Compliments are free to produce. That is exactly why founders have so many of them and know so little.
What separates good feedback from agreeable feedback?
Cost. Not to the founder, to you.
An agreeable answer takes no effort and carries no risk. "Looks great, I'd use it" can be said in two seconds by someone who has not thought for one of them. It also cannot be wrong, which is the tell. Feedback that cannot be wrong is not information.
A useful answer requires you to check something real about yourself: whether you would actually change what you currently do, whether you would pay, whether you already have a workaround you are quietly fond of. That checking is work, and the answer often makes you look unhelpful.
What does a good validator actually do?
Three things, reliably.
They separate what they think from what they would do. Those are different reports and only the second one is worth anything. "That's clever" is a thought. "I'd still open the spreadsheet" is a behaviour.
They name the exact moment they lost interest. Not "the onboarding was confusing" but "I stopped at the point where it asked me to connect an account, because I was not going to give it access on a first look." Founders can act on the second. The first just becomes a slide.
They are willing to say they are not the customer. This is the hardest one, because it feels like withdrawing from the conversation. It is the single most valuable thing a validator can say, and almost nobody volunteers it.
Why is being agreeable so hard to resist?
Because the other person is clearly hoping. Someone shows you a thing they made. Their hope is visible. The social cost of puncturing it lands entirely on you, in the moment, in front of them. The cost of not puncturing it lands on them, later, invisibly.
That asymmetry is the whole reason founders cannot get honest answers from people who know them. It is not that friends lie. It is that the incentive to be kind is immediate and the incentive to be accurate is somebody else's problem next year.
Being paid, and being anonymous to the founder, removes most of that pressure. That is not a trick. That is the design.
Is honest the same as harsh?
No, and confusing the two is the most common mistake new validators make.
Harsh is a tone. It usually signals that someone is performing rigour rather than doing it. "This is a terrible idea" contains no information. It cannot be acted on, checked, or argued with.
Honest is a level of specificity. "I would not pay for this because I already do it in a shared doc and the doc is free and my team already lives in it" is a complete, checkable, useful answer, and it is not remotely unkind. The founder now knows exactly what they are competing against.
What does this skill turn into?
Pattern recognition, mostly. After you have looked hard at thirty ideas, you start noticing the same failure shapes early: the product that is a feature, the market that is one customer repeated, the pricing that assumes a budget that does not exist anywhere.
That is a transferable read. It is the same instinct that separates people who evaluate deals well from people who get excited well. It travels into any work where you have to judge something before there is proof, which is a growing share of digital work rather than a shrinking one.
How does this fit into real working life?
It fits the way judgment work generally fits: in blocks, on your own schedule, priced by the session rather than the hour of presence. Most validators do it alongside other things. The people who get the most out of it treat it as a practice rather than a task queue, because the return is compounding rather than per-unit.
A note on the wider judgment economy, kept honest. There is a family of adjacent activities where people are paid or rewarded for calling things right early, including angel investing and equity crowdfunding platforms. It is worth knowing that judging demand for a product and judging an investment are related but genuinely different skills. The first asks whether anyone wants this. The second adds questions about price, dilution, timing and liquidity, and it puts your own money at risk. Anything involving capital carries risk of loss and is regulated differently by country. Validating is not an investment product and nothing here is investment advice. The connection worth drawing is narrower and more useful: the habit of separating what people say from what they do is the same habit in both rooms.
Key takeaways
- Feedback that cost the giver nothing tells the founder nothing.
- Report behaviour, not opinion. "What I'd do" beats "what I think" every time.
- Name the exact moment you lost interest, not a general impression of it.
- Saying "I am not your customer" is the most valuable sentence a validator has.
- Honest is specific. Harsh is just a tone, and it carries no information.
- The underlying skill is early pattern recognition, and it transfers well beyond any one platform.
FAQ
Do I need industry expertise to be a good validator?
For most consumer and small business products, no. You need to be a plausible user or a plausible buyer and you need to report yourself accurately. Specialist categories are a different matter and are usually routed to people with the relevant background.
What if I genuinely like the idea?
Then say so, and then say what would have to be true for you to pay. Enthusiasm plus a condition is useful. Enthusiasm on its own is decoration.
Am I being unhelpful if I have nothing bad to say?
Not necessarily, but check whether you have anything specific to say at all. If you cannot name one moment, one comparison, or one thing you would keep doing instead, you probably have not looked closely enough yet.
Does giving negative feedback hurt my standing?
No. Founders pay for accuracy, and accuracy is what gets a validator asked back. Vague agreement is what gets quietly filtered out.
How do I get better at this?
Volume and honesty about your own misses. Keep a note of the ideas you were wrong about in both directions. The ones you dismissed that worked are more instructive than the ones you praised that failed.