Anton Mironenkov – “If bananas are not sold, then something is wrong”

Managing Director of X5 Technologies Anton Mironenkov told how artificial intelligence helps to predict our purchases and where the company finds the most promising technologies

The pandemic is the engine of progress

— What is innovative retail today? And how has the perception of it changed over the past few years?

— This is, first of all, the internal culture that is developing in retail companies — the willingness to constantly do something new, change and optimize internal processes, come up with various interesting things for customers. And what we are seeing today is seriously different from the approaches five years ago.

The teams that are engaged in digital innovation are no longer concentrated in the IT department, but are located inside the business functions – operational, commercial, logistics departments. After all, when you introduce something new, it is important first of all to understand what the buyer expects from you and how all the processes work. Therefore, in the corporate culture of X5, the role of the owner of a digital product, which determines the vector of development of platforms that set the rhythm of the company’s processes, is becoming increasingly important.

In addition, the rate of change in business has increased dramatically. Five years ago it was possible to introduce something, and for another three years it remained a unique development that no one else has. And now you just made something new, introduced it to the market, and in six months all competitors have it.

In such an environment, of course, it is very interesting to live, but not very easy, because the race for innovation in retail goes on without a break.

— How has the pandemic affected the technological development of retail?

— She pushed to be more progressive in the introduction of new technologies. We understood that there was no time to wait, we just had to go and do it.

A vivid example is the speed of connecting our stores to delivery services. If earlier we connected from one to three outlets per month, then last year the pace reached dozens of stores per day.

As a result, the volume of online sales of X5 in 2020 amounted to more than 20 billion rubles. This is four times more than in 2019. Moreover, the demand that arose against the backdrop of the coronavirus remained even after the restrictions were lifted. People have tried a new way of buying products and continue to use it.

— What was the most difficult for retailers in adapting to pandemic realities?

– The main difficulty was that at first everything happened at once. Buyers massively bought goods in stores and also massively ordered online, assemblers rushed around the trading floors and tried to form orders. In parallel, the software was debugged, bugs and crashes were eliminated. Optimization and change of processes was required, because a delay at any of the stages could result in hours of waiting for the client.

Along the way, we had to address health security issues that came to the fore last year. In addition to the mandatory antiseptics, masks, disinfection of premises, technology also played a role here. To avoid the need for customers to stand in line, we have accelerated the installation of self-service checkouts (more than 6 have already been installed), introduced the ability to scan goods from a mobile phone and pay for it in the Express Scan mobile application.

Ten years before Amazon

– It turns out that the technologies necessary to work in a pandemic were already available, they only needed to be launched or scaled up. Were any fundamentally new technological solutions introduced last year?

— It takes time to create new complex products. It often takes more than a year from the start of their development to the final launch.

For example, assortment planning is a rather complicated technology. Especially considering that we have many regions, types of stores, and the preferences of buyers in different locations differ.

During the pandemic, we simply would not have had time to create and launch a product of this level of complexity. But we launched a digital transformation back in 2018, when no one was counting on the coronavirus. Therefore, when the pandemic began, we already had ready-made solutions on the way that helped improve work.

One example of the launch of technology during the corona crisis is the Express Scan service. These are contactless secure purchases using a mobile phone based on the usual Pyaterochka and Perekrestok. A cross-format team of more than 100 people launched this project in just a few months, and, bypassing the pilot stage, we immediately moved on to scaling. Today, the service operates in more than 1 of our stores.

— How do you assess the level of digitalization of Russian retail in general?

— We in the company discussed for a long time how to correctly compare ourselves with others and understand whether we digitalized well or badly. As a result, we came up with an internal indicator – the digitalization index, which covers a fairly large number of factors.

Global leaders in digital services such as Amazon have achieved over 80% performance. But in e-commerce there are no physical processes that we have. Digital marketplaces do not need to change the price tags on the shelves – just change them on the site.

It will take us about ten years to reach this level of digitalization. But this is provided that the same Amazon will stand still. At the same time, if the same digital giants decide to go offline, they will have to “catch up” with our level of competence.

— In any industry there are underestimated and overestimated technologies. In your opinion, which technologies are undeservedly overlooked by retailers, and which ones are overestimated?

— In my opinion, technologies that allow you to plan and manage operations in the store through task management are greatly underestimated. So far, a lot here depends on the experience and knowledge of the director: if he notices any shortcomings or deviations in the work, he gives the task to correct it.

But such processes can be digitized and automated. To do this, we implement algorithms for working with deviations.

Sometimes not statistics is used for this, but image recognition, video analytics. The camera looks at the shelves, checks the availability and volume of goods and warns if it is about to run out. Such systems help to allocate the time of employees more efficiently.

If we talk about overvalued technologies, then I would mention electronic price tags. Of course, they are convenient and allow you to change prices more often without the physical participation of a person. But is it necessary at all? Maybe you should come up with a different pricing technology. For example, a system of personalized offers, with the help of which the buyer will receive goods at an individual price.

Big network – big data

— What technologies can be called decisive for retail today?

“The maximum effect now is given by everything related to the assortment, its automatic planning depending on the type of stores, location and environment.

Also, this is pricing, planning promotional activities, and, most importantly, sales forecasting. You can make the coolest assortment and the most advanced pricing, but if the right product is not in the store, then customers will have nothing to buy. Given the scale – and we have more than 17 thousand stores and each from 5 thousand to 30 thousand positions – the task becomes quite difficult. You need to understand what and at what moment to bring, take into account different areas and formats of stores, the situation with roads, expiration dates and many other factors.

– Is artificial intelligence used for this?

— Yes, the task of forecasting sales is no longer solved without the participation of AI. We are trying machine learning, neural networks. And to improve the models, we use a large amount of external data from partners, ranging from the congestion of the tracks and ending with the weather. Let’s say that in summer, when temperatures are above 30 ° C, sales of beer, sweet soft drinks, water, ice cream jump sharply. If you do not provide a stock, the goods will run out very quickly.

The cold also has its own characteristics. At low temperatures, people are more likely to visit convenience stores instead of large hypermarkets. Moreover, on the first day of frost, sales usually fall, because no one wants to go out. But on the second or third day, we see increased demand.

In total, there are about 150 different factors in our forecasting model. In addition to sales data and the already mentioned weather, these are traffic jams, store environments, events, competitor promotions. It would be unrealistic to take all this into account manually.

— How big data and artificial intelligence help in pricing?

— There are two large classes of models for making pricing decisions. The first is based on market prices for a particular product. Data on price tags in other stores is collected, analyzed, and based on them, according to certain rules, own prices are set.

The second class of models is associated with building a demand curve, which reflects the volume of sales depending on the price. This is a more analytical story. Online, this mechanism is used very widely, and we are transferring this technology from online to offline.

Startups for the task

— How do you choose promising technologies and startups in which the company invests?

— We have a strong innovation team that keeps abreast of startups, monitors new technologies.

We start from the tasks that need to be solved – the specific needs of our customers or the need to improve internal processes. And already under these tasks solutions are selected.

For example, we needed to organize price monitoring, including in competitors’ stores. We thought about creating this technology within the company or purchasing it. But in the end, we agreed with a startup that provides such services based on its price tag recognition solutions.

Together with another Russian startup, we are piloting a new retail solution – “smart scales”. The device uses AI to automatically recognize weighted items and saves about 1 hours of work for cashiers per year in each store.

From foreign scouting, the Israeli startup Evigence came to us with a solution for product quality control based on thermal labels. In the first quarter of this year, such labels will be placed on 300 items of X5 Ready Food products, which are supplied to 460 Perekrestok supermarkets.

— How does the company work with startups and what stages does it consist of?

— To find companies for cooperation, we go through various accelerators, we cooperate with Gotech, and with the platform of the Moscow government, and with the Internet Initiatives Development Fund. We are looking for innovations not only in our country, but also in other countries. For example, we work with the Plug&Play business incubator and international scouts — Axis, Xnode and others.

When we first understand that the technology is interesting, we agree on pilot projects. We try the solution in our warehouses and stores, look at the result. To evaluate technologies, we use our own A / B testing platform, which allows you to clearly see the effect of a particular initiative, compare with analogues.

Based on the results of the pilots, we understand whether the technology is viable, and we plan to launch it not in 10-15 pilot stores, but in the entire retail chain.

Over the past 3,5 years, we have studied about 2 different startups and developments. Of these, 700 reached the scaling stage. It happens that the technology turns out to be too expensive, more promising solutions are found, or we are not satisfied with the result of the pilot. And what works great in a few pilot sites often requires huge modifications to be rolled out to thousands of stores.

— What share of solutions is developed within the company, and what share do you buy from the market?

— We create most of the solutions ourselves — from robots that buy sugar at Pyaterochka to unique multifunctional data-based platforms.

Often we take standard boxed products – for example, to replenish stores or manage warehouse processes – and add them to our needs. We discussed assortment management and pricing technologies with many developers, including startups. But in the end, they began to make products on their own in order to customize them for our internal processes.

Sometimes ideas are born in the process of communication with startups. And together we come up with how the technology can be improved in the interests of the business and implemented in our network.

Moving to smartphone

— What technologies will determine the life of retail in the near future? And how will the idea of ​​innovative retail change in the next five to ten years?

— Now online and offline in grocery retail work as two independent areas. I think they will merge in the future. The transition from one segment to another will become seamless for the client.

I don’t know what exactly will replace the classic stores, but I think in ten years they will change a lot in terms of space and appearance. Part of the operations will move from stores to consumer gadgets. Checking prices, assembling a basket, recommending what to buy for a dish chosen for dinner – all this will fit in mobile devices.

As a retail company, we want to be with the customer at all stages of the customer journey – not only when he came to the store, but also when he decides what to cook at home. And we intend to provide him with not only the opportunity to buy in the store, but also a lot of related services – up to ordering food from a restaurant through an aggregator or connecting to an online cinema.

A single client identifier, X5 ID, has already been created, allowing you to recognize the user in all existing channels. In the future, we want to extend it to partners who work with us or will work with us.

“It’s like creating your own ecosystem. What other services are planned to be included in it?

— We have already announced our subscription service, it is in the R&D stage. Now we are discussing with partners who can enter there and how to do it as conveniently as possible for buyers. We hope to enter the market with a trial version of the service before the end of 2021.

Consumers make decisions about the choice of products even before going to the store, and their preferences are formed under the influence of the media sphere. Social media, food sites, blogs, podcasts all shape consumer preferences. Therefore, our own media platform with information about products and food will become one of the channels of interaction with our customers at the planning stage of purchases.

Subscribe also to the Trends Telegram channel and stay up to date with current trends and forecasts about the future of technology, economics, education and innovation.