A marketing data warehouse with agents, for a flat fee

01

The moment

€1.23 a year. That is what my setup of a marketing data warehouse in BigQuery costs to run per year, because you only pay for what you store and query. That number shaped my offer: when the technology costs almost nothing, the work of setting it up right and running it is what a client pays for.

My answer is a marketing data warehouse in BigQuery with agents on top. All of a business’s marketing data sits in one place, and on that place run agents that write reports, check campaigns and answer questions someone would otherwise spend hours on in spreadsheets. I build it, I run it, and I sell it as a flat fee, not as agency hours.

I describe it in such detail because it is an offer someone with your background can build. You need no team and no agency for it. You need an understanding of marketing and the willingness to sell results instead of hours.

02

What went wrong

What can go wrong here I mostly see in what I do not know yet. I do not know exactly what a typical setup really costs, which is why there is no number for the flat fee here yet. I do not know how far it can be standardised. I believe a large part is the same for everyone and a small part different for each, but how big the small part is I am learning with every conversation.

And I see a mistake I watch others make and want to avoid myself: selling the warehouse as a project instead of a service. A project ends with a handover, and then it goes stale. A service keeps running, gets looked after and becomes more valuable over time. Whoever bills something like this as a one-off project by the hour is selling exactly the model that is losing right now.

The third possible mistake is saying yes where no would be right. Not every business needs a warehouse. If a client has a single ad account and their enquiries come by phone, the better answer may be a clean form and the habit of looking once a week.

03

What I know now

What I know now: the expensive part was never the technology. It was always the work of setting it up right, and exactly that work is the offer. How many client accounts the €1.23 figure rests on, I will add once I can confirm it across several setups.

I also know in which order it works. First I collect with the client the questions they really ask every week and the numbers their decisions rest on. Then I connect the sources those numbers come from: ad accounts, website, forms, bookings or sales. Only when the numbers fit together and we agree what each one means do the agents come in. That order is not pedantry. It is the reason the Monday report ends up being right and not just sounding good.

I know the flat fee is fairer for both sides. The client knows in advance what it costs. I have a reason to build it so well that it runs without me. With hours it would be exactly the other way round. When an agent does in minutes what used to take a person hours, billing hours no longer makes sense.

I know that most small businesses give their data away today. Not to someone, but to disorder. The numbers are there, in every tool, every day. They are just never brought together. Agents make affordable for these businesses, for the first time, what only corporations used to have: someone who looks at the numbers every Monday and says what changed, as a short text in the inbox instead of a dashboard nobody opens.

And I know that a warehouse belongs to the client. If we stop working together one day, the data stays where it is, clean and readable. That is an honest selling point you can use well as an independent, because it creates trust. And the value grows: after a year the system knows the seasons and the campaigns that worked.

What I try, I document in the open, including when it fails. You should see what holds and what does not before you put your own time or money into it.

04

Open question

The price is open. The flat fee will be fixed once I know what a typical setup really costs. Until then I would rather talk about a concrete situation in the conversation than about a number on a page. When the number stands, it goes into the changelog here.

05

Changelog

29 October 2026
First version. Nothing here is rewritten silently: every later change gets its own entry in this changelog.