Anti-fraud: how to save 100 million rubles in retail

Aim

To analyze the data and establish control over sales, product movement, use of discounts and loyalty programs. Stop abuse and fraudulent activities.

Context:

A large federal retail chain contacted Marketing Logic. The company, due to its large scale, extensive network, and turnover of goods and funds, lost up to 100 million rubles a year in total due to fraudulent operations, abuse of loyalty programs, hiding product shortages, and other less common deception schemes. The company's security and financial monitoring service, due to the large volume of transactions, identified only a part of dubious cases and in the vast majority (almost 100% of cases) already after the fraud was committed.

Solution

The main task of the project was to develop software for operational monitoring of operations and timely notification of the company's financial monitoring and security services about potential fraud or scam. Technically, the project consisted of 3 stages: the collection and analysis of a large amount of data on network operations with the allocation of a separate segment of all suspicious ones. Creating neural networks and training them on the collected data. The third stage was the implementation of the system and integration with the existing software in the company.

During the project, the joint team of the retailer and Marketing Logic analyzed the sales and product distribution operations, as well as the document flow that accompanies them. To identify suspicious transactions, a list of cases was compiled, on the basis of which control rules and algorithms were formulated for analyzing the array of transactions from the internal IT systems of the trading network. As we analyzed and configured the system, we prepared and optimized more than 50 cases for continuous use. In other words, they can be called types of fraudulent schemes that the system instantly recognizes and notifies of potential danger. Data analysis at the stage of AI preparation and training allowed us to adjust business processes and remove a number of vulnerabilities from them. In case of a suspicious transaction, additional control actions were provided.

After implementation, the system for processing and analyzing all business processes related to the movement and sale of goods allowed us to identify tens of thousands (!) of cases of abuse of official authority and violation of instructions, which allowed the company to save from 60 to 100 million rubles a year.

Further analysis of the existing Antifraud system showed the possibilities of the next step: adding new data sources, for example, integration with video analytics systems, allows you to refine existing cases and develop new models for predicting fraudulent actions. This revision will allow the client to save an additional 10-15 million rubles annually.

Result

The Marketing Logic product has reduced the risks of fraudulent transactions, strengthened control over the commodity and document flow, financial flows and actions of the company's employees. The combined effect allowed the retailer to save up to 100,000,000 rubles in the first year of use.

"The damage from internal fraud can be both obvious – in the form of overestimation of discounts due to the use of bonus and discount cards, fraud with refunds, the sale of expensive goods under the guise of cheap, "virtual" provision of services, and indirect – in the form of overestimation of personal sales or time worked, concealment of shortages, holding the goods until the discount period, as well as in the form of many other more or less common variants of deception of the company. There are separate cases related to compliance with service technologies, internal regulations and the reflection of operations in systems that do not cause direct damage, but significantly increase the load on the back office», - Dmitry Galkin, Managing Partner of Marketing Logic, talks about the implicit potential risks that the Antifraud system is also focused on.