Time series forecasting
& anomaly detection

The open-source standard for time series.
Unit8 helps enterprises go from experimentation to production-grade forecasting systems.

darts · live forecast
0+
GitHub Stars
0M+
Overall Downloads
0+
Data & AI Specialists
0+
Unit8 Projects Delivered
Trusted By

Delivering results for industry leaders

Unit8 partners with enterprises across industries to turn data into competitive advantage.

Merck Selexis Firmenich Daimler CHUV Medgate Helvetia Swiss Re GridBeyond Alpiq Tecan IWB
Use Cases

Proven across industries

From demand planning to predictive maintenance, Darts powers mission-critical forecasting systems across sectors.

Wind turbines in an open landscape

Energy & Utilities

Electricity price forecasting, heat demand prediction, production scheduling, and renewable energy optimization. We help energy companies make smarter operational decisions and reduce costly balancing errors.

Grid Ops Heat & Power Renewables
Case Study Swiss Utility

Heat Demand Prediction for Power Plants

Inaccurate heat demand forecasts caused poor electricity production scheduling and high balancing fees. We deployed a Darts-based framework on Azure DevOps that cut forecast error from 15.8% to 5.7% MAPE, with predictions running 3,000x faster than the previous model.

3,000x
Faster
5.7%
MAPE (was 15.8%)
Case Study Energy Company

Electricity Price Forecasting

The internal trading team needed reliable price forecasts to improve decision-making. We built a backtesting pipeline, benchmarked multivariate configurations, and delivered probabilistic forecasts that reduced the deviation between predicted and actual prices by 25%.

25%
Lower Deviation
Reliable
Prediction Intervals
Cargo plane at an airport loading dock

Transportation & Logistics

Cargo demand forecasting, route optimization, and fleet management. We help aviation and logistics companies plan capacity and reduce costs with data-driven decision-making.

Aviation Logistics
Case Study Airline Trade Association

Air Cargo Demand Forecasting

The client needed granular cargo demand forecasts (KG and USD) to launch a new commercial offering. We delivered an end-to-end pipeline with Darts, Optuna tuning, and MLflow tracking, reaching parity with top-tier competitors in a 5-week PoC.

22.8%
WAPE (6-mo horizon)
5 wks
PoC to Parity
Warehouse with organized inventory shelving

Retail & Supply Chain

Demand and sales forecasting across millions of SKUs, inventory optimization, promotions planning, and stockout prevention. We help retailers and manufacturers turn historical sales data into actionable supply chain decisions.

Retail Supply Chain
Case Study Global Retail Manufacturer

Sales Forecasting for 10M Products

Finding the best forecasting method across 10 million store products was infeasible manually. We built a Darts-powered platform to automatically benchmark, tune, and select models at global scale, outperforming the existing solution in both accuracy and efficiency.

10M
Products Covered
+New
Use Cases Enabled
Modern factory production line

Manufacturing

Supply chain demand forecasting, production scheduling, and facility health monitoring. We help manufacturers replace manual planning with intelligent, data-driven procurement and operations.

Production Procurement
Case Study Medical Equipment Manufacturer

Supply Chain Demand Forecasting

Demand planners relied on manual back-and-forth between sales and manufacturing. We built a Darts model with hierarchical reconciliation, streaming live data from SAP, delivering 10-60% accuracy improvements and enabling procurement to negotiate better supplier contracts.

10-60%
Accuracy Gain
Better
Supplier Contracts
Industrial power plant machinery

Predictive Maintenance

Anomaly detection on physical variables like temperature and vibration to spot slow-shifting degradation months before failure. Shift from reactive shutdowns to planned, cost-effective maintenance.

Hydro Plants Manufacturing
Case Study Swiss Energy Company

Hydro-Electric Plant Monitoring

Slow-shifting anomalies in plant equipment were invisible to human surveillance. We built a predictive ML model with automatic threshold-based alerting. Engineers now detect abnormal changes months in advance and plan maintenance proactively.

Months
Earlier Detection
Reduced
Unplanned Downtime
Case Study Swiss Energy Company

Predictive Maintenance Feasibility

The client had no data analytics team and wanted to understand where predictive maintenance could add value. We delivered a replicable 5-step methodology, prioritized machinery parts, and validated the approach with a steam turbine mini-POC.

5-Step
Methodology
POC
Steam Turbine
Financial analytics dashboard on a screen

Finance

Cash flow forecasting, accounts payable and receivable prediction, and working capital optimization. We help finance teams replace spreadsheet-based projections with reliable, ML-driven financial outlooks.

Cash Flow Payables
Case Study European Enterprise

Outstanding Payables Forecasting

Manual monthly payables forecasting was error-prone and left cash flow unpredictable. We built an ML-based framework with Darts integrating ERP and payment history data, giving the finance team a reliable 15-month outlook for working capital planning.

30%
Less Forecast Error
15-mo
Reliable Outlook
Unit8 Darts Offerings

From proof of concept to
production-grade forecasting

Unit8 built Darts and actively maintains it. Our structured engagements help you unlock its full potential, from a first exploration to a fully operational forecasting system.

4–5 weeks

Forecasting Exploration

Discover what's possible on a brand-new forecasting problem

A lightweight, exploratory engagement for teams venturing into forecasting for the first time or launching a new use case from scratch. No existing baseline required. Works from a simple data export.

  • Forecasting feasibility assessment
  • Data quality review and preprocessing
  • Benchmark models trained on your dataset
  • Results report with a clear path forward

Works from a one-off data export. No live data pipeline or prior forecasting setup required.

5–6 weeks

Forecasting Improvement

Prove the value of modern forecasting on your own data

A focused, time-boxed engagement to assess your data, benchmark model architectures, and deliver an optimised forecasting model. Includes a clear roadmap for what comes next.

  • Baseline vs. Darts model comparison
  • Top architectures shortlisted and tuned
  • Documented code and results deck
  • Productionisation roadmap and next steps

Best for teams validating forecasting potential before committing to a full system.

9–12 weeks Recommended

End-to-End Forecasting

A production-grade forecasting system, built end to end

Starts with either a Forecasting Improvement or a Forecasting Exploration, then extends with your choice of add-on modules (MLOps, monitoring, downstream integration), delivered as one continuous engagement.

  • Everything in the Forecast Sprint
  • Retraining pipeline and model versioning
  • Performance monitoring and drift alerting
  • Forecasts pushed to downstream systems

Start with either sprint. At week 5, choose add-on modules to extend. No upfront commitment to the full scope.

Specialty Offerings

Predictive Maintenance

Purpose-built engagements for industrial anomaly detection and equipment health monitoring. Powered by Darts’ time series and anomaly detection capabilities.

2–3 weeks

Feasibility Assessment

Understand whether predictive maintenance is right for your plant

Before investing in a monitoring system, Unit8 conducts a structured assessment of your equipment park, data landscape, and failure history to determine the feasibility and expected value of a predictive maintenance approach.

  • Equipment inventory: types, operating conditions, and criticality
  • Analysis of common failure modes and historical breakdown statistics
  • Sensor data review: availability, sampling rates, and quality
  • Review of historical incident logs and maintenance records
  • Identification of highest-value monitoring targets
  • Deliverable: prioritised target list and recommended approach

Assessment findings can be directly applied to scope a Monitoring System engagement.

4–5 weeks · POC 9–12 weeks · Production

Monitoring System

From rapid prototype to production-grade anomaly detection

Unit8 builds a predictive ML model to monitor key equipment variables (temperature, vibration, pressure) and alert operators when values deviate from the expected normal operating range, enabling proactive maintenance before failures occur.

  • Normal operating range (NOR) modelling with causal signals
  • Probabilistic quantile-interval prediction as NOR proxy
  • Scheduled anomaly detection with configurable alert thresholds
  • Automatic alerts when values deviate from predictions
  • POC: single equipment type, validated on historical data
  • Production: multi-sensor deployment, live ingestion, dashboard

Start with a 4–5 week POC to validate the approach on your equipment data. Extend to a full production system if results confirm value.

Always-on Expertise

Support & Expert Access

Independent of any project. Available to any team using Darts, whether or not we built your system.

Expert Support Retainer

Ongoing · quarterly or yearly

Continuous access to Darts core developers and forecasting experts. Troubleshoot, get code reviews, architecture guidance, and upgrade support on demand.

Starter
10 h / quarter 24h response
Professional
40 h / quarter 24h response
Premium
80 h / quarter 4h priority
Troubleshooting Code Review Architecture Guidance Upgrade Support Performance Tuning

Expert Sessions

Ad-hoc · 1–4 hours · no commitment

Ad-hoc access to Darts expertise. Book a session when you need targeted guidance, with no project required.

  • Quick Q&A 1h

    Model choice, errors, or performance issues

  • Architecture Review 2h

    Pipeline design feedback and improvements

  • Model Selection 1–2h

    Use-case-specific guidance on approach

  • Forecasting Readiness half-day

    Pre-project assessment of data and expectations

Assessment sessions can be credited against a future project.

Why Unit8

We didn't just
adopt Darts.
We built it.

Unit8 is a leading Swiss data services company with 170+ specialists and 350+ delivered projects. We created Darts and combine deep forecasting expertise with hands-on implementation experience.

Creators & Core Maintainers

Direct access to the team that designs and ships every Darts release. We shape the roadmap and know the internals better than anyone.

Engineering Meets Consulting

A unique mix of data science, engineering, and business consulting.

Trusted Technology Partners

Strategic alliances with the leading AI & data platforms.

OpenAI Microsoft AWS Palantir NVIDIA Snowflake Databricks Dataiku Fivetran digitalswitzerland Data Innovation Innovaud
Track Record
Successful Data & AI projects 350+
Data & AI specialists 170+
Data platform initiatives 25+
Open source community 9.3K+ stars

"Our goal is to turn Darts from a world-class open-source asset into a commercial engine that delivers enterprise value - while keeping the community at its heart."

U8
Unit8 Leadership
Lausanne, Switzerland
Client Voices

What our partners say about us

Our partnership with Unit8 has been instrumental in helping us accelerate how we enter new markets. Their structured approach to problem solving and professional team of capable data scientists has allowed us to deliver best in class results for our customers. This is a very strategic partnership for us.

BG
Brian Gahan
CIO, GridBeyond — AI-Powered Energy Technology
GridBeyond

Our partnership with Unit8 for developing forecasting and machine learning models at SIG is extremely positive. The experience and efficiency brought by Unit8 strongly contributed to accelerate data science initiatives and make an important step towards a more efficient smart city ecosystem.

FJ
Frédéric Joss
Head of Data & Innovation, SIG — Services Industriels de Genève
SIG

The combination of Unit8's hands-on implementation experience, industry expertise, and business understanding was key in bringing different teams and stakeholders together. Our collaboration resulted in a feasibility study for predictive maintenance on a power plant turbine and improved forecasting models based on Darts.

ME
Dr. Martin Eschle
Head of Procurement & Executive Board Member, IWB Industrielle Werke Basel
IWB

Through a strategic partnership with Unit8, we've transformed a foundational concept into a cutting-edge Machine Learning platform that significantly advances predictive maintenance. This tool aligns perfectly with our existing systems and offers flexibility to adapt to diverse power plant technologies.

MB
Martin Boden
Project Engineer, Hydropower Innovation, Alpiq
Alpiq
Contact Us

Let's talk about your business needs

Whether you're exploring forecasting for the first time or looking to productionize an existing setup, our team is ready to help. Tell us about your use case and we'll find the right solution together.

Explore a Use Case

Share your challenge and get expert guidance

Technical Consultation

Discuss architecture, models & production readiness

Partnership Inquiry

Explore long-term collaboration opportunities

Get in touch