Growth you can actually measure

What people usually arrive with

If you recognise any of these, that is roughly the conversation worth having.

Acquisition costs keep rising and nobody can say why.

A rising cost is often a symptom of something other than auction prices: an exhausted audience, a channel that stopped delivering, or measurement that has been counting the wrong thing for six months. Until those are separated, you optimise campaigns while the problem sits elsewhere.

Every channel reports success, yet the customer count doesn't move.

Every platform claims the same sale, so the reports add up to more than reality. Until conversions are counted in one place under one definition, budget flows to whichever channel shouts loudest, not to the one that actually brings customers.

No one knows which channel brings people who stay longer than a month.

In a subscription business the result is decided by the third and tenth order, not the first. A cheap channel can bring customers who never return, while a more expensive one brings people who stay a year. That only becomes visible once ad data is joined with order history.

Meta reports three times more conversions than your sales show

Three dashboards, three versions of the truth and no basis for choosing between them. Usually no system is lying – each one is simply counting something different.

Half your Facebook and Instagram traffic lands as “direct”

In-app browsers anonymise the click. Budget spent, sales happening, attribution missing – so you optimise blind.

Cookie consent is costing you data you did not have to lose

Consent Mode bolted on with a plugin, tags firing before consent is resolved, refusal buried under “settings”. Legal exposure and a measurement hole at once – and the conversion modelling that would fill part of that hole never starts, because it has nothing to work from.

Prices on the site disagree with the cart and with the product feed

Three price sources maintained by hand. The customer sees one, pays another, Merchant Center sees a third – and one day suspends the account.

The dashboard shows numbers, but nobody knows what to do with them

A report answering questions nobody asked. Measurement should end in a decision, not in a chart.

What I do

Growth diagnosis – Where you actually lose customers and money, what can be recovered, and in what order

I start by counting where the business actually loses: acquisition, the first repeat purchase, or retention. The output is a ranked list of causes by how much can be recovered and at what cost – with numbers you can verify, and with the places where the data has no answer marked as such.

Measurement layer – Tags, consent, warehouse

Web GTM, server-side GTM, Consent Mode v2 and export into BigQuery. Including the awkward environments: multiple subdomains, single-page applications, Advanced Matching and event match quality. Measurement that survives cookie blocking and in-app browsers.

Attribution and modelling – Which number to trust

Multi-touch models joining GA4, Google Ads, Meta, CRM and subscription data. Deterministic identity resolution and offline conversions that close the loop from click to CRM transaction. Model changes run in parallel with the existing one and validated on data before they replace it.

Sites with measurement in the foundation – Not a plugin added later

A site built so that it measures from day one – a shared event contract, schema.org generated from code, static HTML legible to search engines and language models alike.

Campaign management – Judged on data, not on the ad panel

Google Ads and Meta run day to day, evaluated against warehouse numbers rather than platform claims. Account architecture too: consolidating campaigns where fragmentation degrades the automated bidding signal, and matching the bidding strategy to real conversion volume.

Full technical scope →

Measurement, live

This page reports through the same event contract I deploy for clients. Below you can see what actually leaves the page – click anything here, or answer the consent banner.

Event names are set once. Data collected under a different name stays under that name forever – which is why the vocabulary is agreed up front, before you accumulate six months of measurements that cannot be compared with each other.

Where this comes from

More than twenty years across research, marketing and data – since 2004. I spent over a decade in market research and consulting, including leadership and board-level roles at ARC Rynek i Opinia, one of Poland's largest research agencies – running research teams and working with enterprise clients in banking, telecom, healthcare and FMCG. I was measuring human behaviour before anyone called it digital analytics.

Today I lead growth, acquisition and marketing analytics at Tendios Technologies in Barcelona – an AI-powered public procurement platform operating across Spain and the EU. Before that I spent three and a half years consulting for SaaS and subscription businesses, on projects large and small, in Poland and abroad.

I know subscription businesses from the inside – I co-own Fitness Catering, a subscription meal-delivery brand. I spent years working for several others in the same category and co-owned the price comparison engine that serves it. I know the economics of recurring revenue from the ledger side: what it actually costs to acquire a customer who stays, and why a discount war over the first order pays off for almost nobody.

The studio is new, the experience is not. Full track record on LinkedIn.

References

Excerpts from LinkedIn recommendations – from people I worked with on both sides: in e-commerce and on the board of a research agency.

More recommendations – there are a dozen or so – on the LinkedIn profile.

How I work

1. First I check what you actually measure. It usually turns out to be less, and different, than the dashboard suggests.
2. Then I fix the measurement, not the report. A handsome dashboard on bad data is worse than no dashboard, because it looks credible.
3. The data stays with you. Your warehouse, your account, your access. I get permissions, not ownership.
4. I leave documentation, not dependency. Event names, metric definitions and the reasons behind decisions written down, so the next person does not start from zero.

Contact

Tell me what you are trying to measure and what you cannot see. I will tell you whether it is an hour's problem or a project – even when the answer is that you can handle it yourselves.

Let's talk.