BI system: what it is, what it's made of, and why implementation comes down to data
A BI system is a set of software that pulls data from a company's operational systems — 1C (the accounting and ERP platform most Russian companies run on), the CRM, the bank, cash registers — brings it to shared master data and calculation rules, and shows metrics in reports and dashboards where you can drill down from a figure to the individual transaction. The charts the user sees are only the top layer. Beneath them sit extraction from the sources, a data warehouse and a data model that records what counts as revenue, a customer and a deal. These lower layers are where implementation gets stuck: the platform itself is installed quickly, while reconciling master data and agreeing on how metrics are calculated take up most of the project.
Below: what BI is in plain terms and which four layers it is built from. Why BI figures diverge from 1C and how to fix it. Which platforms are available in Russia after Western vendors left, and why rankings name different leaders. What makes up the cost of ownership, and when BI is overkill and a dashboard or a bot report is enough. Finally, what to prepare before starting and why an off-the-shelf platform covers only the presentation layer.
What is a BI system in plain terms?
A BI system collects figures from all of a company's systems in one place, checks them against each other and presents them so that an executive sees the picture without asking accounting or exporting to Excel. BI stands for business intelligence, often rendered in Russian as "business analytics". What sets it apart from an ordinary report is that it combines several sources and keeps history, and users can break a figure down by branch, sales rep or document themselves.
In practice, BI answers questions a single accounting system can't. What is the margin by customer if sales are run in the CRM and cost of goods is calculated in 1C? Which sales rep brings in revenue, and which brings in overdue receivables? How does this quarter's revenue compare with the plan and with the same quarter last year?
| 1C report | Excel summary | BI system | |
|---|---|---|---|
| Sources | one database | any number, but manually | several systems, automatically |
| Freshness | as of when the report is run | as of the compilation date | per the extraction schedule |
| Who compiles it | the system | a finance person or analyst, from scratch every period | nobody, once set up |
| Metric calculation | that database's rules | in the formulas of a specific file | in a shared model, the same for every report |
| Drill-down | within the database | down to whatever rows were exported | from the total to a transaction in any source |
| Multi-year history | loads the working database | across different file versions | kept in a separate warehouse |
The BI market in Russia is large and growing. According to a preliminary TAdviser estimate published in February 2025, it exceeded 63 billion rubles in 2024. Most of that money goes on platforms and their implementation at large companies. For smaller businesses, understanding how the system works is more useful than memorizing vendor names.
What does a BI system consist of?
BI is built from four layers: sources, warehouse, data model and presentation layer. Sources are the systems where data originates. The warehouse is the database the data is regularly extracted into. The model is the set of rules by which data is linked and calculated. The presentation layer is the reports and dashboards the user sees. Platform reviews mostly compare the presentation layer, even though the quality of the figures is determined by the first three.
| Layer | What it does | Example | What breaks without it |
|---|---|---|---|
| Sources | store the original transactions | 1C, CRM, online banking, online cash registers, planning spreadsheet | without access to a source, the metric is entered by hand |
| Extraction and warehouse | pulls data on a schedule, keeps history, removes duplicates | nightly extraction of 1C documents and CRM deals into a separate database | reports slow down the live 1C, history is lost when documents are reposted |
| Data model | links tables through shared master data, stores metric formulas | "revenue = shipments excluding VAT minus returns", CRM counterparty = 1C counterparty by taxpayer ID (INN) | every report calculates its own way, figures contradict each other |
| Presentation layer | shows reports and dashboards, manages access rights | owner's screen, sales rep report, branch breakdown | the data is there, but it's awkward to view |
Extracting and preparing data is called ETL: extract, transform, load. It is at the transform step that you decide what to do with a counterparty entered in the CRM as "Romashka" and in 1C as "Romashka-Trade LLC". The platform can't settle this on its own: only the business knows the answer.
The layer beneath the presentation layer is a large market in its own right. According to a Strategy Partners estimate (February 2025), the data management segment — data platforms, DBMS and data warehouses — amounted to 32 billion rubles in 2023. Analysts forecast it will grow to 72 billion by 2030. The segment includes databases for all of a company's needs, not just analytics.
Why don't BI figures match 1C?
Figures diverge when BI and 1C calculate the same metric by different rules or look at the data at different moments. Typical causes: different transaction dates, VAT, documents reposted retroactively, returns, duplicate counterparties and extraction timing. Each is fixed with an explicit rule in the data model and regular reconciliation with the books, not by replacing the platform.
| Discrepancy | Where it comes from | How it's fixed |
|---|---|---|
| Monthly revenue doesn't match | BI counts by payments, accounting by shipments | the model fixes one definition; the other is shown as a separate metric |
| Difference equal to VAT | one report includes VAT, the other doesn't | a rule that "all amounts exclude VAT" and a label on every metric |
| Last month "changed" in 1C | an accountant reposted documents after the extraction | reloading open periods on every extraction, a change lock date in 1C after the month is closed |
| Returns and adjustments | in 1C they are separate documents, and not all of them made it into the extract | every document type that changes revenue is listed explicitly in the model |
| One customer, two records | the counterparty was created in the CRM and 1C independently | matching by INN and KPP (taxpayer ID and registration code), a designated system of record for master data |
| Transactions between your own legal entities | sales between group companies are counted as revenue | an "intragroup" flag on the counterparty, excluded from consolidated metrics |
| During the day, figures are "yesterday's" | extraction runs once a day at night | last load time shown on screen, more frequent extraction for metrics needed during the day |
Not one row of this table is solved by configuring the platform. All of them are solved by agreement: which definition is correct, who owns the master data, when the period is closed. Without a data owner, agreements don't hold. In a February 2024 forecast, Gartner expects 80% of data and analytics governance initiatives to fail by 2027. Gartner VP Saul Judah names the main reason as the absence of a real or manufactured crisis: without a pressing business problem, nobody follows the rules for working with data.
The working habit that keeps BI in line with the books is a control reconciliation after every month close. BI totals for revenue, cost of goods and receivables are compared with the 1C trial balance. Any discrepancy above the threshold is investigated before anyone starts using the figures.
Which BI platforms are available in Russia after Western vendors left?
The choice of Russian platforms is already substantial: Computerra's March 2026 ranking alone included twelve. The market has been restructuring since 2022, when Western vendors began to leave. Choose based on your own sources and data storage requirements: rankings use different methodologies and name different leaders.
Among Western vendors, Tableau announced on March 7, 2022 that it was stopping license sales in Russia and that previously purchased licenses would not be renewed. The shift is shown by data from the integrator Navicon, cited by CNews on January 13, 2025: by its estimate, over three years the share of foreign BI deployments fell from 90% to 23%, while the share of Russian ones rose from 9% to 68%.
The leaders depend on who is counting and how. In the CNewsMarket ranking "Business Intelligence (BI) Systems 2026", published on June 8, 2026, the top five were Yandex DataLens, PIX BI, Luxms BI, Sber's Navigator BI and the Foresight analytics platform. In Computerra's ranking of March 20, 2026, among twelve platforms, Visiology took first place, AW BI second and Luxms BI third. The rankings use different methodologies, and neither of them knows your sources.
| Platform type | Examples | When it fits | What to check |
|---|---|---|---|
| Cloud | Yandex DataLens | starting without your own server, a small team | where the data is stored, whether it contains personal data |
| On your own server | Visiology, Luxms BI, PIX BI, Foresight, Polymatica | requirements to keep data inside your perimeter, many users | cost of licenses and servers, whether the contractor has specialists |
| Accounting system reports | built-in 1C reports and report builders | one source, standard metrics | whether they'll be enough once a CRM or a second database is added |
| Custom presentation layer | a web dashboard or bot on top of your own data model | non-standard access rights, alerts, embedding into operational systems | who supports it after launch |
A separate question is the Russian software registry. For government customers it is mandatory: Government Resolution No. 1236 bars them from admitting foreign software to procurement, with exceptions the customer must justify. Since January 1, 2025, Presidential Decree No. 166 has prohibited government bodies and customers from using foreign software at significant critical information infrastructure (CII) facilities. CII entities — and they include many private banks, energy, healthcare and transport companies — must also check the requirements of Federal Law 187-FZ, Russia's critical information infrastructure security law: amendments made by Federal Law No. 58-FZ of April 7, 2025 added a transition of significant facilities to Russian software. If a company is neither a government customer nor a CII entity with significant facilities, this requirement does not apply. For such a company, a registry listing signals that the rights holder is in Russia and the product will be supported.
What makes up the cost of owning BI?
The cost of owning BI consists of six items: licenses, storage and servers, source connections, the data model, support and training. The license is the first thing you see on the invoice, but the other items drive the budget more. They recur every time accounting changes: a 1C update comes out, a new CRM arrives, the accounting team introduces a new document type.
| Item | What's included | What it depends on |
|---|---|---|
| Licenses | fee per seat or per server | number of users and the chosen platform |
| Storage and infrastructure | database for extracts, a server or cloud resources | data volume, depth of history, storage requirements |
| Source connections | extracts from 1C, CRM, bank, cash registers | number of systems, whether they have an API, quality of documentation |
| Data model | master data reconciliation, metric formulas, reconciliation rules | state of master data and number of exceptions in accounting |
| Support | fixing extracts after updates, new metrics | how often accounting and processes change |
| Training and roles | who builds reports, who only views them | team readiness and number of roles with different rights |
License price levels can be checked on vendor websites. For example, under Yandex Cloud's pricing rules as of October 2026, Yandex DataLens is free for individual use, with one seat. Once an instance has two or more seats, each one is billed: 990 rubles a month including VAT. For on-premises platforms, the server and its administration come on top of licenses.
The other rows can't be calculated from a price list. Connecting a standard 1C configuration and connecting 1C customized over ten years by three different programmers are two different jobs. That is why a BI implementation estimate is given after a source audit, and any price quoted without one is worth questioning: what exactly does it include?
When is BI overkill and a dashboard or bot report enough?
BI is overkill when there is one source, a handful of metrics and one or two people looking at them. For such needs, 1C's own reports, an owner's dashboard with five to nine figures or a morning summary in Telegram are enough. Full-scale BI justifies its complexity when there are several sources, many users with different access rights, and new questions about the data every week.
| Situation | What's enough | Why |
|---|---|---|
| All data in a single 1C database, reports are satisfactory | 1C reports | a second layer on top of the same source only adds a shell |
| The owner needs five to nine figures every morning | a scheduled summary in Telegram | it gets opened every day, the barrier to entry is zero |
| Several sources, one or two viewers, a stable set of metrics | a dashboard on your own data model | you need to combine sources, but not free-form analysis |
| Several sources, dozens of users with different rights, new questions every week | a BI platform on a warehouse | self-service breakdowns and access control justify the complexity |
The threshold is often crossed gradually. First the owner gets a summary in a bot, then a screen with drill-down appears, and a BI platform becomes necessary once department heads start building reports. The data model is reused at every step, so switching the presentation layer doesn't wipe out the work.
Which figures an owner should see every day and how often is covered in the article "The owner's dashboard". To start without a big project, our catalog of ready-made solutions has two that fit: auto-reports, which compile a summary of sales and metrics and send it to Telegram on a schedule, and consolidated dashboards, which bring sales, cash and sites together on one screen.
Where to start?
Start with a list of management decisions that should rest on figures, and a list of metrics with written definitions. The platform is chosen last. The company does most of the preparation itself, and this list immediately shows the scope of the work ahead.
- The decisions analytics is needed for. Which questions the executive asks every week, and which ones currently get the answer "I'll check and tell you tomorrow". Five to seven such questions define the contents of the first set of dashboards.
- Metrics with definitions. For each one, in writing: what it includes, which date it's counted by, with or without VAT, which documents reduce it. Finance and sales agree on the definitions before development. How management reports themselves are structured is covered in the article on management reporting.
- Source map and access. Where each metric lives: 1C, CRM, bank, cash registers, spreadsheets. For each system: a technical user, read access and a test copy of the database. If a metric lives only in someone's spreadsheet, it is moved into a system first; seven signs help tell when it's time to move off spreadsheets.
- Master data and its owners. Counterparties, products, departments, employees. Each master data set has one system of record; the others read from it. Ways to connect systems to each other are described in the piece on systems integration.
- A reconciliation rule. Which BI totals are compared with 1C after the month close, what discrepancy threshold is acceptable and who investigates when it's exceeded.
- An access matrix. Who sees margins, who sees salaries, who sees only their own branch. Access rights in BI are designed together with the model, not added after launch.
- A post-launch owner and support. An employee responsible for definitions and master data, and an agreement on technical support for the extracts. Accounting changes constantly, and without support the model drifts away from it within a few months.
The first three items need no development budget and on their own settle some of the arguments over figures. Items four through seven explain why the cost of BI can't be quoted over the phone: it is determined by the number of sources, the state of master data and the number of exceptions in your accounting.
Off-the-shelf BI or a data model built for your accounting?
An off-the-shelf platform covers the presentation layer: charts, a report builder, access management. The data model beneath it is built by each company for its own accounting, because everyone's procedures, master data, access rights and exceptions are different. The working setup is often hybrid: a ready-made presentation layer on top of a model built for your sources.
The difference shows up on the very first metric. One company counts revenue by shipments, another by payments, a third by certificates of completed work adjusted for returns from prior periods. One excludes intragroup sales, another has none, a third has half its customers entered twice after a CRM switch. The platform will display any of these variants equally nicely, including the wrong one. We covered the choice between an off-the-shelf product and a system built around your process separately: custom system or off-the-shelf.
If accounting is already consolidated in a single system, analytics is built on top of it with almost no separate reconciliation layer. In a fintech project, transactions, liability reconciliation and financial accounting run in one system, and end-to-end analytics shows management the actual state of the business in real time. When there are several systems, the data model and extracts are built as a separate project, and the presentation layer is either taken off the shelf or built for the task.
At INCUBE AI, we build this lower part as custom development: extracts from 1C and the CRM, master data reconciliation, a model with fixed definitions, reconciliation rules and a presentation layer on the chosen platform or our own. We work under contract, data stays in Russia, and we support the system after handover. If the figures in your reports contradict each other, tell us about your task: we'll go through your sources and metric definitions and use them to estimate the scope of work.
Sources and references
- TAdviser: Russian BI systems market — estimates, trends, largest vendors, summary of the review on the SL Soft website, February 27, 2025 — preliminary estimate: the Russian BI market exceeded 63 billion rubles in 2024
- Strategy Partners. "Segment analysis of the Russian corporate software market", February 2025 — "Data management" segment (data platforms, DBMS, data warehouses): 32 billion rubles in 2023, forecast of 72 billion rubles in 2030
- CNews, January 13, 2025: Navicon — Russian BI market grew 30% by the end of 2024 — according to the integrator, over three years the share of foreign BI deployments fell from 90% to 23%, while the share of Russian ones rose from 9% to 68%
- CNewsMarket: "Business Intelligence (BI) Systems 2026", June 8, 2026 — platform ranking, top five: Yandex DataLens, PIX BI, Luxms BI, Navigator BI, Foresight Analytics Platform
- Computerra: ranking of Russian business intelligence platforms 2026, March 20, 2026 — 12 platforms, top places for Visiology, AW BI, Luxms BI
- Gartner, February 28, 2024: forecast that 80% of data and analytics governance initiatives will fail by 2027 — forecast, Gartner VP Saul Judah
- GlowByte on Habr, March 11, 2022: Tableau suspends operations in Russia — decision of March 7, 2022 to stop selling licenses and not renew purchased ones
- Yandex Cloud: DataLens pricing rules — free for individual use (one seat); with two or more seats, 990 rubles a month including VAT for each, as of October 2026
- Resolution of the Government of the Russian Federation No. 1236 of November 16, 2015, Garant — ban on admitting foreign software to state and municipal procurement, with exceptions the customer must justify
- Decree of the President of the Russian Federation No. 166 of March 30, 2022, Garant — since January 1, 2025, government bodies and customers are prohibited from using foreign software at significant CII facilities
- Federal Law No. 58-FZ of April 7, 2025, Garant — amendments to 187-FZ "On the Security of Critical Information Infrastructure": transition of significant CII facilities to Russian software
Frequently asked questions
How is a BI system different from 1C reports?+
A 1C report shows data from one database and calculates it by that database's rules. BI combines several sources into one model: 1C, the CRM, the bank, cash registers and planning spreadsheets. So revenue by sales rep from the CRM can sit next to payments from 1C and the plan from the financial model. History is kept in BI's own storage, so comparisons with previous years don't load the working database. If all the figures you need already live in a single 1C database and its reports are enough, BI will mostly add a new shell around the same data.
Can BI connect directly to the 1C database, without a data warehouse?+
Technically, yes: many platforms read 1C data through the platform's standard interface or ready-made connectors. For one source and a modest data volume, this works. Trouble starts when there are several sources: master data has to be reconciled again in every report, and heavy queries slow down the database the accountants work in. A warehouse removes both problems. Data is extracted on a schedule and cleaned once, and every report is built on one model.
Which BI platform should a small or mid-sized business choose?+
To start, a cloud platform with a low entry barrier is usually enough. For example, under Yandex Cloud's pricing rules as of October 2026, Yandex DataLens is free for individual use, with one seat. If an instance has two or more seats, each one is billed: 990 rubles a month including VAT. If data storage requirements rule out the cloud, companies look at platforms installed on their own servers. The platform choice is secondary, though: any of them works on a reconciled data model, and any of them shows wrong figures on an unreconciled one. A reliable test is one report built on your real data extract, not on the vendor's demo data.
How long does a BI implementation take?+
The timeline depends on the number of sources and the state of master data, so an honest estimate comes after a data audit. Installing the platform and building the first charts can be done quickly. Most of the time goes into extracting data from 1C and the CRM, reconciling counterparties and product catalogs, agreeing on how metrics are calculated, and checking against the books over several closed periods. Until the presentation layer matches accounting for at least a couple of months, managers keep looking at the old reports.
Do you need an analyst on staff to use BI?+
To view ready-made dashboards, you don't need a dedicated analyst. You do need a data owner: someone responsible for master data, for metric definitions and for making sure the figures match the books. Often the CFO or chief accountant takes this role. The technical side — data extracts, the warehouse, fixing breakages after 1C updates — is easier to keep under a support contract. Without an owner, the model drifts away from the books within a few months, and trust in the figures disappears.
Does a BI system have to come from the Russian software registry?+
For government customers, yes: Government Resolution No. 1236 bars them from admitting foreign software to procurement, with exceptions the customer must justify. Since January 1, 2025, Presidential Decree No. 166 has prohibited government bodies and customers from using foreign software at significant critical information infrastructure (CII) facilities. If a company is neither a government customer nor a CII entity with significant facilities, this requirement does not apply. CII entities — banks, energy, healthcare, transport — should separately check the requirements of Federal Law 187-FZ: 2025 amendments added a transition of significant facilities to Russian software. It is also worth checking where the data is physically stored: extracts from 1C and the CRM almost always contain personal data of customers and employees, and Federal Law 152-FZ, Russia's personal data protection law, applies to them.