Crossroads and GlowByte Consulting conducted an analysis of visitors' impressions in the store

2021: In-Store Visitor Experience Analytics

The Perekrestok retail chain and the company GlowByte completed a research project to analyze the impressions of visitors in the store. Glowbyte announced this on April 21, 2021. According to the results of the study, the trading network revised the internal processes that directly or indirectly affect NPS, and held special service marathons for all retail employees - about 28,000 people throughout Russia.

The purpose of the project was to determine the impact of various aspects of the store on customers.

The Crossroads and Customer Experience GlowByte teams began with a study of already accumulated data in the company based on the results of an NPS study, which was conducted through calls to customers after they visited food chain stores. Together with the NPS question itself, customers were interested in what they liked and did not like during the visit. After, customer comments were categorized into 10 main topics: assortment, quality of goods, convenience of visit, etc. It was necessary to determine the degree of influence of certain aspects of the store's work on customers. "Crossroads" and GlowByte solved the problem in several stages.

At the first stage, 58 ideas-triggers of customer emotions were formulated to digitize 10 main categories of customer impressions, for example, "The client noticed an expired product," "Could not find the product after permutations in the store. I don't like products being swapped. " Then, in the second step, data from different sources were collected by selecting the depth of data in 1 calendar year. Data warehouse tables and offloads from various systems were used as source sources, on the basis of which an attribute showcase for analytics was compiled.

After that, analysts calculated the distributions and correlations of the resulting set of metric factors and divided them into 9 groups by how they affect client perception. For example, there are "negative" factors: their growth leads to an increase in the proportion of negative emotions, and at the same time does not affect the change in the proportion of positive emotions.

As a result of the project, the teams received a number of insights for the business. For example, they noticed that men more often scold and less often praise various aspects of the work of shops. What can be interpreted as "if a man was sent to the store, then he is a priori dissatisfied with everything," or it is possible that "men on average tend to notice more negative aspects." But regardless of interpretation, this is actionable insight: if a large proportion of male customers go to the store, then perhaps it should be adapted a little to this audience. In many ways, this is confirmed by the long-known principle of one-size-does-not-fit-all: approaches to the client should be different depending on the segment (both by gender and by which generation the client belongs: boomer, X-generation, millineal or zoom).

To test the insights and hypotheses received, Perekrestok launched A/B testing in 60 supermarkets.

Customer orientation is at the heart of the entire strategy for the development of our network, "says Ivan Brattsev, head of the Client Experience Department of the Perekrestok retail network." The ultimate goal of this study and the subsequent testing is to assess how NPS affects the company's RTO. After that, we will be able to implement the most useful and relevant changes for the business. "

This case is interesting because when implementing, we did not dwell on the study of the NPS index alone as a metric of client satisfaction. We dug deeper into the analysis of customer experience drivers, which served as a source of deep insights for business, "emphasizes Yevgeny Pimenov, co-head of Customer Experience practice in GlowByte.

2019: Completion of the second phase of the Marketing Communications Optimization Project

On October 17, 2019, Perekrestok announced that it had completed the second stage of the marketing communications optimization project with GlowByte. This is the first solution to implement SAS Marketing Optimization (SAS MO) in retail.

According to the company, as part of the project, the teams developed a set of predictive UpLift models and models that predict the average check in the framework of the action for each campaign. They take into account more than 200 client attributes and calculate future profitability, so the forecast result allows you to select the best offer at the moment and increase the efficiency of communication with the client by 30%.

The key goal of the project was to increase the effectiveness of marketing communications. First, the campaign start was switched from manual mode to automatic mode using SAS MA. Then, having a pool of automated campaigns, it was necessary to build a system that sends the client the most profitable offer, provided that several campaigns are available to him. This was made possible by setting up SAS MO, which solves the problem of prioritizing offers at the client level, taking into account external restrictions and unique parameters.

The previous strategy of interaction with customers was based on manual selection and expert prioritization of marketing proposals. For optimization, models for predicting the response to different types of communications were used. They showed effectiveness, but the system lacked a method of prioritizing offers for an individual client. As a result, the consumer received one not always optimal offer for him. This reduced the likelihood of response and resulted in lower campaign conversions and increased costs.

The updated system is based on the calculation of the projected revenue from marketing communication in the context of each client. It calculates the probability of response to a sentence in different conditions:

• if communication is available;
• without communication.

Predicts the amount of customer purchases during the promotion period. This approach is complicated by the fact that it is necessary to obtain comparable results on each individual model. 'Crossroads' the first of the retailers in the Russian Federation solved the complex problem of optimizing targeted marketing with the help of SAS MO and received positive results.

The approach is based on the idea that the effectiveness of a particular communication will be different for each client, and it can be predicted based on a transactional history. For example, the model calculates that the effect of communication X will be higher than that of communication Y, although in manual selection, the manager can give communication Y a higher priority based on the results of previous launches.

Models that are previously run separately are now loaded into a single system, at the exit of which the optimal list of 'one client - one sentence' is formed. All this works in conditions of restrictions on the number of communications and bonuses provided by the loyalty program.

As the initial set of marketing offers, five marketing communications were used with the offer of an increased bonus in the selected categories of goods.

To evaluate the effectiveness of SAS MO optimization, the Crossroads and GlowByte teams launched two scenarios:

• Selection of the best offer for the customer according to business rules;
• Select the best offer for the customer based on SAS MO optimization.

Each scenario involved customers selected according to the same conditions and the same set of marketing offers.

The main criterion for comparing the two scenarios was the metric of the increase in the average turnover during the promotion period per participant. Based on this criterion, as of October 2019, the results of SAS MO were more than 10 times higher compared to the results of communication according to business rules and 30% higher than standard marketing communication launched on a regular basis. Also, the results in the section of each communication were higher when launched using SAS MO.

The goal of "Crossroads" ‒ to introduce a system of automatic selection of offers for each client by the end of 2019 and launch SAS MO at a larger volume of customers. The sales network also plans to expand the list of marketing offers for selection in the system. Further steps will be to add other types of communication to the system.

In 2019, we set ourselves an ambitious goal - to automate up to 80% of targeted marketing campaigns, as well as launch more than fifty regular campaigns, the selection of which is based on predictive models. We plan to implement the launch on the basis of SAS Marketing Optimization. We expect that this will help us automate the launch process and increase revenue from marketing campaigns by 15%. said Mikhailova Ekaterina, head of the monetization department of the client base of the Perekrestok trading network