How to optimize assortment with predictive models?
Companies specialized in retail or evolving in the sector of large-scale distribution have the same goal: to substantially increase their turnover. While this is legitimate in view of the investment made, it is not necessarily automatic. To achieve a high level of revenue, they must continue to refine their marketing strategy. This means optimizing stocks, promotions and pricing policies, as well as optimizing their product assortment. Such performances are not easy to achieve, unless you use predictive models. Let's take a look at how they work to understand how they can help a company optimize its product assortment.
A good assortment strategy to improve the customer experience
What is an assortment plan?
The assortment is defined as a set of products or references arranged in a store or in a catalog and which adequately meets the expectations of the target clientele. The assortment plan is the process that will lead to the selection of key products. Building an assortment plan means, for a company, to define the quantity of stocks, the ranges to be referenced, as well as the width (number of categories), the depth (number of items contained in the same category) and the breadth (total number of references offered) of a product assortment.
The variables are determined by the DNA of the store, but also by the competitive environment in which the store operates. Although essential, these two criteria are not sufficient. The profile of the target customer must also be taken into account when selecting the products to be sold online or in a physical store.
In combination with other factors (product prices, geographical location of the point of sale, use of digital geolocation tools to inform people in the area, promotional offers), a well constructed assortment plan:
- increases the number of visitors to a store or the flow of visitors online;
- increases direct conversion opportunities; and
- strengthens relationships with consumers.
While building an assortment plan is an important step in the success of any business strategy, it is critical for a company to manage this step well so as not to face a significant drop in sales.
How to define a good assortment strategy?
Managing your assortment strategy means that you have to offer your customers products that they like. The key to identifying the products that should be part of the assortment is to know enough about the target customer. It is therefore important to gather as much information as possible about the the users' profile of a point of sale or a store. This makes it easier for the company to accurately predict which products will be purchased.
With its resources, a company is able to efficiently manage the collection of information. Multi-channel contact strategies (website, phone, email, mobile application, point-of-sale questionnaires, social networks, etc.) are certainly used for this purpose, but there is an even more reliable way to access basic information about a customer: the receipt.
On this piece of paper, you can find essential information such as:
- the number of products purchased;
- the number of products per category;
- product varieties;
- shares of premium products;
- date and time of purchase;
- the total amount;
However, collecting customer information is the first step. To take advantage of the data collected, it must be processed and analyzed. Processing the information provides the company with statistical data that is easy to interpret and that will allow it to define an effective assortment strategy for its various outlets.
While the information contained on a receipt can guide the company in selecting the products that work best at a particular location, the use of a predictive model will allow the company to establish real buying behaviors and make predictions with a high level of accuracy.
Predictive marketing announced as the future of sales and marketing
Predictive marketing has become increasingly popular in recent years due to advances in processing technologies. Predictive marketing allows companies to stand out from the competition, accelerate consumer conversion and improve the customer experience. It is therefore with good reason that companies have adopted this marketing strategy to manage their activities. What does it actually involve?
With a predictive marketing strategy, a company can anticipate customer needs in order to offer them products and/or services likely to interest them, as well as personalized promotions (discounts, gifts, additional services, etc.). To anticipate a purchase, the company uses behavioral data that it has collected on each customer (with their consent) through different contact channels. This data is then analyzed and processed using different approaches.
Although the use of classic segments allows to obtain some statistical data, the accuracy rate they show is nevertheless quite low. Making predictions based on approximate data is not necessarily in the interest of large retailers who are primarily looking to increase their sales.
To optimize the efficiency of their predictions, most of them favor technologies based on artificial intelligence (AI) that exploit weak signals to provide reliable and accurate statistics on a customer. Machine learning is the most commonly used algorithm because of its ability to process a wide range of data in record time.
This is not the only benefit of this technology. With machine learning, companies are able to discover the right moment in the customer journey to link purchase intent to action. With the right offer, they are twice as likely to identify high-value customers and get a quick return on investment.
Why is the use of continuous learning technology so beneficial to an organization in managing its product mix?
Machine learning supports good assortment management
In the marketing area, the applications of machine learning are varied. The algorithm developed by this technology not only allows to optimize promotions and advertising campaigns, but also favors the increase of sales thanks to a segmentation of the clientele by identifying the products and/or services that concern each user. The identification by this algorithm of the items that consumers are most likely to buy, is very beneficial to companies that will be able to better optimize their product assortment.
In concrete terms, machine learning will proceed to a cluster (separation of data into homogeneous groups) in order to highlight the structure of an individual or collective behavior or to a discriminant analysis.
In both cases, the data processing will consist of two phases that complement each other quite well:
- a learning phase; and
- a forecasting phase.
The learning phaseDuring this phase, the machine will learn the logic of the model it must integrate. It is a matter of providing it with data that it will analyze. The data can come from internal sources (receipts, loyalty cards, etc.) or external sources (clouds, social media, mobile terminals, etc.).
This learning phase can be supervised or unsupervised. In either case, it will result in statistics that can be measured and interpreted by human intelligence. The more complete the learning, the more accurate the predictions obtained will be.
The forecasting phase
After the learning phase where the machine analyzes a more or less important flow of data, it will be able to draw quite precise predictions. Thanks to the information in its possession, the machine will target the possible assortments for each point of sale by defining the width, depth and density of each assortment and by taking into account the financial objectives set.
For example, the results of a machine learning can be a three-year statistic on the number of items purchased in a given category by senior citizens.
Thanks to the machine learning technology and its continuous learning algorithm, a company can thus:
- Determine appropriate items for a store;
- Develop targeted purchase quantities within revenues and units;
- prepare manufacturing demand plans;
- schedule total sales;
- Plan inventory plans prior to a season or by a period within a season;
- implement flow and receipt strategies;
The optimization of the assortment by means of a predictive model, does not only concern the sales forecast, it also takes into account the planning of receipts. By using machine learning, a company is therefore able to plan the purchase of articles and to choose the ideal period to send order orders to the order managers for processing.
In conclusion, assortment optimization increases revenue and improves customer service. By using predictive models, a company is more likely to succeed with its assortment plan and achieve its financial goals.