What counts as evidence in your validation, and why "I talked to some users" isn't enough

Here's a pattern we see constantly: a founder writes "I've validated this, I talked to users and they loved it" in answer to a validation question, expecting that sentence to carry real weight. It doesn't, and understanding why will make you better at using Product Nerve AI, and probably a sharper founder in general.

The four levels of evidence. The Validation Engine recognizes a spectrum, not a yes-or-no, when it comes to claims about validation:

None means exactly that, no validation has happened yet, and that's a perfectly honest place to start.

Claimed is a statement without specifics: "I talked to users, they liked it." This is the level most founders default to, and it's the one that moves the needle least, because it's unfalsifiable. Anyone can say it.

Described means you've given specific, checkable detail: how many people, roughly when, what they actually said, what segment they came from. This is meaningfully stronger, because it's the kind of claim that could be checked, even if we're not checking it directly.

Shown is the strongest level: you've uploaded actual evidence to your project's knowledge base, interview notes, a survey export, a spreadsheet of signups, payment records, anything the system can actually see and read. This is the only level that can move an assumption to fully Validated.

Why this matters for your score, not just your confidence. This isn't just about how convincing your report sounds. Evidence level directly caps how validated your assumptions can be classified, and it directly caps your Confidence Index, the measure of how much weight to put on your score at all. A high score built on Claimed evidence throughout will carry a note that the confidence behind it is limited, an honest signal to you that the number is provisional.

What this means practically. If you've actually done customer research, don't just describe it, upload it. Interview transcripts, notes, screenshots of survey results, waitlist exports, anything real. It takes two extra minutes and it changes how the engine treats every claim connected to it. If you haven't done real research yet, the honest answer is "none" or a properly described early conversation, not an inflated claim that will just get flagged as unsubstantiated anyway.

This evidence structure exists for one reason: to make sure your validation reflects reality, not how persuasively you can write about your idea. The system rewards founders who've done the work and shown it, not founders who are good at sounding validated.