Improved forecasting accuracy by up to 30%, reducing inventory costs and boosting revenue
Fostered a data-driven culture with advanced AI tools
Collaboration with Google Research and Google Professional Services reduced development time and improved efficiency
Minimized waste, supporting OTTO’s sustainability goals
Optimized supply chains for better product availability and customer satisfaction
OTTO used Google Cloud's TiDE model and Vertex AI to improve their demand forecasting accuracy by 30%, optimizing inventory, pricing and reducing waste.
OTTO used Google Cloud's TiDE model and Vertex AI to improve their demand forecasting accuracy by 30%, optimizing inventory, pricing and reducing waste.
OTTO elevates demand forecasting with Google Cloud’s AI solutions
OTTO, a leading German ecommerce retailer, embarked on a journey to enhance the accuracy of its demand forecasting processes to improve inventory management and customer satisfaction. OTTO’s forecasting team, Team Lumen, implemented Google Cloud’s advanced AI tools, including Vertex AI, BigQuery, and Google Kubernetes Engine (GKE), to achieve this. The integration of the Time-series Dense Encoder (TiDE) model revolutionized their demand prediction capabilities.
In practice, OTTO’s employees leverage TiDE to analyze vast datasets containing multivariate time-series data, such as seasonal demand patterns, pricing changes, and external factors like promotions. For example, when planning seasonal inventory for gaming consoles, TiDE’s advanced architecture delivers highly accurate predictions by capturing short- and long-term dependencies.
Dr. Christian Wiel, a key contributor to the project, highlights: "TiDE’s responsiveness to shifts in demand ensures accurate forecasts, even for highly dynamic products like iPhones."
The adoption of Google’s solutions resulted in an up to 30% improvement in forecasting accuracy, reduced inventory costs, and increased customer satisfaction through timely product availability.
One of the key improvements is TiDE’s ability to adapt rapidly to short-term fluctuations in sales behavior, such as sudden changes in demand due to significant price adjustments. This responsiveness has minimized the propagation of short-term sales peaks, which previously distorted longer-term forecasts.
Dr. Christian Wiel
Senior Data Scientist and Developer
Working with complex time-series data was a challenge. TiDE’s architecture effectively captures short- and long-term patterns for accurate forecasts, even with large datasets. Its efficient design scales easily, and it supports external data like weather and macroeconomic indicators, providing valuable flexibility and cost-effectiveness.
Till Böhmert
Senior Cloud Architect
OTTO's transformational advancements and impacts with Google Cloud
For highly stable products, such as men's jeans, TiDE converged to results comparable to our baseline. However, for more dynamic items like iPhones, the model demonstrated superior responsiveness to shifts in demand levels and provided better adaptability to changes.
Dr. Christian Wiel
Senior Data Scientist and Developer
In practice, OTTO employees use these technologies to improve product forecasts. For example, a demand analyst will extract historical product sales data from BigQuery and use it in Vertex AI to train forecasting models. These forecasts are then used to adjust stock levels and define pricing strategies, thereby better meeting consumer needs and maximizing sales. This includes, for example, managing products with atypical demand behavior, such as seasonal items or new products with uncertain demand.
The Time-series Dense Encoder (TiDE), trained on Vertex AI and deployed on GKE, has enabled OTTO to improve the accuracy of demand forecasts by up to 30%. TiDE is particularly useful for handling complex multivariate time series, such as those linked to the evolution of sales of electronic products or clothing items. This model is capable of capturing both short- and long-term temporal dependencies, which is crucial for predicting demand trends more accurately.
A key benefit for OTTO is the ability to reduce stock-outs and better match supply to actual demand. For example, by more accurately predicting demand for products such as iPhones or men's jeans, OTTO can proactively adjust production and orders, avoiding both overstocks and shortages.
OTTO's teams, particularly those responsible for demand forecasting, also use Google Kubernetes Engine (GKE) to manage large-scale models, enabling smooth and rapid deployment of adjustments and experimentation on real-time data. This infrastructure enables OTTO to work effectively with massive datasets while optimizing operational costs and improving efficiency.
The benefits of this solution also include financial advantages. For example, thanks to improved sales forecasting, OTTO can adjust its stock levels more accurately, reducing the costs associated with excess inventory while increasing product availability. In addition, improved accuracy enables dynamic pricing strategies, allowing OTTO to maximize revenues during periods of high demand and minimize losses during off-peak periods while ensuring fairness using device-, geo- and gender-independent methods.
In summary, OTTO benefits from the integration of Google Cloud solutions to create more accurate and intelligent forecasts, thereby reducing costs, improving the customer experience and optimizing strategic decisions.
Revolutionizing tomorrow's world: OTTO's vision for AI-powered innovation
OTTO is transforming its data management approach with Google Cloud and Vertex AI, by extending their use of AI and predictive analytics to deliver enhanced customer services.
For example, OTTO leverages advanced analytics on Google Cloud to understand customer behavior and optimize operations.Thanks to Google Cloud's machine learning capabilities, OTTO's teams gain valuable insights that enable them to make proactive and strategic decisions.
Key benefits of using Google Cloud solutions include optimized inventory management thanks to more accurate forecasting, and an overall improvement in operational performance. By leveraging Google Cloud tools, OTTO can respond nimbly to rapid market changes while boosting its logistics efficiency.
By enhancing the integration of AI and machine learning, OTTO is positioning itself as an innovative player in online commerce. This allows them to offer customers smarter, more efficient services while optimizing internal operations to remain competitive in a dynamic and evolving market.
TiDE's capacity to model the entire lifecycle of a product has further enhanced our ability to manage items from launch to discontinuation, ensuring more consistent and reliable forecasts throughout the product’s lifespan.
Dr. Christian Wiel
Senior Data Scientist and Developer
OTTO is a leading German ecommerce retailer specializing in diverse product categories, known for its commitment to innovation, sustainability, and operational excellence. With a strong focus on leveraging advanced AI and cloud technologies, OTTO aims to enhance customer satisfaction, optimize supply chain efficiency, and maintain its position as an industry pioneer.
Industry: Retail
Location: Germany
Products: BigQuery, Vertex AI, Vertex AI Workbench, Vertex AI Pipeline, Google Kubernetes Engine (GKE)
