9 blocks. Real tests. One confidence score.

Find out if your idea holds up
before you spend a single rupee building it.

Write a falsifiable hypothesis, run a real test, log strong / weak / fail — the score is a weighted sum you can audit.

Hosted surveyAd testLive waitlist pagePricing widget
Scroll to explore

The 9 blocks, in sequence

Customer SegmentValue PropositionMarketing ChannelEngagementIncome SourceAsset ValidationCore OperationsStrategic AllianceExpense Validation
9
validation blocks, in sequence
0–100
auditable weighted score
3
real tests per block

Startups don’t die from bad ideas. They die untested.

The same three failure modes repeat in startup post-mortems:

42%
No market need
Built before any falsifiable evidence the demand was real.
29%
Ran out of cash
Spent on building before margin, CAC, and LTV were confirmed.
23%
Wrong team or ops
Operating and partnership assumptions never stress-tested against reality.

How Litmus actually works

Exactly what happens inside every block — in four moves.

Step 1 · Hypothesis

Fill in the blank, not a blank page

2–3 falsifiable sentences per block, each self-rated 0–100 — the lowest-rated is what you test first.

I believe who struggle because why enough to action
Chip-driven, not vagueSelf-rated 0–100 eachSame rigor, all 9 blocks

Step 2 · Real tests

Three named tests per block — not opinions

Demand

hosted survey

~40% hint

Reach

your own ad

~8% signup hint

Intent

live waitlist page

real visits + signups

Pricing / WTP widgetContribution margin, your currencyLTV:CAC to 1 decimal · ~3× benchmark

CAC auto-pulls from Marketing Channel — never re-entered.

Every threshold is a hint, never a gate — you can always log a fail.

Step 3 · Verdict

A score you can audit

Demand 40 + Reach 25 + Intent 35 = 100

strong = full · weak = half · fail = 0

Validated ≥70Promising 40–69Not yet <40

Fixed weights, visible cutoffs — no hidden adjustments.

Autosaves 600ms after every change

Step 4 · Redo it right

Re-run one stage. Iterate the whole block.

Re-run

Clears one stage’s result; the rest stays untouched.

Iterate

Archives score + band to history, restarts at step one.

Past attempts stay visible, never overwritten.

Your test page is already hosted

Every test gets a live /l/ link — visits and signups counted straight into the stage judging them.

litmus.lapaas.com/l/your-idea
“32% signed up from the page”
One shareable /l/ linkLive visit + signup countsNo separate form tool

Drop it into an ad, a DM, or a bio.

The 9 blocks, in working order

Same shape every time: hypothesis → named tests, logged strong / weak / fail → weighted 0–100 verdict.

01

Customer Segment

Weighted 40/25/35: hosted survey, small ad test, live waitlist page.

3 stages, weighted 40/25/35: Demand runs a hosted survey (hint: roughly 40% should say they need it), Reach runs a small ad test against your own waitlist page (hint: roughly 8% signup rate), and Intent is that waitlist page itself, tracking real visits and signups with a computed signup %.

02

Value Proposition

A WTP widget (≥5% buy-conversion) and a must-have survey (≥40% “very disappointed”).

Write the promise as 'We promise [who] — better than [alternative] because we're [why],' then test it with a pricing/willingness-to-pay widget (flags real WTP at ≥5% buy-conversion) and a Sean-Ellis-style must-have survey (flags must-have status at ≥40% 'very disappointed').

03

Marketing Channel

Real spend, real response — its CAC feeds Income Source’s LTV:CAC math.

Pick and test acquisition channels with real spend and real response — the CAC this produces feeds directly into the Income Source block’s LTV:CAC math, so your channel numbers aren’t sitting in isolation.

04

Engagement

Log real usage evidence for why people come back — not gut feel.

Hypothesize how and why people come back, then log real usage evidence per stage instead of guessing at retention from a gut feeling.

05

Income Source

Three currency-aware calculators — Pricing/WTP, Contribution Margin, LTV:CAC.

Three currency-aware calculators: Pricing/WTP (visitors → buy-clicks → conversion %), Contribution Margin (price minus delivery cost, shown in your venture’s currency plus margin %), and LTV:CAC (ratio to 1 decimal plus payback-in-months, benchmarked around a 3× threshold — CAC is pulled straight from your Marketing Channel results).

06

Asset Validation

Stress-test the skills, relationships, IP, and equipment you’re assuming you have.

Stress-test what you’re assuming you already have or need to build — skills, relationships, IP, equipment — with the same hypothesis-then-evidence structure as every other block.

07

Core Operations

Does delivery hold up at the volume your other blocks assume?

Validate whether your delivery process holds up under real conditions at the volume your other blocks assume, logged as strong/weak/fail evidence like everywhere else.

08

Strategic Alliance

Are the partners your plan leans on real, willing, and available?

Test whether the partners or suppliers your plan leans on are real, willing, and available — with the same falsifiable-hypothesis format, not just logical on a slide.

09

Expense Validation

Cost assumptions vs real quotes — closing the loop with Income Source margins.

Pressure-test your cost assumptions against real quotes and real numbers, closing the loop with the margin math from Income Source.

AI Co-Founder

Stuck on a block? Ask your AI co-founder.

It knows your venture and your blocks — not a generic chatbot in another tab.

Pressure-test a hypothesisSanity-check a priceTighten a waitlist headlinePlan the next test
Bring your own API key today · Litmus credits coming soon

AI co-founder

“Your Customer Segment score is 81, but Income Source is still unproven — validate whether they’ll actually pay before you scale marketing spend.”

Chat never moves your score — only logged test results do.

Learn How to Validate

Expert guides and frameworks to help you build the right thing.

Chime Business Model: How Fee-Free Banking Built a Nasdaq-Listed Neobank with 10.2M Members
Fintech•23 min read

Chime Business Model: How Fee-Free Banking Built a Nasdaq-Listed Neobank with 10.2M Members

Complete breakdown of how Chime disrupted traditional banking with no-fee accounts and early direct deposit, IPO'd on Nasdaq in 2025, and reached its first GAAP-profitable quarter serving underbanked Americans.

Read article
Kraken Business Model: How a Security-First Crypto Exchange Built a $10B Infrastructure
Fintech / Crypto•33 min read

Kraken Business Model: How a Security-First Crypto Exchange Built a $10B Infrastructure

How Kraken transitioned from a pioneer Bitcoin portal into a global digital asset ecosystem, focusing on institutional security, staking yields, and a 'bank-grade' regulatory bridge.

Read article
Reddit Business Model: Monetizing the 'Front Page of the Internet'
Social Media / Information•25 min read

Reddit Business Model: Monetizing the 'Front Page of the Internet'

How Reddit turned internet culture into a public company by licensing data to AI giants and building a localized ad engine.

Read article

FAQ

Quick answers about the method, the scoring, and the AI.

No. A canvas is a static diagram you fill in once. Litmus is sequential and evidence-gated: each of the 9 blocks makes you write a falsifiable hypothesis, run 2-3 named real-world tests (a hosted survey, an ad test against your own waitlist page, a pricing widget, a margin calculator), and log a strong/weak/fail outcome before a confidence score is computed. Nothing renders a score until you’ve logged real results.

Each block’s test stages carry fixed weights that sum to 100 (Customer Segment, for example, is Demand 40 / Reach 25 / Intent 35). A ‘strong’ result earns the full stage weight, ‘weak’ earns half, and ‘fail’ or an unlogged stage earns zero. The stages sum into a single 0-100 confidence number, banded as Validated (≥70), Promising (≥40), or Not yet (below 40).

Yes, on purpose. The ‘log result’ button is only gated by minimal input (you typed a headline, or entered the numbers a calculator needs) — never by whether you hit the hint-stated benchmark. You can log a fail even when the on-screen hint says “~40% is a good sign” and you got 12%. Litmus is built to let you see a true negative, not to hide one from you.

Both, depending on scope. ‘Re-run’ on a single stage clears just that stage’s logged outcome so you can redo one test without touching the others. ‘Iterate’ on the Verdict screen archives your current confidence score and band into a visible iterations history, resets the whole block to hypothesis step one, and increments an iteration counter — so you can rebuild the case with a sharper hypothesis and still see what you tried before.

Your riskiest assumption
is testable this week.

Pick the block you’re least sure about, write the hypothesis, run the first stage. Free to start, no card required.

A companion Litmus handbook is in the works. The workspace is the product — free today.