Real-time financial dashboard for an e-commerce store

The problem

An online store selling boilers and heating equipment. Orders come from the website into Bitrix24, payments arrive on two bank accounts, and the shop has a cash desk — pickup-on-store payments, refunds, small cash purchases before depositing into the bank. The «how much did we actually earn this month» picture used to be assembled by the accountant once a quarter and always with discrepancies: the bank said one thing, CRM another, the cash book a third. Decisions on purchasing, promotions, bonuses were made by feel, not by data.

Before

• The financial result was assembled by hand once a quarter — 2–3 weeks of accountant work plus discrepancies nobody had time to untangle.
• «Payment vs order» reconciliation was sporadic: money left the bank — but matching it to a specific order took time.
• Bank statement categorisation (800–1000 lines a month) — by hand.
• Oddities — a big purchase from a new supplier, an old order edited after closing, a non-typical client payment — surfaced 2–3 quarters later or never.
• Marketing, cash, suppliers, clients — each category in its own spreadsheet.

What we built

• A single web dashboard — all the company's money in one screen: bank (both accounts), Bitrix24 orders, cash-desk operations.
• Cash-desk transactions before bank deposit — via an internal Telegram bot: the cashier types «yesterday gave the installer 250 for a callout» → the bot parses it and files it into the right category.
• Automatic categorisation of bank statements: the system files 95% of lines by itself (suppliers, taxes, advertising, payroll, rent). Edge cases go to manual review and no longer get lost.
• «Order vs payment» reconciliation: if an order is closed but no money came in — or money arrived without an order — it lands on a «needs review» list.
• Telegram alerts to the leadership chat on triggers: «not paid in 14 days», «first big purchase from this supplier», «closed order edited retroactively», «margin below plan». Nothing is missed.

Results

• The director sees the real financial result of the day at any moment — no waiting for the quarter.
• «Bank vs CRM» discrepancies surface within a day — they used to surface 2–3 weeks later or never.
• The accountant spends ~10 minutes a day on categorisation instead of 2 days at the end of the month. The system proposes a category, they confirm.
• Anomalies became a starting point for conversations with managers: «this order was edited retroactively — what happened?»
• New hires in accounting and analytics get to an operational picture in a day, not a month.