Ben Stokes   |   August 22, 2023

Forecasting the Future of Fresh Prepared: How AI Revolutionizes Production Planning

How does the deli team do more with less?

Grocery retail has certainly faced its fair share of challenges in recent years. The pandemic turned the industry on its head, forcing grocers to quickly pivot to new ways of doing business, such as online ordering and curbside pickup, while also dealing with increased demand and supply chain disruptions. Supply chain issues have persisted, causing shortages of certain products and higher prices for others. Now, rising inflation coupled with changing consumer behavior patterns are adding to the pressure. Consumers are becoming more price-sensitive, while also seeking out healthier and more sustainable options. 

Consumers are becoming more price-sensitive, while also seeking out healthier and more sustainable options. 

The increasing momentum from consumer spend towards prepared foods in grocery stores as a convenient and cost-effective alternative to restaurants has become more evident due to these factors.According to a report from Nielsen, sales of prepared foods in grocery stores have grown by 30% over the past four years, compared to just 3% growth for restaurants. This trend is likely to continue, as consumers prioritize convenience and value in their food choices.  

According to a report from Nielsen, sales of prepared foods in grocery stores have grown by 30% over the past four years, compared to just 3% growth for restaurants.

As a result, the pressure is on grocery stores and associates managing fresh departments to deliver high-quality prepared foods that meet consumer demand, while also managing the complex forecasting and production processes behind the scenes. They ultimately need to do more in-store, with less labor support than ever.  

The Production Conundrum 

The Production Conundrum 

Production Associates are tasked with reducing shrinkage and driving sales while maintaining a consistent and data-driven approach to creating production plans. Planning for production of prepared goods can be a complex task, especially for items required every few days, such as deli sandwiches or blueberry muffins that might only be required on certain days of the week. The complexity of localization of assortment, shift-based requirements, and labor availability adds to the challenge.  

“Kitchen Math is not sustainable … or what has worked until now, cannot work in the future”

Traditionally, production planning for prepared goods has relied on institutional knowledge from store associates, who often made intuition-based predictions for how much will sell based on the day of the week. They would often add up numbers in their head and rely on cardboard pieces or scrap paper available near them to take notes on the quantity of ingredients being consumed by the current production run and doing the “kitchen math” on how much more needs to be ordered. That’s not all, production departments face a four-sided challenge when it comes to balancing execution with operational efficiency.  

production departments face a four-sided challenge when it comes to balancing execution with operational efficiency.  

The Four-Sided Challenge in Prepared: Challenges faced by Fresh Departments 

Challenges in the realm of production planning may not be immediately apparent but become increasingly evident as associates deal with more complex situations. These challenges can be categorized into four distinct segments: Production Challenges, Operational Challenges, Labor Challenges, and Revenue Challenges. Each of these segments presents unique and conflicting situations that must be addressed by associates.

These challenges can be categorized into four distinct segments: Production Challenges, Operational Challenges, Labor Challenges, and Revenue Challenges. Each of these segments presents unique and conflicting situations that must be addressed by associates.

Production Challenges:In a grocery retail environment, production challenges encompass the complex task of efficiently managing inventory, ensuring optimal product availability, and minimizing waste while balancing unpredictable customer demands. These challenges demand meticulous planning, streamlined processes, and adaptable strategies to meet the ever-changing dynamics of the industry. Store Associates typically face situations where they tend to ask themselves questions that look like: 

  • Should I wait for the dough to proof, or should I mix in yeast into the warm water? What is the right sequence of steps? 
  • I received a production run for 3pm and one for 6pm, should I combine the two? 

Operational Challenges:  These refer to the various obstacles faced by associates in the day-to-day operations of managing fresh departments such as Deli, Bakery or Seafood. These challenges can encompass a wide range of areas, including supply chain issues, vendor interactions, and forecasting, among others. Associates dealing with operational challenges may find themselves faced with tasks such as ensuring accurate inventory levels, managing stock replenishment, handling unforeseen disruptions in the supply chain, addressing customer complaints and inquiries, and maintaining overall store efficiency and profitability. These challenges require effective problem-solving skills, adaptability, and the ability to make quick and informed decisions to keep the store running smoothly and meet customer expectations. Associates often face decision dilemmas such as: 

  • My vendor told me a delivery for mushrooms is on the way, Should I subtract those quantities from my next ingredient order? 
  • Does this order quantity include projected spike for the long weekend? 

Labor Challenges:Managing and organizing the workforce effectively is also a big part of the associate experience and typically require them to deal with labor challenges such as inadequate staffing levels, high employee turnover, scheduling conflicts, and skill gaps among the workforces. These challenges could potentially lead to increased workloads, decreased productivity, customer service disruptions, and overall operational inefficiencies. In order to be able to produce an optimal fresh assortment for the day, Associates may need to balance staff availability, ensure optimal coverage during peak hours, address employee absences, and maintain a skilled workforce, all while striving to meet customer demands and deliver a seamless shopping experience. While running a production shift they may need to arrive at answers for common questions such as: 

  • How many associates would we need for the 12-noon production run tomorrow? By when should they report at the Deli counter? 
  • The production associate finished 50% of MTO (Made-To-Order) orders during his shift, how should he communicate what was completed and what is not to the associate coming in next? 

Revenue Challenges: These challenges are probably at the core of every other challenge and are a big part of the store associate deliverables and revolve around the difficulties that arise in generating and maximizing sales and profits. Associates may need to analyze sales data and identify trends or patterns that impact revenue generation. They might also need to strategize and implement effective availability strategies to prevent stockouts or overstocking, as well as collaborate with marketing teams to develop impactful campaigns that drive customer engagement and boost revenue. Rapid rise of online marketplaces also has the department managers facing questions around: 

  • How do I aggregate orders from different marketplaces? 
  • Should I offer Deli sandwiches to our mobile app shoppers? 

Powering through all these challenges plaguing fresh departments in grocery needs one real superpower: AI-Driven Forecasting. It equips associates with advanced Machine Learning algorithms that arrive at precise sales forecast.

This in turn enables associates to make informed decisions on pricing, promotions, and inventory management allowing them to align supply with demand.  

How it works 

Accurate AI Driven forecasts will help predict the number of finished products that will be needed and will be able to do this by production runs. It can then break these forecasts into requirements for raw ingredients, so you could plan for lead times for prep steps and supply dynamics. Also, AI-driven forecasts can combine many variables such as seasonality, weather, local community preferences, and promotional spikes and aggregate the impact of orders from different marketplaces by channel.  

By ingesting variables on demand patterns, such as seasonality, trends, promotions, weather, holidays, and events, AI-driven forecasts can  find the optimal way to use all available information to make the best possible predictions. This ability helps streamline production operations by ensuring that the right amount of food is prepared at the right time, reducing waste, and increasing efficiency. Grocery chains generate vast amounts of data every day, and only AI can make sense of such large data sets. By continuously learning from new data, AI algorithms become more accurate over time, allowing grocery chains to better understand their customers’ needs and preferences. As a result, they can make better decisions about what foods to stock and how much to prepare, leading to more satisfied customers and increased profitability. 

Forecast for Blueberry muffins

If we were to predict how many blueberry muffins we should produce in a day, here is how we would approach it. Obviously, all of these numbers are hypothetical, and we’ll keep it simple:

The first step is to be going to be to estimate customer demand for our production plan item.  Hopefully we have a rich sales history but even if we don’t, that’s okay.  Our AI algorithm is designed to work with whatever it has available to it.  In any case, by learning from the way the production plan item’s sales history interacts with the calendar (time of year, time of month, day of week, proximity to holidays, and some stuff that’s a little more abstract), with department and overall store sales, with weather, with special events, with changing price levels, and whatever else our clients may identify as important, the AI algorithm makes the best possible projection for item demand given the calendar, weather forecasts, store sale forecasts, upcoming events, and the expected pricing levels.  Historically, sales projections have been confined to understanding the effect of one or two variables at a time.  And the need for human interpretation restricted detailed forecasts to a handful of items.  AI changes everything by facilitating micro-decisioning where highly multi-variate forecasts can be done with unprecedented granularity.   

But we digress.  Let’s just say we ran our AI demand forecast and it told us our next production run will sell 100 blueberry muffins. 

AI-Driven Forecasting allows retailers to identify the drivers of projected revenue

A baseline forecast of 100 muffin sales is set, followed by adjusting the forecast for promotions running in the store, which is done by multiplying it with a promotion adjustment factor of 1.2 (depending on client preference, the promotion adjustment factor can also be automatically determined by the AI forecast). The forecast is further adjusted based on upcoming events by multiplying it with an upcoming event adjustment factor of 1.5, considering an increase in demand due to Super Bowl Sunday.

Again, the AI forecast could determine the event adjustment factor based on previous spikes and sales history that it has ingested. A minimum quantity required for the display section is applied to ensure that the muffins are well-stocked. If the final forecast falls below the minimum quantity, it is increased to the minimum display quantity. Finally, a safety stock of 20 is added to the final forecast to account for unforeseen circumstances such as unexpected spikes in demand, delays in delivery of raw ingredients, or supply chain disruptions. This layered approach to forecasting and inventory management has seen resounding success at Upshop’s largest customers who currently use Production Planning.  

Making sure store associates can clearly see the impact of every variable that influences the resulting end forecast plays a major role in building trust and adoption of the system in turn driving ROI of bringing in the right technology to streamline operations. Rest assured that whenever a shopper chooses a blueberry muffin, whether in-store or through your online delivery channels, they can be confident in receiving the freshest baked goods your store offers. 

 About Ben Stokes

Ben Stokes, Ph.D., is the Vice President of Data Science at Upshop. With 25 years of experience in computational predictive analytics, he began his career as a researcher in particle physics phenomenology. For the past seven years, he has served as an architect and team leader, specializing in operations ML/AI applications. Ben excels at utilizing his expertise to tackle previously insurmountable challenges for Upshop’s retail partners. His ability to solve complex problems is unparalleled, bringing valuable solutions to the table. Ben and his team of data scientists work tirelessly to build cutting-edge solutions that enable Upshop’s retail partners to better serve their customers by leveraging data-driven decisions to guide business strategy, right from the fresh perimeter all the way to DSD operations. 

Related Posts

Ben Stokes

Vice President of Data Science