In Denmark and across the world, Vestas has embodied the principles of intelligent innovation for 125 years. It therefore comes as no surprise that the Vestas Service Analytics team has recently collaborated with Dataiku to optimize their shipment patterns, which is estimated to save Vestas millions of euros in the process.
From its humble beginnings as a small manufacturing company on the west coast of Denmark, the company has grown into a global leader in sustainable energy solutions, with 29,000 employees working to design, manufacture, install, develop, and service wind energy and hybrid projects all over the world. To date, with over 160 GW of wind turbines installed in 88 countries, Vestas has already prevented 1.5 billion tons of CO₂ being emitted into the atmosphere.
In recent years, the Service Analytics team at Vestas has played a key role in keeping the company at the forefront of a sustainable future, enabling business decisions and processes with data products and insights across the entire value chain. But how have they achieved this? We spoke with Mohamed Musthafa Shahul Mohamed, Data Science Lead at Vestas, to learn how his team is bringing data to the core of the business.
Bringing Simplicity to a Complex Industry
If each industry faces a unique set of complex analytics challenges, those facing the sustainable energy sector are among the most complex. In the case of the Service Analytics team at Vestas, they have to consider not only external, customer-facing products, but also internal stakeholders across the Operations, Finance, Supply Chain, and Commercial teams. All of these teams work together to answer big questions for the company such as how and when to deliver a turbine part from point A to point B.
A year ago, Service Analytics recognized that a more robust data operation could help them simplify and improve logistical challenges like this. They understood, in particular, that data science based solutions in predictive asset maintenance, field capacity planning, inventory management, demand and supply forecasting, and price planning would provide critical support to the internal customers of Vestas.
Until that point, the data team ran a mighty but more traditional business intelligence (BI) based analytics operation, querying BI-dashboards, deriving insights, and building data products in a less automated manner. The big shift happened when they decided to upgrade the team’s maturity, with an eye toward building solutions that used machine learning and advanced analytics. As part of this transition, a Center of Excellence (CoE) for advanced analytics was put together with the aim of identifying transitional areas within Service Analytics. This involved building proof of concepts (PoC) to showcase their machine-learning capabilities, upskilling the team, and identifying tools and a technology ecosystem that would support their journey over the long run.
Turning Up the Dial With Dataiku
Over the past year, Dataiku has served as the cornerstone platform for Service Analytics’ CoE. As any data team leader knows, finding the right platform is a careful calculation that balances several data science, business analytics, and IT considerations at the same time. Importantly, Mohamed was looking for an Enterprise AI platform that would help Vestas Service Analytics along its advanced analytics journey, empowering the team to mature and upskill as its capabilities increased.
After conducting an internal study of available data science platforms, Mohamed was drawn to Dataiku for five main reasons:
- It provides a simple tool for data preparation, exploration, model building and model deployment in a single platform.
- It offers support for citizen data scientists through auto-machine-learning features, with augmented functionality at every stage of the data science and machine learning cycle.
- It increased time-to-market, offering targeted business solutions for Vestas’ industry- and function-specific analytic solutions.
- It offers an agnostic toolset allowing existing business users to focus on getting value from their data whilst collaborating with more technical code based users.
- It allows users to access data stored on local machines or from Azure cloud; as well as offering good integration with Snowflake and other cloud services.
Given these benefits, the team decided to onboard the Dataiku instance and begin working on use cases and PoCs that could demonstrate the power of Dataiku and the improved time-to-value it enabled.
By far the greatest benefit derived from working with Dataiku has been the fruitful collaboration between the data team and Dataiku’s support team. Over the span of a few months, Dataiku has helped to upskill Mohamed’s team at Vestas, which is now empowered to drive and deliver powerful projects on its own.
Together, Vestas Service Analytics and Dataiku have succeeded in improving the maturity and capability of the data team at Vestas, and in demonstrating the value of data-driven and machine-learning solutions for the enterprise. To date, the most impactful use case that the Service Analytics team has built with Dataiku has been its express shipment recommendation model.
Use Case: Express Shipment Recommendation
With 160 GW of wind turbines to keep in good condition around the globe, Vestas needs to ensure that it can ship the right parts to the right sites within the right window of time — all without incurring unnecessary costs. To this end, the company offers not only regular but also express shipping to its internal customers, ensuring that time-sensitive problems are solved promptly and effectively.
But what happens when express shipping is used to service repairs that are not, ultimately, time-sensitive? There is a very high cost associated with express shipping, which often involves the transportation of parts by air. But Service Analytics found that 52% of express-shipped materials were not being put to use for at least 4 months, and sometimes even longer.
The goal, then, was to create a recommendation tool that would help dispatchers and planners know whether, for a given request, express shipping should be used. The data team decided on a proof of concept in the form of a stand-alone web application built within the Dataiku ecosystem that would make recommendations on shipment type.
The biggest obstacles to building this model, prior to Dataiku’s involvement, was utilizing the existing internal data sets to build a deployed machine learning model and web application. Working with the Dataiku team enabled Vestas Service Analytics to enhance the technical skills of its existing team, improve data access and data quality for the business. In addition, the close collaboration with the Dataiku data science team, allowed Vestas to become independent when delivering future projects on the platform. Within about one month, the Service Analytics team delivered the project as desired, with a deployed machine learning model, API’s and a stand-alone application for end-users.
With the model delivered, Service Analytics were able to prove their theory about the value that could be derived from changing shipment patterns across the company. Because Dataiku makes it easy to monitor, realign, and tweak existing processes as they run, Mohamed and the data team hope to improve their model based on tester feedback, and eventually to integrate it into Salesforce. Though the savings generated by the express shipping recommendation model will only fully materialize over time, the tool when globally implemented is estimated to reduce express shipment costs by 11-36%.
Analytics At Scale
Looking ahead, the data team hopes to continue along its AI-maturity path, and to grow the number of use cases they can build in order to empower the business to make decisions that both align with their sustainability objectives and reduce unnecessary costs. They plan to take advantage of the many features of Dataiku that not only make project development and deployment easier, but also those that help to educate, upskill, and empower data analysts and scientists in their effort to derive greater business value from their data.
As Mohamed puts it, “Dataiku’s platform has a matured ecosystem in terms of capabilities available in the platform, proposed capabilities pipeline, an academy for learning, and a large user community and support structure, which is quite helpful for teams like ours that are transitioning into the data science and machine learning space from BI.”