Why I Built GhostView

The most dangerous feedback a pre-PMF founder gets is the kind that feels honest.

A user interview goes well. The person nods at your key features, says the pricing sounds fair, mentions three things they'd want. You walk away feeling like you learned something. Then you look at your session recordings and watch the same person spend four minutes staring at your onboarding screen, never find the button you told them about, and leave without doing the one thing the whole product is built around.

User interviews are curated. People are polite. They tell you what they think you want to hear, or what they think is the most useful answer, or what they genuinely believe—which turns out to be different from what they actually do.

Session recordings don't lie. They just show you what happened.

The problem is that session recording tools are built for growth-stage companies. They assume you have a product marketing team analyzing heat maps of a checkout flow visited by ten thousand people a day. They're priced accordingly. And they track everything by default, which makes a lot of people uncomfortable about adding them to an early product.

GhostView is the version I wanted for pre-PMF work.

It's invite-gated, which matters when you have twenty users, not twenty thousand. You generate a link. The person installs a lightweight snippet. Their sessions are recorded and you can replay them. It tracks the events you care about and exports everything to JSON so you can take your data with you.

The part I built for myself is the AI synthesis. If you've watched three sessions from three different users across three different days and you want to know what the pattern is—what the shared confusion is, what the moment was that made them leave—you can ask. It reads across sessions and gives you an answer. It's optional, but it's the feature I personally needed.

The live site is at ghostview.dev. It's in early access.

What you build with imperfect information is usually fine. What you build with information that's wrong in a specific direction—the direction of "users say yes"—tends to fail in expensive ways. That's the problem GhostView is trying to fix.