Automated demand forecasting's breakfast session

"I am glad to join this breakfast session last Thursday, and the topic is quite relevant to my daily operation work. My job responsibility is to manage the stock demand and supply planning," said a manager from Venchi.

The breakfast session turned out great, with a full house, allowing local business professionals to make invaluable cross-industry contacts. Following the coffee session, Edmond and Laurent Chindeko from #Explora shared their insights about the what's, why's, and how's of #demandforecasting and #replenishmentplanning.

As the world begins its slow pivot from managing the COVID-19 crisis to recovery, new consumer behaviours span all areas of life, from how we work to how we shop and how we entertain ourselves.

This consumer centricity increases the strength of the retail supply chain. Traditional aggregate forecasts are no longer sufficient. Companies must be able to predict and manage demand per day, per individual SKU, and at the store level, almost near real-time.

To predict consumer behaviour, we need to have enough data about historical sales, promotion shoppers, and causal factors for the company to estimate the safety stock level correctly.

(Storytelling: How to achieve demand planning & replenishment excellence using Azure Microsoft & Databricks | Photo: Odilia)

Many thanks to Edmond and Laurent Chindeko, who demonstrated some replenishment use cases commonly shared in F&B/retailers/logistics that can be enhanced using the latest technologies and data science.

There are a couple of key takeaways in those use cases we shared around demand planning and replenishment procedures, and they are listed in the bullet points below:

• Traditional solutions can't scale to fine-grain. Fine-grain demand forecasts can capture the patterns that influence demand closer to the level at which that demand must be met.

• A scalable solution predicts the optimal stock profile across all products and branches. It can improve store operations efficiency, shorten order processing time, reduce "stock-out" and "overstock", and increase accuracy with different causal data.

• Data is knowledge. It supports domain-specific business decisions.

(Get Started: Accelerating Data-Driven Innovation | Photo: Odilia)

Special thanks to Tiffany Law from #Microsoft, who shared the common challenges slowing down #bigdata and #ML projects. She saw many instances of siloed data across teams and departments, which made developing unified data pipelines very challenging.

To become truly data-driven, an organization must implement a scalable unified analytics platform. She showcased one interesting use case about Starbucks, which unified petabytes of streaming and historical data to deliver real-time and accurate store-level forecasts.