restrictions and one-ways avoided and each worker’s hours reflected. Forward-only deliveries may not need a return trip.

3. Check that cost is an input

Cost has several factors: route distance and time, owned-fleet resource cost, workforce cost including agency and subcontracted labour, and tractors and trailers. The model gives better results when cost goes in as an input parameter, with fuel surcharge refreshed and not held static.

4. Interrogate the engine

Ask what happens when a hard statutory constraint meets a cost objective, what is proprietary, how the engine performs on your own data and what happens to your roadmap when a needed constraint is missing. A penalty-based model and an infeasibility-based model diverge here.

5. Require explainability, oversight and traceability

Check that the platform can say why a route was generated or a carrier chosen, that humans oversee the decisions that matter while routine work is automated, and that every decision is time-stamped and geo-tagged. The line between automation and approval should be configurable per decision class.

6. Define what may be relaxed, and what may not

Run scenarios in parallel and compare cost and effect, with hard constraints identical across every scenario and only the tradeable layer moving. Driver working hours cannot be relaxed by adding buffers, and the breaks governed by hours-of-service rules cannot be touched.

7. Set the hierarchy of control

Plan and execute at region and area level while headquarters sets the rules and the cost basis, so only decisions that need to reach the execution team are passed down and everything else is decided centrally.

Rules Checked After vs Penalised vs Infeasible: Key Differences

What to Look for in a European TMS

Rules inside the model. Look for driver hours, zones and dock capacity applied during optimisation, with a clear answer on whether breaches are infeasible or penalised.

Cost as a decision variable. Check that fleet, labour, agency, trailer and fuel costs shape the plan and that fuel surcharge can be refreshed.

Absorbing change by configuration. Look for new cost drivers and regulations taken as configuration, not a vendor development cycle.

Explainability and oversight. Check for plan-level explanations, configurable autonomy, human approval and time-stamped, geo-tagged traceability.

Control hierarchy. Look for headquarters, regional and area levels that can each be modelled, with country-level constraints held separately.

Evaluation Criteria in Action: Real-World Results

Neither deployment is in Europe. They show what central rules with local execution deliver.

A global freight operator with 4,500+ drivers across 51 service centres. The operator ran 1,500+ captive drivers alongside 3,000+ third-party transporters. With centralised dispatch across both, the execution rate rose from 75% to 92%, uncovering $14M+ in annualised capacity across 25 sites. Read the freight operator case study.

A global food and beverage leader in Thailand and across Southeast Asia and MENA. Thailand alone runs 100+ distribution centres across 33+ cities serving 150,000+ outlets, with 18+ million orders a year and 5,000+ vehicle dispatches a month. With Locus’s dispatch, capacity, carrier and hub agents, it reached 97%+ SLA adherence, 22% lower procurement cost and 15% better rider time efficiency. Read the F&B leader case study.

Common TMS Evaluation Mistakes to Avoid

Judging on demo polish. On the surface every optimiser looks clean. The difference shows when a hard rule meets a cost objective on your own data.

Treating cost as a report. A cost number that was calculated after routing is not a number the optimiser worked to reduce.

Letting every region plan for itself. Leakage hides where nobody owns the total, so decide what is controlled centrally before rollout.

Treating human oversight as a checkbox. Article 14 asks that people can understand, override and stop a system, and Annex III dates are fixed, so the platform you choose in 2026 is the one running when they arrive.

How Locus Answers the Three Questions

Locus, the world’s first Decision-Intelligent, Agentic TMS, was recognized by Gartner for seven consecutive years and appears in the 2026 Hype Cycle for Supply Chain Execution and Logistics Technologies. It is named a Leader in the QKS Group SPARK Matrix for Transportation Management Systems and holds the #1 position in Route Planning in the G2 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.

On the first question, Locus models 250+ real-world operating constraints in one decisioning layer across owned fleet, contracted carriers and purchased capacity, and its engine treats statutory limits as infeasible, not as a penalty. On the second, cost enters the model as an input, and a new cost driver is taken as configuration. Locus reports 11 years of core algorithm development and 10+ patents on routing algorithms, figures it states are internal and not independently audited. On governance, plan-level explainability, human oversight and traceability are native, the autonomy level is configurable per decision class, and scenario design with hard and soft constraints lets teams run scenarios in parallel, compare cost and effect and choose the plan that goes out. Country, region and area logic can each be modelled, with execution controls sitting with operations teams. Six governance mechanisms keep teams in control. Locus has optimized 1.5B+ deliveries for 360+ enterprise customers across 30+ countries. The constraint data still has to be right, and that sits on the operator’s side. To see how Locus answers your three questions, schedule a demo.

For leaders asking how to evaluate a European TMS, Locus gives one governed decision layer where rules sit inside the model, cost is an input and new regulation is a configuration, with every decision explained and overseen.

Frequently Asked Questions

What should you ask when evaluating a European TMS? Ask three questions: do the rules sit inside the optimisation model or are they checked around it, does cost enter as an input or get applied to a fixed route, and can the engine absorb the next regulation by configuration. Add explainability, oversight and control hierarchy.

Why design lanes before contracting drivers? A lane designed first can be matched to the contract that suits it. Contract first and the lane has to fit a contract already signed. Designed lanes can run as static routes alongside dynamic optimisation.

What is the difference between treating a constraint as a penalty and as infeasible? A penalty puts a price on breaking a rule, so a large enough saving can outweigh it. Infeasible means the plan is not allowed. Driver hours are statutory, so the second approach is the one that holds.

Why should cost be an input to the optimiser? Cost applied after routing only reports the plan. Cost inside the model, including fuel surcharge refreshed over time, changes which plan is chosen. The difference is whether the number changed the plan.

What does the EU AI Act require for human oversight? Article 14 requires high-risk systems to be designed so people can understand their limits, monitor them, override or reverse outputs and stop them. Annex III obligations apply from 2 December 2027, and operators must assess whether their driver allocation is in scope.

How should a multi-region business control planning? Headquarters sets the rules and cost basis, regions handle local constraints and area teams execute with only the exceptions they need. Without that hierarchy, each region plans for itself and nobody owns the total.