AI & product classification for pricing in cosmetics sector

For a consistent and controlled price image

HThis document gathers insights from our team of Big Data and Pricing experts experienced with European retail leaders. Our mission is to solve complex data problems that directly damage the profitability and price image of retailers, developing cutting-edge technological solutions based on the latest advances in AI and machine learning. This document gathers insights from our team of Big Data and Pricing experts experienced with European retail leaders. Our mission is to solve complex data problems that directly damage the profitability and price image of retailers, developing cutting-edge technological solutions based on the latest advances in AI and machine learning.

The aim of this article is to address a major issue which pricing and marketing managers come across in cosmetics and parapharmaceutics when segmenting their catalogs. How can we work at the pace of the market and maintain a segmentation - and therefore a pricing - that is consistent with the evolution of the catalog and trends?

If a brand is focused on consumer perception, then products’ classification and segmentation according to common characteristics (type of product, capacity, etc.) or value attributes (brand perception, innovation, current trend, etc.) is an essential first step for the industrialization of effective pricing. Classifications are largely used by marketing teams. But when pricing teams need to use them they come across a few obstacles.

Pricing complexities due to market volatility

Aside from the complexity generated from catalogues containing numerous products, the care and the beauty sector face an even more challenging obstacle because of the volatility of the market. This sector of activity is extremely dependent on:

trends in fashion & beauty magazines
a high rate of incoming brands or e-commerce sites
publications and opinions shared by specialized skin care and beauty influencers

A product that has been commonplace for years can become trendy overnight. A single high profile publication in the press or an article written by an influencer can launch a new cosmetic / skincare trend. Online magazine Glossy sums it up in one of their articles: “TikTok’s army of skin-care influencers have built up devout followings and have the ability to make or break sales for brands” CeraVe is arguably the brand which benefited the most from TikTok’s popularity. Tik Tok’s algorithm highlights content based on your reactions to previous content, regardless of who you follow. The algorithm found that content from influencer @skincarebyhyram was performing well, so it showed his content to a large audience which in turn boosted the visibility of his favorite brand CeraVe.

This unpredictable environment is causing the marketing and pricing teams of beauty retailers to constantly redefine established classifications. However, retailers in this sector indicate that these classifications are still done manually and often “by using common sense”. The consequence? Outdated and potentially unreliable classifications which could damage product price positioning.

Why are there problems with the coherence of the offer?

Generally speaking, traditional tools (Excel, BI, etc.) require you to choose between exhaustiveness and speed of work.

They do not allow you to work simultaneously on classifications for the entire catalog when it holds tens or hundreds of thousands of references. Teams must therefore cut up the product catalog. This is a waste of time and it damages the overall consistency of product pricing. Another solution would be to exclude certain product groups, which would make the strategy less precise. However, limiting the analytics to certain categories due to a lack of technological tools means running the risk of missing out on untapped margin potential or of being inconsistent in the eyes of the consumer in relation to the brand’s overall positioning.

Given the size of the catalog and the classification data, users are not able to work at market pace. Working in real time implies having automatic classifications on the one hand, and on the other, being able to integrate and modify existing classifications, on-the-fly. For these adjustments to be relevant, they need to be integrated into the strategy right away.

To strike the right balance between exhaustiveness and speed, the tools most used by retailers today focus on the three main criteria that define a product and are systematically available in the repository: the type of product, the brand and the size (or capacity in the case of health and beauty products). Integrating additional criteria specific to each product family and based on a repository that is not always complete is a real technological challenge in managing large quantities of data. However, from a strategic point of view, going beyond this would enable those in charge of pricing to classify products more accurately by integrating value attributes and not just product characteristics. This approach is very important in cosmetics where the consumer’s choice is not limited to “basic” criteria. For example: “shampoo, Garnier, 250 ml, dehydrated hair, general use” are already much more refined characteristics than “shampoo, Garnier, 250ml”. We can also go further with value attributes for the customer by adding, for example, “novelty, self-cocooning, sulfate-free formula, natural ingredients, local production, Instagram popularity” to the previous classification.

To what extent is AI relevant for catalog classification for cosmetic retailers?

We keep hearing about AI but we don’t always know how to use it and get the full value out of it. AI will allow teams to automate existing product classifications. By relying on the existing product database, it is possible to train an AI to recognize several types of products and to associate attributes to them. The algorithm will be able to instantly classify a product such as “Lipstick, Dior, red, matte, grape, Christmas edition” from a single photo or product description submitted to it and this product will be directly integrated into a “high sensitivity” pricing strategy during the Christmas period.

As we said before, unpredictability is key in the cosmetics industry. Pricing managers, product managers or category managers need tools to work more efficiently, but above all, they need to react very quickly. Mercio is committed to provide retailers with this agility. Color choice is a key element in cosmetics, whether it be lipstick, blush, foundation, etc. If nude suddenly becomes the trend, the pricer will be able to instantly update the “color of the moment” into the classification, and maintain consistent pricing.

A classification tool must therefore provide a clever mix between automation thanks to AI and a manual “corrective” part. This means allowing the person in charge to keep control of the AI by rectifying or sending new information related to an evolution on the market, whether this be an incoming product, a new trend etc.

How do you increase performance on classifications and pricing?

Data science is a great starting point to creating value insights for a brand that does not have an existing classification. If the retailer already has classifications, data science will help gain accuracy by verifying the validity of the classifications in place and by enriching them.

Receipts constitute a massive volume of data for retailers, and this data is continually being renewed. This is not an obstacle for AI but rather provides it with a great opportunity to learn. By analyzing sales receipts, a machine learning algorithm can identify purchase behaviours and links between products that are very difficult to perceive “with the naked eye”. For example, cases where an increase in the price of an electric sander would lead to an increase in manual sanding equipment of the same brand. Once the classification is identified and automation is in place, teams can "feed" the AI algorithm with new information. This article shows a first example of classification. All the data available to the retailer will allow him to gain finesse: store receipts crossed with web shopping carts, web visits, and insights into loyalty programs, for example. Thus, new classifications can be identified: "routine" product, impulse product, etc.

In conclusion

AI and business experts form a team, they feed each other. AI enables business experts to make swift classifications. Business experts, in turn, send new information to the AI to adapt to a given context which could include incoming products, a competitive and fluctuating environment or new trends.

AI eliminates the time-consuming management of classifications over time and teams can focus on optimizing pricing strategies based on different classifications, performance indicators and the retailer’s business strategy (e.g., optimizing private labels, favoring the web, targeting new customers, etc.). A pricing tool that is powerful and flexible enough to integrate these classifications in real time allows us to test the efficiency of a classification, compare several classifications, and change the strategy very quickly if necessary.