Advertising without surveillance

Part 1 explained why this website needs no cookie banner: we track no one. The obvious follow-up question came promptly — then how do you know whether your marketing works?

The answer has been running here in day-to-day operation for five weeks. This part shows it in detail, with the tools and with what the numbers actually deliver.

The dilemma, named honestly

Marketing without measurement is a game of luck: you produce content, post on channels, and whether any of it works, nobody knows. The standard solution — Google Analytics — solves the problem and creates three new ones: your visitors’ data flows to an advertising corporation, the website needs the consent dialog, and you yourself become part of exactly the surveillance machine this company argues against. For a brand called “Own Your Stack”, that would be satire.

The third way: count anonymously, host it yourself. For us that is Umami on our own infrastructure — there are comparable tools; the category matters more than the pick. No cookie, no fingerprinting, no profile. What gets counted is the event, not the person.

What you see — and what actually suffices

Daily practice consists of four questions, and none of them requires knowing who someone is:

What gets read? Page views per post. Shows which topics carry — and which of the author’s favorite topics interest nobody.
Where do readers come from? Referrers: search engine, LinkedIn, newsletter, direct visits. Shows which spoke of the visibility machine actually turns.
Is the campaign working? Every teaser link carries a campaign tag (UTM). That traces every blog post back to the LinkedIn post that brought the readers — as an event, not as a person.
Where do they go next? Outbound clicks: newsletter signup, contact page, source links. That is the conversion that counts.

And because the tool has an API, the morning routine is the agent’s job: it pulls the numbers, compares them with the previous week, and reports anomalies. “Post X is picking up through search, campaign Y delivered nothing yesterday.” Two sentences instead of a fifteen-minute dashboard session. That, too, only works because the tool is API-first — more on that on Thursday.

What you deliberately give up

Transparency is part of the method, so here is what this way cannot do, stated plainly:

Retargeting — “chase every visitor of the pricing page with ads for three weeks” — is not possible and never will be. Person-level funnels — “person A read X first, then Y, then bought” — do not exist; we see frequencies, not life stories. Demographics — age, income, interests — are missing entirely.

The honest assessment after five weeks: for ad-driven mass sales, that would be a loss. For a company that builds trust with expert content and whose customers pick up the phone themselves after reading, none of it is needed. With anonymous frequencies we make the same decisions others make with personal profiles — which topics, which channels, what frequency. The rest of the data hoarding is, for this type of business, ballast with legal risk attached.

The business view for Switzerland

For Swiss SMEs, the third way has an extra benefit that is often overlooked: you can show it to your own customers. The measurement that touches no personal data is also the one that needs no explaining in the customer meeting, in the data protection audit, and under the nDSG, the Swiss data protection act. Data protection here is not a compliance burden that hinders marketing — it is the selling point that marketing is.

And the effort? One afternoon of installation on an existing server, one counting script in the page header, one line in the privacy policy. The campaign discipline — every link gets a tag — is a matter of habit and becomes routine after a week.

The takeaway test: open the analytics of your own website and mentally strike out everything that identifies a person. Which of your actual decisions of the past three months would have changed? If the answer is “none”, you know what to do.

Part 2 of the Privacy series. Thursday brings the checklist behind all of this: choosing software when an agent is supposed to operate it.