This is data engineering. Not an AI startup.
ACS — Abovyans Consulting Services. We build one unified data model around your business and hand it back in the familiar interface of Excel and Power BI. Inventory, sales, purchasing, e-commerce and finance stop arguing about the numbers — because there's now one number.
We don't roll out someone else's ERP over eighteen months. We build a tool around your business, on your terms — and it stays inside your own infrastructure.
The math gives you everything you need. Whether you add AI on top is your call.
90% of the value is careful data architecture, DAX measures and honest math. Calm, without the hype and theatre around "neural networks".
We prove it with work
Concrete numbers, real cases, formulas. No adjectives without figures. If something got faster, we show by how much and why.
We don't compete with SAP
A big ERP is years and millions. We solve 80% of the same job for 5% of the budget — no rollout pain, in Excel and Power BI.
The data is yours
The model and all the data live on your side. We build a tool that stays with you — not a key to our cloud.
Three principles every project rests on.
They run through all of our work — and they're the best answer to "but we already have reports".
MVP-first delivery — into use, fast
The product goes into use as early as possible, so the business starts getting value now — not in a year.
No year-long rollouts. The first model goes live in 4–8 weeks, and by the end of the second month roughly 80% of the functionality is delivered. The remaining 20% is built alongside live use — you make decisions on data now, instead of waiting six months for a perfect solution.
first MVP — from 4 weeksBuilt for your business — not a template off the shelf
There are no ready-made solutions on a shelf. Every model is built from scratch around your inputs.
This isn't a stock dashboard. We build around your terms and your metrics — the ones that settled in your business. If you measure "yield per ton" rather than "revenue per square meter", we build around yield per ton. The model speaks your language instead of making you learn someone else's.
built from scratch, for youPareto's 80/20 — to the point, not reports for their own sake
80% of the business result comes from 20% of the effort. Our job is to find and surface exactly that 20%.
We don't churn out reports for the sake of reports. A simple chart that shows real sales efficiency beats ten over-engineered dashboards built "because it looks nice". Less noise — more decisions.
20% of effort — 80% of the resultFrom the first review to a model in production.
Five steps. Transparent, with no surprises mid-project.
Free data review
We look at your exports, reports and e-commerce accounts. We find where the 20% that drives 80% sits — and show the map on the call.
Design the data model
We describe the sources, terms and key measures in your language. We agree on what's in the MVP and what comes in the second wave.
Build the ETL and measures
We set up extraction from your accounting, e-commerce, banks and POS. We normalize the data, write DAX measures, assemble the first working screens.
Into production
We hand over the model with documentation: which measure is where, what the sources are, how to add a channel. 80% of the functionality is already in use.
Develop and support
The remaining 20% is built alongside live use. After that — ongoing support on a clear basis: new directions, channels, measures.
What it actually runs on.
No secrets. "Excel" on the outside — a serious data engine on the inside.
On the outside the client sees a familiar Excel file or a Power BI dashboard. Inside — a columnar engine, a normalized model and hundreds of DAX measures. At larger volumes ClickHouse becomes the core, while Excel and Power BI stay the front everyone understands.
Model & measures
The heart of the system. The columnar engine holds tens of millions of rows; all logic lives in measures, not in formulas scattered across cells.
Storage & ETL
Extraction from accounting, e-commerce, banks and POS, normalization and joins. At larger volumes — a columnar database as the core.
Front for people
What an employee opens every morning. A familiar interface — no new product to learn, no dependence on someone else's closed cloud.
Automation
Where the math pays off directly — automated ad bids from DAX measures over time frames, thousands of bids sent through an API in one run.
One engineer — the one you actually work with.
No call center, no conveyor of account managers. The person who designs your model is the same one who answers your questions.
Mikhail Abovyan
founder · data engineer
Since 2018 I've been building unified data models for mid-sized businesses — from e-commerce sellers to manufacturing and wholesale. Over that time, 10+ companies in continuous production: inventory, sales, purchasing and finance finally pulled into one picture.
I don't believe in the magic of "neural networks that do everything for you". I believe in careful math and honest data engineering — the kind that works every day and doesn't break at the first non-standard question the business asks.
"A business doesn't need beautiful dashboards. It needs one number it can trust — and the decision that follows from it."
I work directly: you talk not to a manager, but to the person who actually designs and builds your model. It's faster, more honest and clearer — for both sides.
A review of your business — free. An hour of your time, our prep.
Before the call we look at your data — exports, reports, e-commerce accounts. On the call we show a map: where your 20% that drives 80% sits. No obligation to go further.