These three providers recur on weather-data procurement shortlists in our early-access cohort: the Met Office commercial service, Tomorrow.io (formerly ClimaCell), and DTN. Each has substantial market presence and publishes claims of strong accuracy and broad geographic coverage. Yet archived actuals from defined operational settings reveal much more distinct skill profiles than those materials imply.
This article covers Climingo's benchmark results for temperature, wind speed, and precipitation across UK and Central European markets as of early 2025. None of these providers commissioned the work. Climingo is paid by the operators, traders, and logistics teams using its platform, with no financial relationship to any benchmarked provider.
Why Independent Benchmarks Differ From Reported Accuracy
Provider accuracy claims usually rely on in-house validation datasets covering chosen regions, periods, and metrics. Prospective customers generally cannot audit them, and they seldom match a customer's market, target variable, or forecast horizon.
Our archived-data benchmark uses forecasts captured when each provider issued them, rather than retrospective hindcasts. We match those files with verified partner actuals for the same locations, variables, and time steps. Skill scores are normalised to a climatological baseline, enabling comparisons across regions with different forecast difficulty. The published method is exportable, so customers can recompute scores with their own tools.
The findings below come from that pipeline. They describe the regions, seasons, and market settings early-access users loaded into the platform. They are not universal rankings.
Met Office: Strong UK, Variable Wind
The Met Office runs highly capable NWP infrastructure. Its UKV, or UK Variable resolution model, operates at approximately 1.5 km horizontal resolution across UK land, giving temperature a structural advantage in complex terrain over global models at 10 km or coarser resolution.
Our benchmark found strong Met Office temperature skill for UK applications at 12h and 24h horizons. In UK energy-demand forecasting, especially during the 07:00 to 21:00 peak when temperature sensitivity is greatest, its temperature scores ranked among the strongest providers assessed in this market.
Wind results varied more. At open-sea sites in the central and eastern North Sea, the global model performed competitively where its marine boundary-layer representation is well tested. Results were less consistent in complex coastal terrain, including the Welsh coast and northwest Scotland, as expected when the resolution benefit is concentrated over UK land rather than offshore transition zones.
For UK precipitation, the Met Office was competitive without clear dominance. In UK operations where temperature is the main variable, its commercial service showed a genuine skill advantage in our data.
Tomorrow.io: Competitive Precipitation, Improving Probabilistic Skill
Tomorrow.io combines several NWP sources with proprietary machine-learning post-processing and inputs from its commercial microwave sensor network. This creates a performance profile distinct from single-model providers.
Our data showed Tomorrow.io performing competitively on Western European precipitation, including UK, French, and Benelux market areas. CRPS for its probabilistic precipitation outputs improved across successive API versions, indicating continued calibration work. For logistics and supply-chain teams using precipitation uncertainty to guide routing, the probabilistic product merits comparison with deterministic alternatives.
For large-scale energy use cases, temperature and wind results were more mixed, including day-ahead demand forecasts for large grid zones and offshore wind generation forecasts. In some Central European locations, we observed winter cold bias in temperature, a possible issue for ML-blended products with geographic gaps in training data. The bias varied by location and season, making operational correction more difficult.
At 0-6 hours ahead, precipitation skill was notably strong, consistent with the microwave network's ability to detect active cells. That short-range capability matters for logistics teams making same-day route adjustments.
DTN: Strong Short Range, Weaker Beyond 48 Hours
In our data, DTN's strength centred on operational weather: short-range wind and precipitation at 0-24 hours, with well-calibrated surface conditions for ground transport and construction. Its product line reflects an agriculture and transport heritage, where short-range accuracy is the main commercial need.
For day-ahead and intraday energy trading, the relevant 12-48 hour horizon makes DTN's short-range strength useful. Its 6h-ahead wind performance was consistently competitive in UK and Benelux data, including good calibration of extremes. For logistics decisions using 6-12 hour planning windows, DTN performed well on UK and northern France routes.
At 48-72 hours, DTN underperformed the other two providers in our data. That fits a product family not historically focused on extended medium-range forecasting. Organisations planning over 3-7 days, such as weekly shipping schedules or power-generation commitments to medium-term contracts, may find less differentiation at these horizons.
What the Regional Data Shows
Cross-provider analysis most clearly shows that no provider leads across every region and variable. Selecting one provider for all use cases can therefore leave value unrealised in some operational settings.
In our data, the Met Office leads UK temperature, Tomorrow.io is competitive for short-range Central European precipitation, and DTN is consistently strong for short-range operational wind in the UK and northern France. A UK energy-trading desk and a logistics operator running cross-Channel corridors face different provider choices, even with the same variables.
Procurement discussions often make one provider choice across variables, regions, and horizons. The data supports a multidimensional assessment, segmented by actual exposure. This is harder to run than one vendor contract, but context-specific skill differences are large enough to justify the added complexity.
An Important Caveat on These Results
This analysis has defined limits. The results come from Climingo's pipeline as of early 2025 and cover the variables, regions, and horizons loaded during early access. All three providers regularly change model configurations and post-processing, so skill profiles can shift, sometimes materially, after updates.
We do not claim the Met Office always beats Tomorrow.io on temperature, or that DTN is always weaker at 72 hours. These are measurements from defined operational conditions and a defined period. Offshore wind in the German Bight or inland logistics in Poland may produce different results. Use this analysis as a starting hypothesis, not a procurement verdict. Benchmarking each provider against your own actuals, market, and decisions is the only way to identify your winner.