Senior-led · Commerce & Real Estate Intelligence

Structure turns
data into
decisions.

Not “ROAS is up 12% this month.” More like: “62% fit — bestseller runs out in 3 weeks, marketing is eating margin in two markets.”

One engine, calibrated to your business. I build the decision layer commerce and real estate businesses are missing — pipelines, reporting and AI agents that turn scattered data into one screen you can act on. Diagnosis to delivery, one senior operator.

DIRECTORY
ATA
Directory
Structure that gives your data meaning.
Data
Signal from every campaign, every market.
The third D
Decision.
Data without Directory is noise. With both, every number points somewhere.
Margin per SKUOrder-level attributionBigQuery pipelinesAnomaly detectionDecision rulesForecastingCompetitor pricingMargin per SKUOrder-level attributionBigQuery pipelinesAnomaly detectionDecision rulesForecastingCompetitor pricing
Philosophy

Most brands have data.
Few have structure.

The Directory principle

A campaign without structure is a folder with no names. I build the architecture first — campaign hierarchy, keyword segmentation, budget logic — so your data flows where it should.

Data that answers questions

I connect Amazon Ads to your analytics infrastructure — BigQuery, Looker, custom dashboards. Every impression, click, and conversion lands somewhere meaningful. Then I act on it.

Decisions, not reports

Every week there's a concrete next move — a bid to cut, a keyword to harvest, a listing to rewrite. That's what decisions look like.

What I do

Three disciplines.
One outcome.

I work across the data layer, the reporting on top of it, and the spend it governs — because separating them is where most agencies lose the thread.

01
Data
Infrastructure
Marketplaces, order system, ad platforms and bank piped into a governed BigQuery model. Raw → clean → master, idempotent loads, no duplicated numbers. One place where the figures finally agree.
BigQueryPipelinesData model
02
Reporting
& Decisions
Margin per SKU, market and channel, refreshed on a schedule rather than assembled by hand. Forecasting, anomaly alerts and decision rules set to your thresholds — so “margin dropped” arrives as “do X by Friday.”
Looker StudioForecastingAI agents
03
Campaigns
Ad spend judged against real margin, not revenue. Structure, bidding and budget allocation measured per SKU and market — including the campaigns that look profitable in the panel and lose money in the accounts.
Amazon AdsROAS vs marginAttribution
Built for your business

Calibrated for two industries —
more coming.

E-commerce

Cross-border commerce

Margin per SKU, per market, per channel — the number your ad platform can’t show you. I ran a beauty-tech e-commerce business selling across nine markets in four channels and nine currencies, and I owned the P&L, so I built exactly what I needed: BigQuery pipelines, order-level attribution, AI agents watching margin and inventory.

The first thing the data showed was that the margin everyone was steering by had been calculated wrong. Correcting it was half the gain. The other half was repricing and moving focus toward the markets that actually earned — DE and PL beat eMAG on delivery cost and product pricing — with competitor prices tracked in real time.

The business was Lumineo; the engagement ran from October 2025 to July 2026.

61%
Margin, DE market · from a mis-measured 8%
24h
Data to decision · was 7 days
42
SKUs
4.8×
Avg. ROAS · portfolio average, all markets
−58%
ACOS · portfolio average
29%
Ad spend saved · varies by platform
Amazon Sponsored ProductsSP-APIListing & A+ ContentHelium 10BaselinkerCompetitor intelligenceDE · FR · ES · IT · GB · PL · CZ · HU · RO
Real Estate · in development

Pricing intelligence for developers

Whether today’s price list is selling fast enough — and where to move it before the sell-through indicator says you’re too late. Built on the same engine: inventory by unit type, absorption per m², competitor pricing, financing-schedule alignment.

Currently in development. No numbers to show yet — get in touch if this is the gap you’re looking at.

Talk about it →
The Engine

One engine.
Built for you.

Every cockpit I build runs on the same engine — the part that doesn’t change between industries. Sources flow through a governed data model: raw → clean → master, idempotent loads, no duplicated numbers. On top sit forecasting and anomaly detection, then decision rules calibrated to your business — the thresholds and priorities that turn “margin dropped” into “do X by Friday.”

What sits on top is never generic. The cockpit shows one number — your operation’s fitness, not twelve charts to interpret — plus navigation: not just where you are, but what’s about to happen and what to do about it.

The engine is reusable. The calibration isn’t. That’s the work.

Who you actually work with

One operator.
Twenty-five years.

I'm Tomasz Kurzątkowski. For 25 years I've been turning data into decisions and ads into sales — across e-commerce, pharma, telco, and the public sector.

Directory has been my own practice since 1998. Today it points at one thing: the decision layer commerce and real estate businesses are missing — pipelines, attribution, and automated monitoring that keep the numbers honest. In e-commerce that has meant Amazon Ads governed against real margin across nine markets and four sales channels.

When you hire Directory, you get me. Not a junior, not an account layer. The person who builds the system is the person who runs it.

NEUCA · Pharma
Built Apteline into Poland's #3 online pharmacy

20% of revenue, and the GDPR-compliant Master Patient File behind it.

ILC / NEUCA Group · Managing Director
Rebuilt the data model and BI for an entire pharmacy network

Data ownership, quality KPIs, and the reporting layer the network runs on.

QED Software · AI/ML
Ran go-to-market for AI/ML into the US

Enterprise advisory on responsible AI and data governance.

Orange Polska · Partnerships
Microsoft and Huawei partnerships across cloud and analytics

Cross-functional transformation programmes at telco scale.

How it works

Four moves.
Repeatable.

A system refined across dozens of product launches and market expansions. The same logic applies whether it's month one or year three.

Step 01

Diagnose

Where your numbers disagree and why — platforms against the order system against the bank. I find what is mis-measured before anything gets rebuilt.

Step 02

Build the data layer

Pipelines, model and reporting built on your infrastructure, in your accounts. Reconciliation rules written down, including which system wins in a conflict.

Step 03

Calibrate

Thresholds, alerts and decision rules tuned to how your business actually runs. The engine is reusable; this part never is.

Step 04

Report & Decide

Live dashboards, monthly strategy calls, clear attribution. Every session ends with a concrete next move — not a slide deck.

Tooling
Google BigQuery· Looker Studio· Python / dbt· Cloud Run· AI agents
Get in touch

Let's build something
that performs.

First conversation is 30 minutes and costs nothing. Tell me where your numbers aren’t driving decisions yet.