How to validate a B2B SaaS idea pre-launch

To validate a B2B SaaS idea pre-launch, run a four-rung experiment ladder before writing production code: 15 customer interviews using the Mom Test, a painted-door landing page, a concierge or smoke MVP, then signed LOIs or pre-payments from at least three buyers. Each rung is a kill switch. Build only after money commits.

Most founders skip rungs and build first. That's why most pre-launch SaaS dies on Stripe day one with a polished product and zero buyers.

What "validate b2b saas idea" actually means in 2026

Validation is not market research. It's not a survey, not a Reddit post, not a Twitter poll. It's the smallest possible experiment that proves a real buyer at a real company will pay real money to solve a specific pain.

For B2B SaaS the signal threshold is lower than B2C. You don't need 10,000 weekly signups. Thirty paying customers at $200 MRR is a signal. Three signed LOIs at $500 per month is a signal. A founder who tells you "I'll wire $1,500 today if you can ship by Friday" is the signal.

The ladder below treats validation as a sequenced experiment. Each rung gates the next. If a rung fails, you pivot or kill, you do not advance. The point is to spend cash and time only on ideas that have already proven they can charge.

The four-rung validation ladder (run them in order)

Validation rung	What you ship	Cost	Pass signal	Kill signal
Customer interviews	Calendly link + Mom Test script	$0	12 of 15 describe pain in last 30 days	<60% have a current workaround
Painted-door page	Single landing page + ad spend	$200-500	10%+ visitor-to-signup, under $10 per signup	<3% conversion after 1k visits
Concierge MVP	Manual service via spreadsheet + email	$0-200	5 customers complete the workflow weekly	Customers churn after 2 sessions
LOI / pre-sale	1-page LOI or Stripe Payment Link	$0	3+ signed LOIs or pre-payments	0 commits after 10 asks

Run them in order. Skipping rungs is how founders end up with a $40,000 MVP and a calendar full of "interesting, get back to me" replies.

Rung 1: Mom Test interviews (15 buyers, not friends)

Rob Fitzpatrick's The Mom Test is the only book on validation that matters. Its core rule: people lie to be nice. Questions about the future ("would you use this?") return polite garbage. Questions about the past return data.

Here's the script. Three questions, no pitch.

1. "Walk me through the last time you tried to solve [specific pain]. What did you actually do?"
2. "How much time or money did that cost you, and how often does it come up?"
3. "What have you spent money on in the last 90 days to fix it?"

You're listening for two things: a specific recent example, and money already moving. If a prospect says "we built an internal Notion doc and a Slack bot" or "we paid Zapier $400/month to fake it," you've found pain. If they shrug and say "yeah, that would be nice to fix," you have a feature, not a problem.

Kill criteria: if fewer than 9 of 15 prospects describe attempting a workaround in the last 30 days, the pain isn't urgent. Pivot the ICP or pivot the problem.

Where to find 15 buyers in two weeks: LinkedIn Sales Navigator (filter by title, post a one-line ask), targeted cold emails to 200 prospects (you'll get 15 calls), founder communities like Pavilion, RevGenius, and CMA. Avoid friends, accelerator cohort mates, and Twitter followers. They'll over-praise.

Rung 2: Painted-door tests (a button that doesn't work yet)

A painted-door test is a landing page that advertises a feature or product you haven't built. Visitors who click "Get started" hit a "We're rolling this out, leave your email" form. The click is the demand signal. The opinion doesn't matter; the click does.

This is how Notion's 2015 reset worked. Ivan Zhao and team had nearly run out of money on the original product, scrapped most of it, and rebuilt around a smaller painted-door promise (blocks for thinking) before relaunching. The signal was clicks and paid retention, not survey results.

Build it in a day with Vercel and a single Next.js page or a Framer template. Keep it simple. One headline (literal value prop, not clever wordplay), one screenshot or 30-second Loom, three bullets of what it does, one signup field, one pricing line.

Then drive 1,000 targeted visitors. Use LinkedIn ads ($300-500 budget, target by job title and company size), three relevant subreddits, two Slack communities. Cost target: under $10 per signup. Conversion target: 10%+ visitor-to-email.

If you spend $500 and get fewer than 30 signups, the headline is wrong, the audience is wrong, or the painted door is wrong. Iterate the headline three times before killing the rung. Most failed painted doors are a copy problem, not an idea problem.

Rung 3: Concierge MVP or smoke test (deliver the value manually)

This is where founders cheat the most. They want to skip to "build the real thing." Don't. Serve five customers manually first.

A concierge MVP is the product delivered with humans, spreadsheets, and email. If your SaaS is going to score sales calls with AI, you score them yourself in a Google Doc and email back the report. If your SaaS is going to auto-generate compliance docs, you write them in a Notion doc and send a PDF. If your SaaS is going to schedule Slack standups, you DM each team member yourself at 9am.

Linear is the cleanest example of this discipline turned into product. Karri Saarinen, Tuomas Artman, and Jori Lallo built it as the issue tracker their team wished they'd had at Coinbase, Uber, and Airbnb. The first version was a tool for themselves and a small group of design-led startups. They didn't pitch a market; they shipped what already worked manually for their own engineering teams, then opened it up.

The concierge phase teaches you the workflow no spec doc captures. You'll learn that the export needs CSV not JSON, that customers want Slack alerts at 8am not 9am, that the actual decision-maker is the head of RevOps not the CFO. None of this shows up in interviews.

Pass signal: five customers complete the workflow weekly for two weeks. Kill signal: customers stop responding after the second session, or the manual delivery takes more than 90 minutes per customer per week (it won't scale even with software).

Rung 4: Letters of intent and pre-sales (the only signal that matters)

Money commits or it doesn't. This rung is the only one that survives contact with reality.

Three options, in order of strength:

1. Pre-payment via Stripe Payment Link. Send a payment link for the first month or first quarter. Offer 50% off for paying before launch. If three buyers wire money before the product exists, you have a real B2B SaaS.
2. Signed Letter of Intent. A one-page LOI saying "buyer commits to a paid pilot of [scope] at [price] starting [date]." Three LOIs is the minimum bar.
3. Verbal commit on a recorded call. Weakest signal. Use only as a stepping stone to the LOI.

Pylon (YC W21, B2B support platform now serving Linear, Hightouch, and Deel) sold its first customer contracts before the product was fully built. Marty Kausas's team ran sales calls on the painted-door promise, signed pilot agreements, then shipped against the spec the first three buyers agreed to. By their Series A in 2024 the playbook was the same: sell the contract, build to the contract.

Use this LOI template:

"[Company] commits to a paid pilot of [Product] beginning [Date], at $[Amount] per [month/quarter] for [Term]. Pilot success criteria: [3 measurable outcomes]. Either party may exit at the end of the pilot."

Send it. If three buyers sign in 2-3 weeks, you've validated. If 0 sign after 10 asks, the offer, the price, or the buyer is wrong.

How to sequence the validation: 6 weeks, no code

Six weeks is the budget. End to end, no production code shipped.

• Week 1-2: Interviews. 15 calls. Mom Test script. Three written summaries per call.
• Week 3: Painted-door page. Build, drive 1,000 visitors, hit 10% signup or iterate the headline.
• Week 4-5: Concierge MVP. Onboard 5 buyers, deliver value manually, document the workflow.
• Week 6: LOIs and pre-sales. Send 10 asks, target 3 commits.

If week 6 ends with zero signed LOIs, you have your answer. The idea isn't ready, or the ICP is wrong, or the price is wrong. Cycle back to interviews with a sharper hypothesis. Don't write code yet.

A useful counter-rule: every week you spend coding before rung 4 passes is a week you're losing. Code is the most expensive form of opinion.

What to build first (and who builds it)

Once three LOIs are signed, you have permission to build. Now the question is who and how.

The default 2026 stack for a B2B SaaS MVP: Next.js on Vercel, Supabase or Neon for Postgres, Clerk or Supabase Auth for auth, Stripe for billing, Resend for email. With AI-assisted coding (Cursor, Claude Code) a single competent engineer ships this in four to six weeks for $20-40k all-in.

Three paths to actually ship it:

• Founder-led with AI tools. If you can write code, do this. Cursor plus Claude can take a non-engineer founder surprisingly far. Cap: about a third of MVPs hit a complexity wall (auth edge cases, async jobs, payment webhooks).
• Hire a freelancer. Upwork, Toptal, your network. Trade-off: 2-4 week vetting cycle, monthly contracts, no replacement guarantee, you become the project manager.
• Book a Cadence engineer for the build. Every engineer on Cadence is AI-native by default, vetted on Cursor, Claude Code, and Copilot fluency before they unlock bookings. For a typical B2B SaaS MVP, a mid engineer at $1,000 per week ships the core product in four to six weeks. If the scope includes architecture decisions, a senior at $1,500 per week is the right call. The 48-hour free trial means you sanity-check fit before any week is paid. If your LOIs are signed and you want a mid engineer on Monday, book your first Cadence engineer and use the 48-hour trial to confirm fit before any money moves.

The right rule: don't book or hire anyone until rung 4 passes. Once it does, the right MVP scope is one user, one workflow, one risky assumption. A mid engineer with a clear LOI-derived spec ships that scope in a month.

If you're a non-technical founder weighing your options, the build vs buy framework decides each feature on differentiation and time-to-ship before any code is written.

Common founder mistakes when validating

• Building before money commits. The expensive mistake. Six months of code, no buyers, founder out of runway.
• Asking friends instead of buyers. Your accelerator cohort isn't your ICP. Family isn't your ICP. Twitter followers aren't your ICP.
• Vanity-metric landing pages. A landing page with 800 newsletter signups but zero LOIs is not validation. Email collection is the second-weakest signal.
• Skipping the LOI conversation because it's awkward. The awkward 15-minute conversation about pre-payment saves six months of misallocated build.
• Treating one positive interview as proof. One yes is noise. Twelve out of fifteen is signal.
• Confusing curiosity with intent. "This is interesting, send me a demo when it's ready" is not a buying signal. It's a polite exit.
• Hiring a senior engineer pre-LOI. The right answer is usually a mid for the MVP build; a senior is over-spec for a 4-week scope.
• Building auth, billing, or admin from scratch. Use Clerk, Stripe, Retool. Save the custom code for the differentiated workflow.

If you're a non-technical founder reading this and wondering whether you need a co-founder to do any of it, the honest 2026 answer is no. You don't need a technical co-founder; you need a clear MVP and someone to ship it once the LOIs are in.

Already at rung 4 with signed LOIs? Book a vetted, AI-native engineer on Cadence and use the 48-hour free trial to ship the first slice before paying for a single week. Weekly billing, no notice period, replace any week if fit isn't right.

FAQ

How long should pre-launch validation take?

Six weeks of focused work, end to end. If you can't get three LOIs or pre-payments in six weeks, the idea probably isn't ready, or your buyer list is wrong. Either is useful information.

How many customer interviews do I need before building?

Fifteen is the floor. Use the Mom Test: ask about how they handled the problem last week, not whether they would buy a hypothetical product. Past behavior is signal; future intent is theater.

What's a painted-door test?

A landing page that advertises a feature or product that doesn't exist yet. Visitors who click signup hit a "we're rolling this out, leave your email" form. You measure conversion (10%+ to signup is the bar) and cost per signup (under $10 for B2B). Opinion doesn't count, the click does.

Should I build the MVP before getting LOIs?

No. Letters of intent and pre-payments are the only validation signals that survive contact with reality. Build only after at least three buyers commit money or sign a one-page pilot agreement. Pylon (YC W21) sold contracts before fully building; the discipline survives to Series A.

When should I hire an engineer to build the MVP?

After rung 4 passes. For a $20-40k MVP scope, a mid-tier Cadence engineer at $1,000 per week ships a real product in four to six weeks. Junior engineers at $500 per week handle integrations and cleanup. Lead engineers at $2,000 per week make sense only when architecture decisions are in scope (a rare case for a pre-launch MVP).