Every hypothesis, shown — not just the winner.
Ask a market question. Ledger checks every angle worth checking, and shows you what it found — including where it found nothing.
how it works
Four steps, no black box.
Ask a market question
One free-text question — the kind you'd ask a smart friend before starting something. "Is there a viable business in X?", "Does anyone actually pay for Y?", "Where does Z break down?"
Up to five hypotheses generated
We derive the angles most worth checking — different monetization paths, audience segments, sub-niches. You see all of them, not just the ones with an answer.
Evidence collected for each
Public-web search with dual-query reformulation via Tavily. Every source URL, snippet, and fetch timestamp is kept — nothing is fabricated, nothing is silently smoothed.
Deterministic verdict per hypothesis
A fixed rule over the classified evidence: strong signal, mixed, or insufficient evidence. Same evidence in, same verdict out. No coin flips.
inside a report
What you actually get.
H1 · Recipe blogs monetized via display ads
strong signalThree independent sources (Mediavine, AdThrive, and RaptiveEasy) document display-ad networks operating specifically for food / recipe content, with public RPM figures and creator payouts.mediavine.com/publisher-guidecites $20–$40 RPM as typical.
Verdict rule: 2+ domains × paid_behavior_verified → strong_signal
H4 · Air-fryer-only content sustains a standalone site
insufficient evidenceSearch returned four listicles ranking air-fryer recipe blogs, but none quantified revenue for an air-fryer-only site (as distinct from a general recipe site with air-fryer coverage). This is a gap in the evidence, not a signal either way.
Verdict rule: 0 rows in verified/claimed classes → insufficient_evidence
questions
Reasonable to ask first.
How is this different from asking ChatGPT the same question?+
ChatGPT will answer from its training data with confident prose and no citations. Ledger runs live web search per hypothesis, keeps every source URL, and never presents an unsupported claim. The trade: 3-10 minutes and €29 versus 30 seconds and nothing.
What if the report is wrong or unhelpful?+
Every hypothesis links to its source URLs — you can trace and verify each claim. If a pipeline error prevents delivery, we automatically retry twice; if it still fails we contact you within 24 hours to re-run or refund. Reports that produced their intended output are not refundable simply for being unwelcome.
How long does it take?+
Typically 3-10 minutes from payment to a delivered report. You get an email when it's ready. Nothing to babysit.
Can I ask any question?+
Ledger is optimised for market questions — viability, monetization, comparison, "does anyone actually pay for X". It's a poor fit for advice questions ("should I quit my job?") and questions requiring real-time data ("what's AMZN trading at?"). See the AI Disclosure for what it should not be used for.
What data do you keep?+
Your email, the question you asked, the report we generated, and cited evidence. Nothing more. See the Privacy Policy — we don't sell it, don't use it for training, don't share it beyond named sub-processors (Anthropic, Tavily, Stripe, Supabase, Vercel, Railway, Resend).