PM portfolio from scratch: what to build when you have nothing to show

Only 17% of PMs have a portfolio (Aakash Gupta’s widely cited survey). That means having one at all is a differentiator before a hiring manager reads a single word. The bar is not perfection: it is existence plus clarity. If you are starting with nothing shipped, the goal is 2-3 case studies that prove you can identify a problem worth solving, argue that the market is real, and make a hard trade-off. Everything else is filler.

What a portfolio has to answer that a resume cannot

A resume lists what you did. A portfolio answers three questions a resume cannot:

• Why was this problem worth solving at all (not just “users complained”)?
• What did you choose NOT to build, and why?
• What would you do differently now?

If your case studies do not answer those three questions, you have a process log, not a portfolio. Process logs get skipped.

Projects that work when you have no shipped PM work

Not all fabricated projects are equal. These four types are credible because a hiring manager can verify the inputs.

Product teardown with a counter-proposal. Pick a real product with known metrics (a failed AI feature, a stagnant consumer app). Diagnose why it is underperforming using the viable/lovable lens: was the market too small, was willingness to pay unclear, did the product force a behavior change users never asked for? Then write a one-page spec for what you would do instead and why. This proves strategic thinking on verifiable material.

A spec for something you pitched at a previous job. If you did BA, ops, or analyst work, you almost certainly pushed for a product change that never got built. Write the PRD now. The problem context is real; the artifact is new. For confidential work, describe the problem, your process, and the outcome in prose; reconstruct any wireframes at low-fidelity; and get written permission from your manager for one or two artifacts if you can.

A volunteer or open-source project. Nonprofits and open-source tools have real users and real constraints. The outcome may be small; the reasoning process is identical to a paid PM role.

An AI-native spec. In 2026, AI-PM roles are where open reqs are concentrated and where direct experience is rarest. A PRD for an agentic feature, an eval framework design for an LLM product, or a “should we build this AI feature?” analysis (covering viable market size, latency and cost constraints, failure modes) is directly relevant to the companies hiring right now. Do not frame this as “demonstrating AI fluency.” Frame it as doing the actual work: see build an eval portfolio project for a structured approach.

How hiring managers actually read a portfolio

The decision sequence is: URL seen on resume, 10-second load, headline and positioning statement, first case study. If the first case study does not hook them in 4 minutes, the portfolio is skipped. Under 5 minutes total is typical before a hiring manager decides to forward or pass.

What that means concretely:

• Your positioning statement at the top is not “Aspiring PM passionate about user experience.” It is one sentence about the problem space you own: something like “ex-BA who works at the intersection of enterprise data workflows and the people who have to use them.”
• The first case study leads with the market argument and the trade-off, not the process. Save the user research synthesis and the wireframes for the body.
• 2-3 strong case studies beat 6 mediocre ones. Each case study should include at least one actual artifact shown (not just described): a PRD excerpt, a prioritization table, a user research synthesis.

Platform: an honest comparison

Notion is the fastest to set up and free. The default URL (notion.site/...) signals disposability to senior hiring managers. Notion public pages also index poorly in search and can feel like a draft. Use it only if you add a custom domain.

Personal site via Webflow, Framer, or Vercel takes 10-15 hours with AI coding tools but gives you a custom domain, full control, and a signal that you can drive a project end-to-end. This is the right choice for anyone applying to companies that move fast in product.

UXfolio and similar portfolio tools provide good scaffolding but the domain itself (uxfol.io) flags the site as a portfolio template to anyone who recognizes it.

PDF does not work. No links, no updates, and it cannot be shared cleanly in an ATS or Slack thread.

The recommendation: custom domain on a personal site, built with AI tools if that is faster. If Notion is where you start, buy the domain add-on and treat it as a draft.

The 2026 bar: viable and lovable, not just shipped

Before 2025, a portfolio proved you could ship. Feasibility is now cheap; showing you shipped something is table stakes. The 2026 portfolio proves two things AI cannot substitute:

Viable: that you identified a problem people and companies will actually pay to have solved, at a market size that justifies the investment. Case studies need a market-sizing argument, a willingness-to-pay signal, and a clear explanation of why this problem was worth building instead of something else.

Lovable: not “clean UI” but that you understood where in the workflow users actually were, anticipated what they would need, and made the product meet them there rather than forcing a behavior change. See feasibility is free for the full reframe.

A case study that says “I ran 20 user interviews and shipped X which increased retention by Y%” is now incomplete. It needs: here is why the market was real, here is what I chose not to build, here is where AI gave us the obvious answer and I pushed back, here is what I would kill if I were starting again.

Hiring managers at AI-native companies (Anthropic, OpenAI, Cursor, Perplexity) report caring as much about your reasoning on feasibility and viability as on UX polish. A portfolio that shows you killed a feature because the market was too small is more impressive than one showing 50 user interviews and a polished prototype.