The system pulls data from the Wildberries seller account, calculates sales and advertising metrics, spots anomalies, and sends a Telegram summary that explains what caused them
MPstats is an analytics platform for marketplace sellers. For its audience of Wildberries sellers, we built a standalone product: a Telegram agent that monitors the store so the seller doesn’t have to.
Sellers need a daily read on how their business is doing: how many units sold, how ads are performing, what’s happening with returns and stock, where sales dipped and why. All of that data lives in the Wildberries seller account, scattered across several reports.
To see the full picture, sellers exported sales, advertising, returns, and stock reports and pieced them together in Excel — hours of work every day. And with hundreds of SKUs, they still couldn’t keep track of everything: figuring out which items were slipping and why meant digging through the data by hand.
Drops in sales or ad performance were spotted late, and tracking down the cause took even longer.
The data had to be collected automatically, and instead of a pile of numbers, the seller needed to see what matters: what changed, where the problem is, and what to do about it.
The system automatically pulls Wildberries data: sales, orders, returns, stock, and finances. Sellers no longer export reports, and they set the collection schedule themselves.
Every day, a summary arrives in Telegram with a dashboard and a list of the biggest problems, with the ones that hit revenue hardest at the top. Metrics are calculated by business, category, brand, and product and compared against the plan and past periods. Sellers set alert thresholds and their sales plan in their own Google Sheet.
For each problem, the system builds a hypothesis from that seller’s own data and suggests next steps. Sellers can ask follow-up questions in the chat in plain language and drill down to a specific product, and every answer is based on real data only.
Here’s what happens to the data, step by step — from collection to answering a question in Telegram.
Core data — sales, orders, returns, stock, and finances — comes through the official Wildberries API. Whatever the API doesn’t provide (the product card funnel, search queries, ad details), the system pulls from the seller account through the seller’s authorized session and the reports they upload. Data is collected on a schedule, for example in the morning, afternoon, and evening.
From that data, the system calculates more than 60 metrics: the sales funnel (impressions, clicks, add-to-carts, orders), conversion rates, ad metrics such as advertising cost of sales (ACoS) and return on marketing investment (ROMI), and margins after commissions and logistics. Each metric is compared against four baselines — the seller’s plan, yesterday, last week, and the previous summary — so you see not just the number but where it’s heading. The system also reconciles its numbers with the Wildberries sales report and warns you if they diverge by more than 5%.
To avoid flooding the seller with alerts, the system compiles the day’s top problems and ranks them by revenue impact, so the costliest issues come first. Sellers set the thresholds for what makes it into the summary, and a filter weeds out false alarms: brief spikes and repeat alerts about the same problem.
For each anomaly, the system builds a hypothesis strictly from that seller’s data — no generic advice — and suggests next steps. Say the seller asks in the chat, “Why did footwear sales drop yesterday?” The system breaks the drop down by brand and product, all the way to a specific item. It answers from real data only and doesn’t make up numbers.
Also used: Celery + Redis — scheduling and background jobs · matplotlib and plotly — charts for the summaries
Sellers no longer export reports and stitch them together in Excel: the data, explanations, and recommendations arrive ready to use.
The same approach works for anyone who handles large volumes of data and analytics.
We build solutions like this around your needs: turnkey AI products, AI agents and automation, integrations, and data.
Our servicesTell us about your project — we’ll talk it through and show you how to make this work for your business
We use cookies and Yandex Metrica to improve our website. Privacy policy
