How we helped a leading spirits manufacturer reduce production costs through enhanced forecasting models

Summary

Client: Our recent client was a leading spirits manufacturer headquartered in the USA, with revenue exceeding $14Bn.

Challenges: The company operated multiple manufacturing facilities and warehouses globally, resulting in complex supply chain dynamics. Inaccurate forecasts led to stock-outs for some SKUs and excessive inventory pile-up with others, resulting in substantial financial losses exceeding $200 million. The client had already developed a packaged forecasting solution however, its complexity made it difficult to reconfigure and manage i.e. made it impossible to utilize any variation in forecasts. 

Solutions: Quantzig analyzed the demand signal of various SKUs and developed an advanced SKU classification framework. This framework enabled the client to categorize and understand the unique characteristics and demand patterns of their products. Besides, Quantzig employed advanced machine learning algorithms to create multiple forecasting models for each SKU. The process was further streamlined by automating the calculation of the MAPE for each model, enabling the selection of the best-fit model for each SKU. We helped the client bring down the 30-day manual process to 4-day automated process using a combination of Azure Synapse, Power Automate, and Power Apps along with a Python backend.

Impact Delivered

All of the above initiatives resulted in:

• 10-15% point improvement in forecast accuracy
• 80% reduction in forecasting process time
• Significant improvement in the production planning process

Industry Overview

• The USA spirit manufacturing industry is a vibrant and influential sector, known for its wide range of spirits, including iconic American whiskeys. With continuous innovation, growing interest in craft distilleries, and a strong export market presence, it remains an essential part of the country’s rich alcoholic beverage heritage.
• The USA spirit manufacturing industry is a significant contributor to the country’s economy. It generates substantial revenue, with total sales reaching billions of dollars annually. 
• An accurate forecasting model fuels the spirit of success for USA manufacturers, uncorking profitability through accurate predictions and optimized inventory management.

About the Client

• Our recent client was a renowned spirits manufacturer headquartered in the USA, with revenue exceeding $14Bn.
• The client had a complex manufacturing process and was operating more than 20 manufacturing facilities and 70 warehouses across the globe. 
• However, the company faced a pressing challenge stemming from its inaccurate forecasting methods, which resulted in frequent stockouts and IHC. 
• The combined impact of these detrimental factors amounted to a significant loss of approximately 20% to their thriving business.
• An innovative and robust forecasting model was urgently needed to combat these detrimental financial losses and secure a prosperous future.  

Challenges

• Due to a manual data manipulation and approval process, the client’s forecasting process was 30 days long. Along with this, forecast accuracy challenges resulted in issues related to production planning and supply chain planning as the teams had to backward correct planning mismatch in the next forecasting cycle.
• The prevailing demand forecasting solution, operating on a rigid “one-fit-all” approach, astonishingly overlooked the crucial signals emanating from the vibrant realm of marketing. The absence of a harmonized framework resulted in a significant disconnect between marketing efforts and demand forecasting, causing missed opportunities and hindering the client’s understanding of consumer behavior. The detrimental consequences were palpable, with frequent stockouts, soaring inventory holding costs, and an alarming 20% loss to the client’s bottom line. 
• Forecast inaccuracy plagued the client’s operations, causing a staggering $200 Million loss due to a combination of SKU pile-up and stockouts for specific items. The inability to predict demand accurately left the client grappling with excessive inventory for some SKUs, while others remained out of stock, leading to a dissatisfied customer base and missed sales opportunities. The consequences of this forecasting shortfall were far-reaching, causing financial losses and hampering the client’s ability to meet the evolving needs of their target audience. They sought to address these issues through collaborative efforts, innovative solutions, and a willingness to embrace forecasting models to enhance their forecasting accuracy and achieve sustainable growth.
• The client recognized a pressing need for a demand forecasting solution that could achieve a minimum of 20% improvement in overall accuracy to combat this current forecasting shortfall. Their existing forecasting methodology lacked the necessary flexibility and adaptability to capture the complex signals emanating from the multifaceted spirits industry, thus leading to frequent stockouts, excessive inventory holding costs, and missed sales opportunities. The proposed forecasting model needed to integrate advanced algorithms, predictive analytics, and machine learning capabilities, enabling the client to harness the vast quantities of data at their disposal to improve accuracy and optimize inventory management. 

Solutions

• To help develop bespoke solutions, our experts developed a four-step process to help the client tackle all the challenges: Demand Forecasting Module, Demand Adjustment Module, Automation Module, and Business Intelligence Command Centre.
• These four steps include: enriching data by adding micro- and macro-economic variables and competitive impact. Followed by improving the analytics model by incorporating advanced algorithms and ensemble models through our K-Fold validation network based on 16 different advanced algorithms.  To tackle the complexity of the package, we created demand adjustment modules to correct the demand from previous demand variations and the current stock on hand. And finally, develop the ability to deep-dive into forecast vs actual variations and identify the root cause of variations for the next cycle.
• To address the client’s pressing demand forecasting challenges, Quantzig leveraged advanced analytics and machine learning algorithms to analyze the demand signals of various SKUs and developed a comprehensive SKU classification framework. This framework served as a foundational cornerstone for the client’s demand forecasting strategy, allowing them to identify and group products based on their unique characteristics, demand patterns, and sales velocity. By incorporating this SKU classification framework into their forecasting possess, the client gained a deeper understanding of their product portfolio, enabling them to fine-tune their forecasting models and optimize inventory levels.
• Further, we enabled advanced machine learning algorithms to read the demand signals of various SKUs and based on that run multiple algorithms on each SKU. This approach helped the client to develop a robust demand forecasting strategy that could adapt to the dynamic nature of the spirits industry, capturing the complex signals emanating from various sources such as marketing, seasonality, and trends. By leveraging Quantzig’s expertise in machine learning and forecasting models, the client was able to deploy a suite of advanced algorithms, including time series analysis, deep learning, and random forecast models to gain a granular understanding of their demand patterns and optimize inventory levels accordingly. The integration of machine learning algorithms marked a significant milestone in the client’s journey towards embracing data-driven insights and enhancing their operational efficiency, enabling them to stay ahead of the curve in the dynamic spirits industry.
• Moreover, we automated the process of calculating the MAPE (Mean Absolute Percentage Error) of each forecasting model, streamlining the client’s demand forecasting process and enabling them to make data-driven decisions with greater efficiency and accuracy. By integrating this automation into their workflow, the client was able to identify the best-fit forecasting for each SKU based on their unique demand patterns and characteristics. This approach allowed the client to optimize their inventory management processes and reduce financial losses by mitigating the incidence of stockouts and overstocking. The integration of automation and machine learning algorithms represented a powerful combination, enabling the client to exceed the evolving needs of their target audience with newfound precision and agility.