Methodology for successful assortment forecasting
In a store, the assortment plays a decisive role since it must allow a company to develop its performance and profitability. To do so, it must be constantly optimized according to seasonality, customer needs, competition, etc. If machine learning and artificial intelligence are used for sales forecasting, we will see that they are also used for assortment optimization.
But how do you go about making successful assortment forecasts? What information should be taken into account to obtain reliable assortment forecasts? How does machine learning help to address assortment issues? We propose you to discover how to make assortment forecasts in order to gain in performance and productivity.
Principles and challenges of in-store assortment forecasting
Assortment forecasting allows to realize an assortment plan adapted to the sales space and adapted to the customers. Before detailing the effective prediction solutions, we will define the issues related to assortment forecasting.
What is the role of an assortment in a store?
In a convenience store, in an online shop or in the retail sector, the assortment plays a decisive role. Indeed, it is thanks to the assortment that a point of sale makes its turnover. The assortment is the list of all the products offered for sale.
We can immediately see the challenges linked to the assortment. To be efficient and effective, the composition of the assortment must be ideal: no out-of-stock items, no overstock, no overpriced products, etc.
The assortment of a sales outlet can vary from one period to another. For example, the assortment can be different in summer and in winter. Similarly, it may vary during vacation periods or for special promotional offers. Finally, several stores within the same group may opt for a different assortment. The geographical location can influence the composition of an assortment: a sales outlet located in the city center of a regional capital will not offer the same range as an outlet located in the countryside.
To summarize, the role of the assortment must meet the "5R" rule stated by Charles Kepner in his 1963 book "Modern Supermarket Operations":
- the right product;
- at the right time;
- in the right place;
- in the right quantity;
- and to the right customer.
Finally, a few years later, Charles Kepner added a sixth "R": the right information. For him, the customer must find the information he needs on his own in order to immediately understand where his product is, how it works, etc.
If a store's assortment meets all of these needs, then it will be competitive, efficient and will generate sales and revenue. But being able to offer the right product in the right place at the right time to the right customer in the right quantity is not easy. Different parameters have to be taken into account to provide reliable predictions that will allow to compose a performing assortment.
How to define the assortment forecast?
After having detailed the definition of the assortment, it is appropriate to ask ourselves how to achieve an efficient assortment. Indeed, all companies have to think about the assortment of their stores.
- Which products should be selected from the assortment?
- In what quantities will the different items be sold?
- At what time of the year is it best to highlight them?
- Which prices or promotional offers trigger sales?
- What are the new consumer trends?
Fortunately, every company can rely on an infallible tool to compose its assortment: the assortment forecast. But what is Assortment Forecasting? Assortment forecasting is an innovative modeling solution. The implementation of this process allows a company to determine the evolution of its assortment over time.
Thus, thanks to machine learning, a company knows exactly how to compose its assortment for the first week of February, for the Easter school vacations, for the last three months of the year, for the back-to-school weekend, etc.
The three steps of assortment forecast optimization
If assortment forecasting appears to be an essential part of a business strategy, how should a company proceed to optimize it? To optimize the assortment forecast, it is necessary to rely on artificial intelligence. Indeed, artificial intelligence allows to predict sales and therefore to optimize assortment forecasting. Here are the three steps to follow to successfully optimize your assortment forecast.
Collection of internal data related to the sales of a point of sale
The internal data of a store comes from different sources. Indeed, depending on whether it is an online or a physical store, different types of data can be collected and can feed the predictive intelligence platform.
The first group of internal data concerns sales receipts. Common to all sales areas, they can provide a lot of information about customer purchases:
- which products are purchased the most, in what quantities and at what prices;
- the time of day, week or month when purchases are made;
- the average basket, the number of sales and the turnover;
- additional products or complementary sales;
- purchase frequency;
- the sales force present, the number of salespeople present in the team;
- products purchased only on promotion;
Next, companies can analyze backorders, open orders, cancelled orders, abandoned carts, etc. This data is more relevant for online stores and also informs the marketing department about the strengths and weaknesses of the assortment.
All of this internal company information is relevant information about the behavior of customers in relation to the assortments offered in the stores. By analyzing its sales receipts, a company will know which assortments are the most effective and profitable.
To optimize assortment forecasting, it is therefore necessary to start with this first step: collecting the company's internal data. As we will see later, this data will not be processed manually, but analyzed by a very high accuracy forecasting platform. But before we look more closely at how it works, let's discuss the second type of data that the platform needs to refine and personalize its assortment predictions.
Collection of external data to characterize customer behavior in a store
External data is the second type of data that can be used to feed the forecasting software. This external data is necessary to contextualize the purchases and the quality of the offer. Unlike the first group of data, called "endogenous" variables, these new data are "exogenous" variables.
The "exogenous" variables include:
- the weather;
- road traffic;
- competitive density;
- the company's market positioning;
- average income and purchasing power;
- events organized around the point of sale;
- legislation in place;
- the inflation rate;
In the case of the health crisis linked to the coronavirus, other variables had to be taken into account, such as regional decontamination, the opening of schools, travel constraints, etc.
These external variables therefore include all the information capable of explaining and contextualizing the behavior of customers in a specialized store, a retail store, a mass distribution store, an online sales site, etc.
The role of machine learning in assortment forecasting
The role of machine learning is to analyze all the data in order to provide estimates on the buying behavior of customers in the short, medium and long term future. This means that for each point of sale, the predictive intelligence platform is able to provide sales estimates for each product.
With this software, the company can implement an optimized planning of its stock levels. Inventory management plays a crucial role in the supply chain. Assortment forecasts give the company the possibility to optimize its supply chain and stock management.
Also, with these assortment forecasts, the company adapts its stock level according to the period of the year and according to its objectives. Logistics teams can anticipate needs and negotiate better purchase prices with suppliers.
Making assortment forecasts means being reactive and competitive. In terms of reactivity, it is important to know that new data can be regularly integrated into the predictive intelligence platform. The new forecasts are provided in real time, which allows the company to adapt and optimize its assortment plan in a few moments. This guarantees that the company will remain competitive, regardless of external events.
In conclusion, to make assortment forecasts, it is advisable to place the customer at the center of the analysis. Indeed, making assortment forecasts means analyzing the customer's behavior, his needs and his objectives. The predictive intelligence platform is a software able to provide reliable and very precise data on the customers' buying behavior and on the socio-demographic context. Thus, each point of sale will be able to compose an assortment that precisely meets the needs of its customers.
We recommend these other pages: