![\"ARDFM](\"https:\/\/rbc-grp.solutions\/wp-content\/uploads\/2026\/11\/cover_arrfr_waves_en.jpg\")
<\/p>\n<\/p>\n
From 1 January 2027, Kazakhstan’s banks must give the regulator direct access to their data through data marts. The ARDFM draft rules are already published, and only weeks remain before the new year. Here is what is required, why “we’ll do it in December” won’t work, and how to turn an obligation into an asset the bank itself can use.<\/strong><\/p>\n The new law “On Banks and Banking Activity in the Republic of Kazakhstan” obliges second-tier banks to maintain data marts (in Russian, \u0432\u0438\u0442\u0440\u0438\u043d\u044b \u0434\u0430\u043d\u043d\u044b\u0445<\/em>). The launch has already been postponed once: instead of June 2026, the provisions take effect on 1 January 2027. Nobody has promised a second delay.<\/p>\n On 16 October 2026 the Agency for Regulation and Development of the Financial Market (ARDFM) published a draft resolution with the rules for building, maintaining and updating the data marts. Public consultation closed on 3 November. At the time of writing the document is still a draft, and details may change. But the outline of the requirements is clear already, and waiting for the final wording before starting work is like waiting for the first snow to buy winter tyres.<\/p>\n A data mart is a separate, structured database on the bank’s own infrastructure. The regulator does not take the data away; it queries the data mart through a centralized platform. The platform’s technology operator is the Digital Development Center of the National Bank.<\/p>\n The law names the basic data categories: the loan portfolio, large depositors, large payments, money transfers and cash withdrawals. The full list of fields will be set by the regulator in technical specifications.<\/p>\n Key requirements of the draft:<\/p>\n The requirements apply to second-tier banks and to branches of non-resident banks.<\/p>\n A report can be assembled by hand, checked by eye and sent by the deadline. A data mart does not work that way: it has to be current all the time, not by the fifth of the month. That changes the nature of the task.<\/p>\n A report is a photograph. A data mart is a live broadcast. In a photograph you can fix the lighting and crop what you don’t want. Live, everything shows as it is: mismatches between the core banking system and the loan module, reference data kept differently in every system, “temporary” manual adjustments now in their fifth year.<\/p>\n That is why a data mart project is 20% technology and 80% getting the data in order. Don’t blame the mirror.<\/p>\n The draft rules do not spell out specific sanctions for an unready data mart, and we are not going to invent fine amounts. But the price of being late is more than fines:<\/p>\n The main planning mistake is to count development time only. Before the first line of code, the bank goes through choosing a contractor, agreeing the contract and arranging access to its systems. Each of these steps easily takes weeks, not days.<\/p>\n Then comes the hard part: taking inventory of the sources, agreeing the calculation logic for each indicator, cleaning up reference data, setting up the loads, quality checks, load tests, and building the operations process. Measure twice, cut once is not a proverb here; it is the project schedule.<\/p>\n By our estimate, a turnkey data mart takes about eight weeks from contract signing, with some stages running in parallel. The arithmetic is simple: a contract signed in November still leaves a chance of a full launch by 1 January. Later than that, only a temporary solution is realistic by that date, to be brought up to the target design afterwards.<\/p>\n The conclusion: if the project has not started yet, every week without a decision shortens the plan. Putting it off until December means choosing the stopgap in advance.<\/p>\n A regulatory data mart is no reason to buy a closed “box” with a licence for years ahead. We build enterprise data warehouses and data marts on open technologies that the world’s technology companies run on. It is the rare case where fashionable and practical are the same thing.<\/p>\n What this gives an executive:<\/p>\n This is how data platforms are built at companies younger than most banks that process orders of magnitude more data. The good news: a bank does not have to be Netflix to use the same tools.<\/p>\n The good news: the data the regulator needs is the data the board needs. The loan portfolio, large clients and cash flows are the basis of management reporting, risk analytics and budgeting<\/a>.<\/p>\n If the data mart is built on a proper data warehouse, the same foundation covers several tasks at once: the regulatory data mart, management reports with no manual assembly, plan-versus-actual analysis and scenario planning in an EPM system such as IBM Planning Analytics<\/a>. Put the data in order once and everyone uses it, from the risk manager to the CFO.<\/p>\n Building a data mart for the regulator alone is like running water into the house and using one tap.<\/p>\n It is a structured database on the bank’s infrastructure to which the Agency for Regulation and Development of the Financial Market has secure access for supervision and early risk detection.<\/p>\n The provisions of the law on data marts take effect on 1 January 2027. At the time of publication, the rules for building and maintaining data marts exist as a draft ARDFM resolution.<\/p>\n Second-tier banks of Kazakhstan and branches of non-resident banks.<\/p>\n The law names the loan portfolio, large depositors, large payments, money transfers and cash withdrawals. The exact set of fields will be defined by the regulator’s technical specifications.<\/p>\n No. The draft rules require automatic loading from the bank’s internal systems and explicitly prohibit manual loading.<\/p>\n No. Primary data stays in the bank. The operator’s platform queries the data mart for each request and does not store primary data permanently.<\/p>\n A DWH is a warehouse where data from the bank’s different systems is gathered and brought to a single form. A data mart is a prepared slice of that data for a particular consumer, in this case the regulator. A data mart is most reliable when there is a warehouse underneath it.<\/p>\n An open stack: StarRocks and Apache Iceberg for storage and queries, dbt for calculation rules and tests, and the Data Vault 2.0 methodology for the data model. The warehouse is managed through a shell and a dashboard portal from RBC Group. The calculation rules and the data model are stored in the bank’s Git repository.<\/p>\n RBC Group builds data warehouses for banks in Kazakhstan on open technologies, is developing a solution for regulatory data marts, and implements IBM Planning Analytics for budgeting and management reporting on the same data. The core of our team has been building data warehouses for banks and financial organizations since 2002. If you need to assess your bank’s readiness for the 2027 requirements, start with a short audit of data sources and data quality: it will show the real scope of work and the timeline.<\/p>\n Contact us<\/a>: marketing@rbc-grp.solutions<\/a>, +7 727 258 1712.<\/p>\n This article is based on open sources as of 13 November 2026: publications in Kazakhstan’s business media about the draft ARDFM resolution (in Russian). It is not legal advice; once the resolution is approved, the requirements may differ from the draft.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":" From 1 January 2027, Kazakhstan’s banks give the ARDFM access to data marts. What the draft rules require, how long it takes and what to do now.<\/p>\n","protected":false},"author":6,"featured_media":6420,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[173],"tags":[],"class_list":["post-6409","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-stati-en"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\tWhat happened<\/h2>\n
What exactly the bank must do<\/h2>\n
\n
Why this is not “just another report”<\/h2>\n
What being late costs<\/h2>\n
\n
How long it really takes<\/h2>\n
What to do now: a plan for the coming weeks<\/h2>\n
\n
What we build warehouses and data marts on<\/h2>\n
\n
\n
How the bank itself benefits from the data mart<\/h2>\n
Questions and answers<\/h2>\n
What is an ARDFM data mart?<\/h3>\n
When does the requirement take effect?<\/h3>\n
Who does the requirement apply to?<\/h3>\n
What data must the data mart contain?<\/h3>\n
Can data be loaded into the data mart manually?<\/h3>\n
Is the bank’s data handed over to the regulator for storage?<\/h3>\n
How does a data mart differ from a DWH?<\/h3>\n
What technologies does RBC Group use to build warehouses and data marts?<\/h3>\n
How RBC Group can help<\/h2>\n
Sources<\/h2>\n
\n