How a New Platform Helped Rosgosstrakh Improve Data Quality for Business Processes

CNews: Viktor, was replacing foreign software with a domestic one merely a tribute to the import substitution trend, or was there more to it? And how did you select your vendor?

Viktor Bushmin: We weren’t looking for a one-to-one replacement for Pentaho. The point is that Rosgosstrakh’s business processes are data-driven — data are used by sales teams and portfolio managers across various products. In recent years, we’ve seen an explosive growth of data, alongside the growth of the insurance business itself. Many processes and IT solutions required scaling. At the same time, we faced restrictions on the use of foreign software, as well as occasional failures in data processing — all of which affect timely decision-making and business task administration.

Viktor Bushmin Rosgosstrakh Business processes at Rosgosstrakh are data-driven

We turned to Cinimex, with whom we’ve been collaborating in data analytics since 2019. We asked our colleagues to analyze how data are currently processed within Rosgosstrakh: which data enter the system “dirty” or duplicated, which don’t enter at all, and where processing failures occur. As a technical assignment, we provided the vendor with a prototype created through “vibe coding.” That script solved the problem, but we needed to industrialize the process. We required an open-source, Python-based system, fully documented, highly templatized, and transparent — so that anyone could easily trace where data come from, how they are processed, and how to add new nodes and columns. The solution also had to comply with the requirements of the VTB Group, of which Rosgosstrakh is a part, as well as with regulatory and federal laws.

Proper documentation was essential so that if a new IT partner came onboard after a procurement cycle, they could quickly understand the solution without interrupting business processes. Our primary goal was to increase performance and fault tolerance in data delivery for key business processes. We built the replacement for Pentaho to be accessible for Data Science specialists, who would maintain and expand its functionality. It was crucial that business units — not just IT — could independently develop the new solution. I believe no IT specialist can understand the meaning and purpose of data as deeply as the business team can.

The solution was developed in Python, which allowed us to rework the architecture, eliminating numerous temporary tables and complex SQL scripts. Early this year, we received the new platform, performed large-scale data reconciliation and quality checks, and in September, the system entered industrial operation — replacing Pentaho in the Rosgosstrakh IT landscape for data preparation processes.

CNews: What tasks does the system from Cinimex solve?

Viktor Bushmin: Everything related to data collection, processing, and analysis. We’re confident the new solution will enhance fault tolerance and improve data quality. It allows us to see the full data lineage — where the data come from, how they are transformed, and where they are stored.

The system also helps us improve customer service quality. The completeness of information about the client, the insured object, and our history with them is critical. For example, preliminary calculations for OSAGO or CASCO policies should give a favorable quote that won’t change drastically later. Various client factors are considered — driving experience, accident history, and so on. This means the client provides minimal information, while accurate and up-to-date data allow us to offer the best rates. This is possible because the decision-making and verification system checks client history across a wide range of available data. The platform guarantees data quality, which lets us focus less on data hygiene and more on service.

CNews: The implementation took several months. How complex is the system to maintain and administer?

Viktor Bushmin: We plan to manage the solution entirely on our own. Our data engineers have sufficient qualifications — especially when supported by AI and corporate chatbots. Cinimex will provide second- and third-level support, helping our team offload routine administrative tasks. Ease of use was crucial — we wanted colleagues from sales, operations, and marketing to see BI dashboards and reports anytime to track sales performance, marketing campaigns, and product opportunities. We deal with massive data volumes, so avoiding duplication and ensuring timely, accurate input into modeling and decision systems is key. With the Cinimex platform, we identified growth points in our processes and expanded our working data landscape.

Before the new platform, we lacked a tool that guaranteed data integrity during transformations. We delayed industrial rollout because we carefully compared data quality between Pentaho and the new system, discovering additional constraints along the way. Gradually, we scaled up to processing 24 tables and over 5,000 fields, with datasets ranging from 1 to 10 million rows. The reconciliation process was lengthy, but by September, the system proved fully reliable in production.

CNews: Rodion, how did you eliminate the need for temporary tables and complex SQL scripts? What technologies and patterns did you use for ETL?

Rodion Martynov: Our task was not to replicate the old system, but to refactor and improve it. To do so, we delved deeply into the business meaning behind every number — every figure represents a process. We even verified and preserved exact numeric precision up to the 12th decimal place where necessary.

Rodion Martynov Cinimex We adopted a modular approach, eliminating single-step processing of millions of records

We chose a modular orchestration approach — launching Python scripts in a Directed Acyclic Graph (DAG) format. It’s neither monolithic nor microservice-based, but a multi-component system with interdependencies.

The modular design avoids single-shot processing of millions of records. Instead, the data array is split into smaller chunks processed independently. We also introduced an internal database layer within the system to track every change across the entire data processing pipeline. Additionally, Cinimex moved business logic into transparent, maintainable ETL processes, improving data quality and simplifying support by eliminating redundant logic and duplicate elements. Data from different sources are now centralized in one tool, ensuring integrity and manageability.

CNews: What types of companies would benefit from this Cinimex solution?

Maxim Zharov: Data synchronization has become standard practice rather than a trend. Tools like DTE (Data Transformation Engine) are now mainstream, so our solution is essentially a custom-built system, not an off-the-shelf product.

Previously, such projects were seen only in large enterprises like Rosgosstrakh, but that’s changing — we’re now seeing growing demand from mid-sized businesses. Thanks to open-source technologies and cloud providers, even mid-sized companies can now build architectures comparable to large enterprises.

Maxim Zharov Cinimex The platform is not only developed by a Russian company but also runs on a domestic OS with Russian databases

In summary, the solution is most relevant for large enterprises with state participation. It’s developed by a Russian company, runs on a domestic operating system, and uses Russian databases — ensuring no potential compliance restrictions, even for organizations like Rosgosstrakh that belong to critical information infrastructure (CII) sectors.

It’s also suitable for the financial sector, where every decision relies on extensive numerical analysis. Finally, as more organizations — even small and mid-sized ones — undergo digital transformation and build data-driven processes, this platform becomes increasingly relevant across industries.

CNews: Can we say that the new platform has become a business management tool for Rosgosstrakh?

Viktor Bushmin: Absolutely. In the financial and insurance sectors, misinterpreted data can lead to massive losses — they are essential for forecasting and fraud prevention.

The Cinimex system has become our business control panel. Just as a driver monitors the speedometer, fuel, brake fluid, and oil indicators to make decisions on the road, our company relies on data indicators to make informed business decisions.

Read the full interview in CNews at the link