Machine as the only tool that deeply understands the shopper’s view on assortment

Author: Milan Havlíček, Sales Director Yieldigo

At first glance, it might seem to be very easy to set the right parameters and conditions for price optimization of goods in supermarkets. However during implementation, numerous concrete aspects that must be taken into consideration when setting prices begin to appear. These aspects can be: the nature of assortment (durable vs. fresh food vs. non-food), the character of the individual articles in terms of shopper’s perception, the physical properties of the articles (quality, pack size, meat content etc.) and, last but not least, seasonal influences and the origin (local vs. national origin). In this article, I aim to explain why the aspects underlying conclusions made by retail chains today may not be valid tomorrow.

Characteristics of assortment
 
When I say that fruit and vegetable differ from beverages, it is no surprise. From the point of view of price optimization, price managers have to look carefully at different characters of goods. The first aspect that differs is purchasing price stability, which is a key factor for margin optimization. While the price of beverages change once per year, the price of fruits and vegetables might change on a daily basis. The frequency of the change of purchase price is not a problem in itself. The problem is changes in purchase prices are not known sufficiently in advance and in many cases are not systematically captured. They might exist only in the heads of buyers, their emails, or excel files. From this point of view, the obstacle is a systematic nature of data inputs, not the assortment itself.

Fresh foods are, in general, frequently purchased by shoppers. They often appear in shopping carts. This is, of course, an advantage in terms of price optimization. It is logical — shoppers often buy yoghurt or milk, as compared to purchasing detergents. For performance of price optimization, the higher frequency of purchases and the higher penetration rate in shopping baskets are advantages. Price optimization tools have thus more opportunities to “learn” the nature of the assortment and to recommend the right price regarding the goal of optimization: increased profitability or increased turnover.

Another important role is the durability of the assortment. Fresh fruit or vegetables can remain fresh on the shelves for a few days, at maximum. While there is nearly no problem with expiration period of detergents. The impact on price optimization is simple — the question of selling goods at a reduced price before going bad or being disposed of is more likely to appear on fruits and vegetables.

Shopper perception of assortment
 
The way how the shopper perceives the assortment is the most important question. The promotion of certain articles plays, without doubt, an important role ; a frequent promotion significantly influences the awareness of expected price levels of the article. On the other hand, information gained from promotional activities can be used for optimization of non-promotional prices . During promotions some factors can be spotted, for example cannibalization among articles that are difficult to spot otherwise.

To a certain extent, the presence of competition limits the level of price changes. Supermarkets are forced to carry out a greater amount of minimalist price interventions. This many marginal interventions are followed by many marginal changes in sales. It is in principle very difficult to measure the real impact of concrete price changes. While price managers play a crucial role in promotions (promotions are often followed by more than a significant increase in sales), the price managers have a very challenging task with price optimization of non-promotional prices. They need an intelligent engine or tool which is able to reveal even subtle nuances. It is essential to add,the average retail shopper has a strong price awareness of approximately 100 articles. It represents only a slight share of the entire assortment. This does not mean that these articles should not be included in a portfolio of the tool carrying out automated price optimization; on the contrary, they must be there. The prices of the rest of the assortment can be optimized in regard to them. For the rest of the assortment, in Yieldigo, we can see interesting findings in regard to the assortment price elasticity. Our experience shows that for 27–30% of assortment, price increase leads to the increase of both sales and profit; this is called price rigidity. On the other hand, price elasticity means price decrease has the effect of increase in sales and profit, this can be observed at a 4–7% of assortment. These observations must be perceived dynamically from the point of view of the assortment and time in question. While beer may belong to the first group in winter, it is likely to belong to the second group in summer.

Physical characteristics of assortment
 
Should the price of goods on shelves follow the logic that a larger package should be about 5% cheaper compared to an appropriate multiple of the standard package size? Or just 4% cheaper? Or 7% cheaper? How do you find the answer? Is it desirable that articles differing only in flavor, e.g. different kinds of flavored mineral water of the same brand and the same package, have the same price? Should the articles with lower meat content or lower fat content in dairy products be automatically cheaper than those with higher content? From our observation on the market we can say there are two points of views on how to answer these questions. Firstly, one is based on the technological characters of the article, e.g. the production process. These are usually price managers and buyers working in the sector for a long time, and managing an assortment and its prices who stand for this opinion. The second one, is based on the shopper behavior and shopper’s preferences. Especially in dairy assortment, the price optimization engines face situations where shoppers prefer products with a lower fat content and are willing to pay more for such products. Similar behavioral patterns can be seen at a popularity of different flavors, some flavors are very popular by shoppers while others are nearly unsold. Then, there is no sense to keep the same price for these articles.
 
Who is Yieldigo?

 
Yieldigo creates an intelligent software plugin for retailers based on modern mathematical knowledge and can simulate shopper’s decision making process while purchasing goods and uses this ability for setting optimal prices. The basic principles, on which Yieldigo was built, are its own scientific research in the field of mathematical stochastic methods and technological scalability enabling Yieldigo to use efficiently great computing power. Based on retailer’s data, the machine learning engine learns automatically new patterns of shoppers purchasing behavior every day which guarantees its high flexibility and ability to respond to new situations on the market.

Yieldigo serves as a plugin needed to connect to the existing pricing system of the retailer and over a month the retailer will see a significant increase in profitability of more than 5%. It’s like to employ a pricing manager who has the best mathematical education, has unlimited working hours, understands perfectly shoppers purchase behavior and has many years of experience with the assortment. If a modern easy to use interface for pricing is added we get a package that is used by Yieldigo clients from retail industries like groceries, drugstores or pharmacies, brick & mortar retailers on the one hand and e-commerce chains on the other hand. Yieldigo’s strength is a quick integration of the plugin which takes just a few weeks.

If you are a small retailer and you are considering price optimization we have good news for you — it’s not just the matter of size of the team, it doesn’t even matter if you have one manager or fifty of them in your pricing team. A machine learning tool managed by one pricing manager is the right way.