Is the demand forecasting solution working for you, or are you working for it?

The retail world is rapidly changing and is already significantly different from what it was a year ago. In today’s challenging realities, it’s becoming more difficult than ever for retailers to stay competitive. One of the keys to maintaining their competitive edge, maintaining profitability and increasing sales is accurate demand forecasting across all categories, including the fresh category. Today, we’ll explore the power of modern artificial intelligence (AI) solutions with machine learning to help avoid costly mistakes in demand forecasting.

Demand Forecasting Solution: A Checklist for a Retailer

There is no shortage of forecasting solutions today, but there is a shortage of systems that can fully account for the complexity and reach of the FMCG retail supply chain.

To understand if your demand forecasting solution is up to date today, here are 6 key questions you need to answer.

  1. How much time and effort do your experts spend creating forecasts, and how much manual intervention does the process require?
  2. Does your demand forecasting system take into account category behavior across stores?
  3. Can you combine all your data for demand forecasting and replenishment across all channels so that it works for your business in a growing omnichannel retail environment?
  4. Are you able to respond quickly enough, on a daily or even intraday basis, to make accurate forecasts to meet changing customer needs?
  5. Can you accurately forecast stocks, seasonal demand, and weather and share these forecasts with suppliers to improve product availability in stores?
  6. Is your demand forecasting system reliable? Is it able to study and predict customer demand by category, product, sales channels?

Having trouble answering? Let’s look at each item separately.

1. How much time and effort do your experts spend creating forecasts, and how much manual intervention does this process require?

Surveys show that 52% of supply chain leaders believe they spend too much time manually managing data.

Take, for example, out-of-stock. When it happens, you lose sales of a certain product for a period. The main burden in such a situation is placed on your employees responsible for forecasting and replenishment, since they must take into account the missing data and enter them manually into the forecasting system.

The output of new products to the market is also quite labor-intensive. With most demand forecasting solutions, employees must manually tell the system how to forecast this new item given the history of a «similar» item. However, there are many types of «novelties». It can be a brand new product, a replacement for an existing product, or a variant of an existing product. All of these changes require significant manual intervention.

Factors such as seasonality, the opening of a new store are also «time sinks» for staff, often causing additional need for manual data processing and error correction.

In retail, maintaining quality data is critical for the prediction engine to work properly. And the use of an effective demand forecasting solution, with advanced data cleaning and training capabilities eliminates manual intervention of the staff, thereby reducing the level of labor costs and the number of errors in the forecast.

2. Do your prediction systems aggregate all categories across the store?

What happens in one part of the store affects other processes of the company. Retailers often use best-in-class forecasting solutions for core categories, but manually forecast categories in the fresh group. Or use a separate solution to manage fresh goods, isolating them from the main forecasting solution.

Indeed, most technologies designed to work with fresh produce are not designed to manage non-food categories as well. Conversely, systems for planning non-food categories do not take into account the nuances and subtleties of managing fresh goods. As such, retailers often operate with disparate demand replenishment systems.

According to surveys, only 36% of supply chain professionals reported working on a single platform.

Fresh goods categories are particularly difficult to predict. These products are sensitive to external events such as the weather and have very short shelf lives. Add to this the requirements for traceability and intraday delivery of some products, then forecasting becomes even more complicated.

Using the main system for general forecasting and an isolated method of forecasting these more complex categories deprives retailers of a holistic view of all their products and categories, which exacerbates the situation.

Your forecasting solution must be able to take into account the nuances of forecasting both the main categories and the categories of the fresh group.

3. Could you use datasets to predict demand and replenishment across all channels?

Omni-channel – must have for modern retail.

Accurate forecasting and replenishment allows retailers to accurately forecast demand across all channels to effectively prioritize and fulfill orders no matter when, where or how they are received from consumers.

To achieve efficiency in omnichannel retail, retailers need a comprehensive solution that can understand the actual demand of the network and then manage the distribution of inventory across each sales channel.

The solution must guarantee synchronized inventory to ensure product availability for both in-store and online shoppers. The system must be able to provide the exact time and place of picking up goods in orders, both for employees and for third-party organizations — delivery partners.

By connecting systems to streamline order management, retailers can reduce order fulfillment costs by selecting the most appropriate inventory source. By being able to confidently identify the best source of inventory needed for any given fulfillment channel (be it shelf, warehouse, dark store, or other), you can always meet customer needs at the lowest cost of replenishment.

4. Could you respond quickly enough to generate accurate forecasts to meet changing customer needs?

The turmoil of recent months has led to significant and rapid changes in buying behavior.

Omni-channel demand, fluctuations in shopping frequency, and a range of other factors greatly impact retailers’ ability to predict what will be in stock, when, and how it will be stocked. Demand for fresh, ready-to-eat and takeaway food remains high, but is more cross-channel than ever before. Retailers must have relevant data to understand how all categories affect each other.

A vital component for every supply chain is complete and accurate visibility of stock availability throughout the chain. This allows not only to confidently predict changes in demand, but also to react quickly when such changes occur unexpectedly. In addition, it makes it possible to support complex logistical supply chains by sourcing and obtaining goods from multiple locations and/or suppliers during a disruption period.

Machine learning-enabled demand forecasting systems are helping to continuously improve the accuracy of forecasts. Algorithms learn relentlessly using historical, current and contextual data without requiring user intervention.

This allows the system to recognize the event as an anomaly that does not normally exist in the demand cycle. Combining historical data before the event, data during the event, and incoming data received after the event, gives a complete understanding of the impact of the failure on the behavior of the buyer in the near and short term.

5. Coud you accurately forecast stocks, seasonal events, weather, and share these forecasts with suppliers to improve product availability?

Events, holidays, and their accompanying promotions are complex for supply chain planning processes. The ever-evolving distribution channels, the struggle for loyalty and the growth of digital consumer awareness have made many methods of stock forecasting and seasonality accounting completely obsolete. Added to these problems are health problems, weather, economic and political changes, and it seems impossible to predict what will happen next.

Traditional systems solve the «lost past» problem by grouping similar products together and making assumptions about their behavior. However, these assumptions do not always reflect real demand.

Speaking of promo forecasting challenges, it should be noted that the supply chain may not even know about the promotion. And this, in turn, can lead to out-of-stocks. In addition, the supply chain must understand the mechanics of the promotion to be able to track the progress of the promotion across all channels and report on the effectiveness of the promotion.

What happens when you share a prediction? In an ideal supply chain, the demand forecast is based on the “single version of the truth”. This forecast must be available to the entire business and suppliers in order to achieve the desired results and meet the goal of the end-to-end supply chain.

But when forecasts are made with errors or omissions, it causes a ripple effect throughout the retail organization and its partners, leading to inaccurate plans, poorly executed promotions, and strained supplier relationships.

A demand forecasting solution must respond to any combination of recurring events (such as school holidays, vacations, sporting events) and unexpected disruptions (such as weather, economic, and medical events) in an automated manner that eliminates human intervention and greatly improves forecast accuracy.

6. How forward-looking is your demand forecasting system? Can it learn and predict consumer demand by category, product, channel?

As we said earlier, the way consumers shop has changed dramatically. Traditional systems, which are not trainable and base forecasts solely on the past, cannot keep up with the constant changes in customer behavior. The solution simply cannot predict future demand without the possibility of learning. In addition, traditional systems are built on algorithms that are given once and depend on a human resource to manage exceptions.

AI-powered demand forecasting solutions with machine learning enable retailers to make more accurate forecasts and reduce errors. Because AI not only easily handles what is difficult for a human to do (input of varied and complex data), but also recognizes trends and anomalies in demand.

The implementation of such solutions provides high quality data, forecasting through various channels and the ability to use a single system that can predict all categories. This in turn saves time, eliminates the risk of manual errors, and frees up employees to do more important work.

Got questions? Do you want to learn more about complex IT solutions for automating forecasting, pricing, loyalty program management and auto-order systems? Get advice from Atriny specialists: write to the mail or fill out the form below.


About Us

Contact us

© 2023.All Rights Reserved.