Validate a UK SaaS idea in a weekend with Claude Code (2026)

Here is the expensive version of the story. A UK builder has an idea on a Tuesday, gets excited, and spends the next six weekends building it. They ship, post it on Reddit, get 47 likes and four sign-ups, none of whom ever log in again. The idea was not bad. Nobody had ever checked whether anyone wanted it.

Here is the cheap version. The same builder has the same idea on Tuesday, and by Sunday night they have a landing page live, GBP 120 of traffic through it, a waitlist of 30 people, three founding-member pre-orders, and a calendar with five interview slots. Now they build - against a queue of people who already paid.

The cheap version is one weekend. This post is the sequence, and the harness you ship to run it is itself a small Claude Code build. Validation is not the opposite of building. It is the first, cheapest thing you build.

Why this matters more than your code

Roughly 85% of SaaS ventures fail inside 18 months, and in 2026 the dominant cause is not bad engineering - it is premature scaling without a demand signal. Builders are better than ever at shipping (you can stand up a first Claude Code project in an afternoon now) which makes the what to build question the binding constraint, not the how. A weekend of validation buys you out of the most common way indie SaaS dies.

The mistake is treating validation as research you do in your head. Asking your friends if they would use it does not count - they love you, they will say yes. Posting on LinkedIn and collecting likes does not count - likes are not pre-orders. Validation has to put something real in front of strangers and watch what they do.

Step 1: Mine the complaints (Saturday morning, 2 hours)

You are not looking for your opinion. You are looking for the exact words real people use when the problem bites.

Three places UK builders find them:

1. Reddit. Search the relevant subreddit for the problem plus words like "frustrating", "hate", or "anyone else". Read the threads, not just the titles.
2. One- and three-star reviews of the nearest existing tool on G2 or Trustpilot. One-stars tell you what is broken; three-stars tell you what is almost right - that gap is your wedge.
3. Upwork / PeoplePerHour briefs where someone is paying a human to do the thing your software would automate. People paying for a manual fix are people with budget for an automated one.

Write down the top three recurring problems and, crucially, the language attached to each. Their words, not your paraphrase. Those phrases are about to become your landing page.

Step 2: Build the smoke-test harness with Claude Code (Saturday afternoon, 1 evening)

A smoke test is a landing page that describes the product as if it already exists, with one job: capture an email when someone wants in. You are buying a conversion rate, not selling a product.

The stack is AI-native and free to host:

• Next.js page on Vercel - the page itself.
• Supabase - the waitlist table.
• Resend - the confirmation email.

Claude Code writes all three in one session. Open the project, give it the headline and subhead you mined in Step 1, and let it scaffold:

claude -p "Create a Next.js landing page for a SaaS waitlist.
Headline and subhead provided below. One email-capture form.
On submit: insert {email, source, ts} into a Supabase table called
'waitlist', then call Resend to send a confirmation email.
Use env vars SUPABASE_URL, SUPABASE_ANON_KEY, RESEND_API_KEY.
EN-UK copy. Headline: <your headline>. Subhead: <your subhead>."

The form handler is about 15 lines:

// app/api/waitlist/route.ts
import { createClient } from "@supabase/supabase-js";

const supabase = createClient(
  process.env.SUPABASE_URL!,
  process.env.SUPABASE_ANON_KEY!
);

export async function POST(req: Request) {
  const { email } = await req.json();
  await supabase.from("waitlist").insert({ email, source: "smoke-test" });

  await fetch("https://api.resend.com/emails", {
    method: "POST",
    headers: {
      Authorization: `Bearer ${process.env.RESEND_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      from: "hello@yourdomain.co.uk",
      to: [email],
      subject: "You are on the list",
      text: "Thanks - you are in. We will email you the moment it is ready.",
    }),
  });

  return Response.json({ ok: true });
}

Deploy to Vercel (git push, done). The waitlist lives in your own Supabase, so when you build the real product it carries straight over - the harness is a head start, not throwaway. This is the same one-weekend rhythm as the weekend vibe-coding playbook, pointed at validation instead of a finished product, and it reuses the exact Resend pattern from the morning-brief agent.

Step 3: Drive real traffic (Sunday morning)

A landing page with no traffic is a hypothesis with no test. You need cold strangers, not your network.

GBP 100 to GBP 150 of paid traffic is enough to read a conversion rate with confidence. Options for a UK builder:

• Reddit ads targeted at the subreddits where you found the complaints - cheap, and the audience is already problem-aware.
• Meta ads if your audience is consumer-shaped.
• Google search ads on the exact problem keywords if the intent is commercial.

Send 300 to 500 visitors. That is enough to separate a 5% conversion from a 1% one.

Step 4: Read the signal honestly

Email-capture rate (cold traffic)	Read
5% or higher	Real demand signal - proceed to interviews and pre-sale
2% to 5%	Grey zone - the idea may be real but the message is off; rewrite the headline and re-run
Under 2%	Nobody wants this as framed - pivot the angle or the audience

But the email rate is the weak signal. People give an email when mildly curious. The two strong signals come next.

Step 5: Mom Test calls (the following week)

Interview 10 people from the waitlist. The trap is asking about the future ("would you use this?") - people are terrible at predicting their own behaviour and kind enough to lie. The Mom Test fix is to ask only about the past:

• "Tell me about the last time you hit this problem."
• "What did you do about it? What did that cost you?"
• "What have you already tried? Why did it not stick?"

You are listening for a problem that already costs them time or money today. If they cannot recall the last time it bit, the pain is hypothetical and the idea is weaker than the landing page suggested.

Step 6: The pre-sale - the only signal that counts

Offer founding-member pricing to the waitlist: a discounted lifetime or annual rate for buying before launch. This is where willingness-to-pay separates from willingness-to-express-interest.

The rule, unchanged for years because it works:

• Three pre-orders at founding-member pricing is your green light. Build.
• Zero pre-orders after a clean landing page and real traffic is your data, not your failure. The idea, the angle, or the audience is off. Pivot one of the three and run the weekend again - you are only out GBP 120 and a Saturday.

Money changing hands is the only validation that proves people will pay rather than that they will nod.

The cost line, honest

Item	Cost (GBP)
Vercel hosting (Hobby)	0.00
Supabase waitlist (free tier)	0.00
Resend confirmation emails (free tier)	0.00
Claude Code (Pro sub you likely already have)	0.00 marginal
Paid traffic to the smoke test	100.00 - 150.00
Total	100.00 - 150.00

Three founding-member pre-orders at, say, GBP 60 each more than covers it. You can validate net-positive.

The point

Validation feels like the boring bit you skip to get to the fun bit. It is the opposite. The fun bit is building something 30 people are already waiting for. The weekend you spend on a smoke-test landing page, a little ad spend, and six honest phone calls is the weekend that decides whether the next six weekends are worth it. You have the tools to build the harness in an evening. Build that first.

---

New here? IdeaStack publishes one deeply researched UK business opportunity every Thursday - real keyword data, competitor analysis, builder prompts. See the latest free report.