Tinkoff RECO

Main article: Artificial intelligence in retail

2020: Tinkoff RECO start

On September 29, 2020 the Tinkoff company reported that it developed and started in a pilot stage own technology of an algorithmic cashback with referral models — Tinkoff RECO. The technology is responsible for selection for clients of an individual cashback on goods, brands, purchases in shops and in different commodity categories. According to representatives of Tinkoff, the algorithmic Tinkoff RECO technology for September, 2020 has no analogs in world practice.

Tinkoff RECO

According to the company, Tinkoff RECO is family of modern AI algorithms (Artificial Intelligence) trained at purchases of 8 million clients of Tinkoff within 2 years in different categories (FMCG, restaurants, the equipment and electronics, clothes and so forth). RECO foresees on the basis of the history of transactions of the client that the person in the future will want to purchase and can offer the buyer an individual cashback on goods necessary for it.

The architecture of technology allows to calculate different models of probability of purchase for a specific brand, shop or goods and to decide whether it is necessary to give a cashback on a specific brand or goods. In process of implementation of Tinkoff RECO in the Tinkoff ecosystem clients will be able to receive offers on goods and specific brands in the section Cashback in mobile application of Tinkoff. For September, 2020 offers with a cashback to 30% for goods and brands of several tens partners of Tinkoff - producers and merchants are already available to a part of clients.

What more people buy according to the card by, that the algorithm learns about his preferences more, and that high probability that it will offer it the most suitable offers and a bigger cashback. We are going to scale this technology so that clients received that cashback which suits them most of all. Vera Leychenko, the chief of the department of the deposit and settlement products Tinkoff told

The RECO technology also helps shop or a manufacturing brand to define from among partners of Tinkoff: what goods sell and what are not present whether it is necessary to start on goods or a brand offline what client needs to offer other type of goods and to what on it to give a cashback.

Tinkoff RECO

Tinkoff RECO is in addition capable to collect automatically a basket of purchases which will be necessary for the client in the nearest future — depending on his requirements, habits and a consumer profile. For example, having analyzed purchases of the client (a diet, the amount, frequency, frequency of purchases and so forth), the algorithm can define that this week it is time for client to purchase these or those goods — for example, laundry detergent or chicken fillet.

Regular, basic goods can be added to cart automatically, a part of products can be added as the recommendation. For example, if the client often eats beef — the algorithm can suggest to replace it with chicken to expand a diet. Also Tinkoff RECO can advise the person to buy more fruit and greens if these products meet in its basket very seldom. The Tinkoff RECO technology is continuation of development of ReceiptNLP service for interpretation of text information from trade checks using neuronets. Service can find the brand name in the text of the check, define up to 70 types of goods, decrypt the reduced name and recognize a product. Possibilities of service are already applied in Tinkoff, for example, for carrying out researches about expenses of Russians in shops or annual drawing up the Advent calendar from which clients learn — how often bought these or those goods what they spent most of all money for in year how many and what bonuses received and so forth.