I treated my portfolio as a product: PRD, North Star and privacy

A Product Owner portfolio has an odd problem: it has to prove I can lead products, yet it's usually built like a nice-looking résumé, with none of the practices I advocate.

So I did the opposite: I treated this site as a product, with a defined problem, audience, metrics, roadmap and decision log. This post is how it went.

Starting with the problem, not the screen

A résumé and LinkedIn show titles, companies and tools. But whoever evaluates a PO wants to know something else: do they understand the problem before proposing a solution? How do they decide when there's a trade-off? Do they measure results?

Since my formal product experience started in 2024, after years in procurement, that problem is even bigger for me. The quality of my reasoning has to be visible. That's why case studies sit at the center of the site, not the list of jobs.

Who it's for

I defined four audiences: recruiters, product leaders, potential clients and the product community. The primary persona is the product leader who might hire me, because they weigh the most in the decision and the case studies were designed to convince them.

That drove concrete choices. Case studies follow a fixed template (context, role, constraints, discovery, decisions, delivery, results and learnings) and have a dedicated block for decisions, with alternatives, criteria, choice and trade-off. It's the part a product leader most wants to see, and the part portfolios usually hide.

The North Star

Pageviews measure reach, not value. The site exists to start work conversations, so the main metric is qualified contacts per month: people who read at least one case study and then clicked through to my LinkedIn or GitHub.

I haven't instrumented those conversion events yet. It's on the roadmap, and owning that is part of treating the site as a product: the metric was defined before the measurement existed.

Decisions and trade-offs

A few decisions I logged:

Decision

Where should content live?

Options considered

• A headless CMS, like the Directus I use at work.
• Notion as the content source.
• MDX files in the repository itself.

Criteria

A single author, occasional publishing, zero cost and full control over the format, including components like this decision block.

Choice

MDX in the repository, read at build time. All content starts as a draft and only goes live when I switch its status to published.

Trade-off

Publishing requires Git and a deploy. In return I get version history, review before publishing and no external dependency.

Decision

How to measure without disrespecting privacy?

Options considered

• Keep Google Analytics running for everyone, with no notice.
• Load Google Analytics only after the visitor accepts cookies.

Criteria

Compliance with Brazil's data protection law (LGPD) and consistency with the site's transparency principle.

Choice

Google Analytics only loads after "Accept". "Accept" and "Reject" carry the same weight in the notice, and the choice can be changed at any time. The privacy policy explains what is collected and why.

Trade-off

I lose data from anyone who rejects or ignores the notice. My numbers get smaller, but reliable where it matters: trends and proportions.

I also chose to have the site in Portuguese and English, with Portuguese as the default, and to flag confidential clients instead of inventing numbers I can't disclose.

Built with AI agents

Much of the code and documentation was produced with AI agents that I directed and reviewed. That enabled something I couldn't do alone in the same time: in parallel, one agent handled automated tests and CI while I worked on the cookie notice.

The PO role didn't shrink, it moved. Instead of writing every line, I defined the problem, chose between alternatives and reviewed the output. That's why every page carries a note saying the site was built with AI assistance and that all content is reviewed by me.

What's next

This site's public roadmap includes, among other items:

• instrumenting conversion events to actually measure the North Star;
• a third case study, on system modernization and developer tools;
• a direct contact option at the end of each case study, the moment of highest conviction.