No black box. Every verdict traces to the evidence.
A verdict you can't inspect is just an opinion with better production values. Here is exactly how Valtr moves from your question to a call you can defend. And where every number came from.
Four stages, from your idea to the verdict.
Your inputs
The idea, the wedge, the market, and your real constraints. Capital, time, location, and skill. Stated plainly, in about twenty minutes.
Grounded research
Valtr pulls the reference set: comparable ventures with outcomes attached, competitors, buyer profiles, and market sizing. Sourced and tiered, never assumed.
The model
Evidence is weighed into the three legs. Demand, supply you can beat, and a reason you win. And the moat pillars. Constraints are matched to what the winners needed.
The verdict
A verdict word plus a conviction level, in plain language framed for who you are. And rerunnable as new evidence lands.
Three legs. Two usually stand; one is yours to prove.
Every call rests on demand for the job, supply you can beat, and a reason you specifically win. Valtr shows the state of each. And names the one your first test exists to resolve.
Is the job real and budgeted?
Named pain, felt regularly, with money already moving against it. The difference between a problem people admit and one they pay to fix.
Can you beat what's there?
Who serves the job today and how well. An opening is only real if incumbents can't close it with a roadmap slide.
Why do you win, and hold?
The edge that compounds with use rather than eroding. Usually the load bearing assumption, and the thing a cheap test can settle.
Every claim cites its source and its tier.
Valtr never flattens a Reddit thread into an IRS figure. Every number shows where it came from and how much to lean on it.
Direct customer outcome data we measure with explicit opt in, lagged 90 days. The strongest evidence Valtr has.
Public outcome data with verified provenance. Named, traceable, and checkable by you.
Aggregated proxy data from named sources, used when direct data is thin. Always labelled as such.
A word, not a grade. Framed for who you are.
Nine sections, plain English, no filler.
Every number traces to a source; every claim has a drill target. Here's what a full read contains.
The verdict
The call and the conviction behind it. The one line read and the move that follows.
Executive summary
Why it's this verdict, and what it takes to keep it one.
Market opportunity
The opening, sized from the whole category down to the slice you can win first.
The field
Who you're up against, and where the open corner sits.
Your ICP
Who buys first, why they buy, and who signs.
Cheapest tests
The fastest experiments that convert the unmeasured leg into evidence.
Economics
Your stated inputs, and what the capital buys before it pays its own way.
Risk & scenarios
What could break it, how likely, and the spread of where it lands.
Provenance
How the read was computed, with every figure tagged researched or reasoned.
How far can you trust it?
Is this just ChatGPT with a wrapper?
No. The verdict is computed on a reference set of real ventures with outcome data attached, not generated from a model's opinion. You can inspect the cohort it was scored on.
Can I recompute the verdict myself?
Yes. Every figure that drives the call is shown, with its source and tier. The math is yours to check and to rerun as your own evidence lands.
What if my idea is too niche for a cohort?
Then Valtr says so. A thin reference set returns a needs a test verdict rather than a false confident number, and points to the test that would resolve it.
How current is the evidence?
Measured data is lagged 90 days for accuracy; verified and proxy tiers are timestamped. Reads are rerunnable as sources update.
See the method on your idea.
Your first verdict is free. A clear call, the evidence behind it, and a full report in minutes.
Get your verdict →