2024 Benchmark of Route Optimization APIs
Since its creation, Kardinal has been developing optimization technology with a central objective: enabling accurate and reliable modeling of complex business constraints, specific to each industry and client.
This algorithmic approach, focused on operational realism, is driven by a strong belief: the relevance of a Vehicle Routing & Scheduling (VRS) engine is not solely defined by its computational performance, but by its ability to faithfully reflect real-world operational constraints.
To rigorously assess this capability—both ours and that of the leading VRS API providers—we conducted an independent and reproducible benchmark during the summer of 2024.
The goal: to quantitatively compare the functional coverage of eight VRS solutions based on a shared set of operational constraints and modeling engine capabilities.
Methodology
Data source: Only public API documentation was used, ensuring a fair and transparent evaluation.
Scope of analysis: 63 representative constraints and features, derived from real-world use cases, and supported by at least two providers to ensure comparability.
Scoring criteria: Explicit, documented, and API-usable support of each business constraint.
Vendors evaluated: Google, NextBillion.ai, Solvice, GraphHopper, Verso, Routific, and Mapbox — in addition to Kardinal.
This benchmark does not aim to compare algorithmic performance (speed or optimality), but to measure functional modeling capabilities — a key factor in demanding operational contexts.
Step 1: Building the constraint evaluation panel
The first step of the benchmark involved establishing a representative reference of business constraints supported by the main Vehicle Routing & Scheduling (VRS) APIs.
To do this, we conducted a comprehensive inventory of constraints explicitly documented in the public APIs of the evaluated providers. This work resulted in a complete database of the features available on the market.
To ensure comparability between solutions and focus on a meaningful common ground, only constraints or features supported by at least two providers were included in the final analysis scope (63 selected out of 87 identified). This selection helps exclude edge cases while maintaining a broad and relevant functional coverage.
The full list of selected constraints, along with their definitions, is available here.
Step 2: Classification of constraints by type
The constraints were grouped into categories corresponding to the main modeling axes of VRS issues:
Advanced optimization engine capabilities: This category refers to structural features of the optimization engine such as lexicographical optimization, cost modeling, continuous optimization, advanced predictive traffic management, or incremental optimization.
Optimization objectives: This category groups the different business objectives that the algorithm can manage.
Speed profiles: This category includes the variety of vehicle types that the algorithm can take into account.
Resource constraints: This category brings together the constraints applied to resources (vehicle/driver), such as capacity, time windows, or rest times.
Order constraints: This category refers to constraints related to mission orders, such as pick-ups/deliveries, skill requirements, etc.
Order sequencing constraints: This category addresses constraints that link multiple orders together (for example, sequencing orders in a specific chain).
Assignment constraints: This category groups constraints that link resources to orders (for example, the compatibility of a resource with specific deliveries).
Step 3: Classification of constraints by industry
The constraints/features have been grouped into categories corresponding to the main industry sectors requiring the use of VRS:
Waste transportation
Fresh produce delivery
Urban couriers
Parcel delivery
Long-distance transport / Full truckload (FTL)
Freight forwarding
Less-than-truckload (LTL) transport
Retail
Field services / On-site interventions
Results
Part 1: Comparative analysis of business constraint coverage by market players
Each engine was evaluated on a total of 63 business constraints covering the following areas: order constraints, resource constraints, optimization objectives, speed profiles, etc. The final score is based on the ratio of constraints supported by each solution.
Rank | Provider | # Constraints | % Coverage | Category | 1 | Kardinal | 72 | 94% | Premium |
---|---|---|---|---|
2 | Google | 44 | 70% | Premium |
3 | NextBillion.ai | 44 | 70% | Premium |
4 | Solvice | 41 | 65% | Advanced |
5 | GraphHopper | 40 | 63% | Advanced |
6 | Verso | 26 | 41% | Basic |
8 | Routific | 26 | 41% | Basic |
9 | Mapbox | 18 | 29% | Basic |
Analysis
The analysis highlights the gaps in functional coverage between the solutions and identifies the players best suited for complex use cases.
The following typology was defined:
- Basic VRS: < 50% coverage
- Standard VRS: between 50% and 70%
- Premium VRS: > 70%
Kardinal stands out with a significantly superior modeling capability, covering 94% of the business reference. This result demonstrates the technical maturity of its engine and its ability to handle multi-constraint cases with high operational complexity.
Part 2: Benchmark of solutions by constraint categories
This radar chart compares the main route optimization solutions according to 7 key categories of features and constraints (detailed in the Methodology section). Each axis represents a key dimension of an optimization engine, and the scores (ranging from 0 to 1) show the coverage rate of each provider in this dimension.
The radar chart below is interactive. By clicking on the names of the providers, you can add/remove them:
Evaluated dimensions
- Advanced engine capabilities: Continuous optimization, custom cost functions, lexicographic optimization, etc.
- Optimization objectives: Handling of multiple business-specific objectives.
- Speed profiles: Diversity and specificity of vehicle speed profiles.
- Resource constraints: Time windows, breaks, capacity limits, etc.
- Order constraints: Handling of pickup/delivery pairs, required skills.
- Order sequencing constraints: Enforced logical sequencing of stops.
- Assignment constraints: Complex rules linking resources and tasks.
Comparative summary
Kardinal: Market-leading engine across all dimensions, with scores close to 1.0. Outstanding performance on critical dimensions such as assignment rules, sequencing, and multi-objective optimization. Best suited for regulated or high-stakes logistics environments (e.g., waste management, sensitive deliveries).
Google: Strong in advanced features and sequencing logic, but less flexible in assignment constraints and fine-grained modeling.
NextBillion.ai: Well-balanced overall, though less customizable in multi-objective optimization.
Solvice: Specializes in lexicographic optimization. Partial support for complex constraints.
GraphHopper: Consistent but lags behind on advanced modeling capabilities.
Verso & Routific: Designed for simpler use cases. Limited performance on complex constraints.
Mapbox: Primarily a mapping solution, with minimal support for business-specific constraints.
Part 3: Analysis of each provider’s ability to meet industry-specific constraints
This section assesses each engine’s ability to handle industry-specific constraints. The scores (%) represent the share of sector-specific constraints supported within each domain.
Industry | Kardinal | NextBillion | Google | Solvice | GraphHopper | Verso | Routific | Mapbox |
---|---|---|---|---|---|---|---|---|
Waste transportation | 97.7% | 79.5% | 79.5% | 68.2% | 65.9% | 54.5% | 52.3% | 31.8% |
Fresh produce delivery | 91.3% | 71.7% | 65.2% | 76.1% | 65.2% | 54.5% | 52.3% | 31.8% |
Urban couriers | 95.5% | 70.5% | 75.0% | 75.0% | 72.7% | 50.0% | 47.7% | 34.1% |
Parcel delivery | 90.2% | 78.0% | 68.3% | 78.0% | 70.7% | 56.1% | 53.7% | 31.7% |
Full truckload transport | 97.3% | 81.1% | 78.4% | 67.6% | 70.3% | 56.8% | 54.1% | 32.4% |
Freight forwarding | 97.6% | 75.6% | 75.6% | 70.7% | 65.9% | 56.1% | 56.1% | 34.1% |
Less-than-truckload transport | 91.8% | 73.5% | 67.3% | 67.3% | 61.2% | 44.9% | 44.9% | 26.5% |
Retail | 90.7% | 72.1% | 67.4% | 79.1% | 69.8% | 53.5% | 53.5% | 30.2% |
Field services | 97.3% | 75.7% | 81.1% | 70.3% | 67.6% | 54.1% | 48.6% | 32.4% |
Analysis
Kardinal demonstrates outstanding cross-sector performance, with coverage exceeding 90% across all industries. It is the only provider to offer such a high level of adaptability.
- NextBillion.ai and Google form a solid duo, with consistent scores around 70–80%, though still below Kardinal across all use cases.
- Solvice performs well in certain markets such as retail, fresh product delivery, or parcel logistics, but falls short in more complex sectors like long-haul transport or waste management.
- GraphHopper shows average coverage, typically between 65% and 72%, without standing out in any particular vertical.
- Verso, Routific, and Mapbox score significantly lower, often below 55%, and even under 35% for Mapbox. Their ability to serve demanding business sectors is therefore very limited.