Oh no, not another AI post from a management consulting firm! Attend a seminar on "Top Prompts for Generative AI" and suddenly every operational expert is now an AI expert!

We understand the skepticism. But what sets Skovinen apart is that we have balanced our core understanding of applying operational solutions, with the addition of formidable technology leaders with decades of experience in AI (including at Google AI and Two Sigma). This unique combination of operational and technical expertise allows us to empower companies to make rapid, data-driven decisions that lead to success.

While Generative AI grabs most of the headlines, many classical operational challenges—such as demand forecasting, capacity planning, OEE and supply chain optimization—can be significantly improved through a blend of traditional consulting methods and "classical" branches of AI like machine learning (ML) and deep learning (DL). And yes, we used quotes around classical because ML and DL are barely a decade and a half old—such is the rapid pace of development in this field.

Tried-and-true techniques like Machine Learning and Deep Learning — if you're not already using them — can dramatically enhance traditional business processes such as demand forecasting.

Demand Forecasting enables an organization to more accurately plan for the investment in, or allocation of, facilities, equipment, raw materials, inventory and resources at the right time. Never before have businesses been under such pressure to deliver products and services more quickly and accurately. The implications of demand forecasting are significant for businesses that rely on projections to run operations in an efficient, cost effective manner.

Yet, it is common practice to base demand forecasting solely on historical demand models, generally failing to incorporate domain knowledge, nor the latest advancements in statistical modeling. In addition, it is commonplace to override projection results and make adjustments based on “gut feeling” without objective justification. Often companies react to demand spikes or forecasting inaccuracies by investing in technology, without addressing fundamental business processes. Without the proper foundation in place to evaluate, make data-driven decisions and take action, innovative technologies prove to be of little value in and of themselves.

Our approach to enhanced demand forecasting (incorporating ML and DL) involves several key steps:

• Establish objective, data-driven fundamentals, such as enabling data sharing across cross-functional teams. Ensure that everyone—from executives to the shop floor—has a basic understanding of the key metrics that signal good or poor performance.
• Conduct expert analysis (by data scientists) of the multi-dimensional aspects of historical demand and relevant variables for each product category or SKU. These variables might include factors like holidays, game schedules, monthly average temperatures, rainfall, snowfall, and zip code-level consumer demographics, in addition to historical demand.
• Assess and apply the appropriate machine learning and advanced statistical techniques for each SKU. We incorporate an ensemble of models optimized for each product category, allowing us to best capture demand patterns and significantly improve forecast accuracy.
• Leverage unstructured data using large language models (LLMs) in specific steps of the overall ML/DL pipeline. We utilize LLMs to convert natural language data, such as external customer reviews (social media) and internal customer service records, into structured tabular data that can be fed into traditional machine learning models, ultimately improving forecast accuracy.
• Finally, establish robust business processes to ensure the effective capture, review, and response to the enhanced forecasts produced.

In summary, we combine operational expertise with advanced AI techniques to help companies make data-driven decisions and improve critical business processes like demand forecasting. While Generative AI garners much attention, proven methods like machine learning and deep learning can significantly enhance traditional forecasting models. By integrating structured and unstructured data and optimizing business processes, our approach leads to more accurate forecasts, enabling businesses to operate more efficiently and effectively.

If you'd like to explore how we can improve demand forecasting at your company, feel free to reach out to our leadership team. We'd be happy to discuss how our approach can make a tangible and significant financial impact on your business much more rapidly than traditional consulting solution providers can offer.

The Skovinen Team
Houston TX / Vancouver BC