The most expensive software is software nobody uses. Every month I scope a build for a founder whose idea is great in their head but hasn't been tested anywhere outside it. Sometimes it works. Often it doesn't. The cheapest way to find out is to validate before you build, not after.
Here are the five validation tactics I see actually work, in order of how much they cost.
1. The 20-conversation rule
Cost: $0. Time: 2–3 weeks.
Find 20 people in your target market and have a 20-minute conversation with each. Not pitch, not sell, listen. Ask:
- "What's the most painful thing about [your problem area]?"
- "What have you tried so far?"
- "Where does that solution fall down?"
- "If you had a magic wand, what would change?"
- "How much would you pay for that?"
If 15 of 20 people describe a similar pain that nobody has solved well, you have signal. If 20 of 20 describe completely different things, you haven't found a problem yet.
2. The fake door test
Cost: $200–$1,000. Time: 1 week.
Build a landing page that describes your product as if it exists. Run a small ad campaign ($300–$500 of LinkedIn or Google ads) targeting your audience. Measure two things: click-through rate (does the message land?) and email signup rate (do they want it enough to leave a contact?).
10–20% signup rate from cold traffic is a real signal. 1–2% is the background noise of the internet.
3. The pre-sale
Cost: $0. Time: a few days.
Take the most engaged five people from steps 1 and 2. Tell them you're building it. Offer a 50% lifetime discount if they pre-pay now. If three of five pull out their card, you have validation that exceeds anything any survey can give you. If zero do, you don't have a product yet. You have a curiosity.
The most expensive feedback you'll ever get is silence after launch. The cheapest is "no, I wouldn't pay for that" before the first commit.
4. The concierge MVP
Cost: $0–$2k. Time: 2–6 weeks.
Manually deliver the value your software would deliver, by hand, for 5–20 paying customers. Examples I've seen work:
- "AI-powered marketing report" → you write the report yourself, send a PDF.
- "Matching service for X" → you read the form, you pick the match by hand, you email.
- "Auto-generated meeting notes" → you transcribe the recording manually.
This is slow per customer, but it's effectively free to set up. And it teaches you what the automated version actually needs to do, which is information you literally cannot get any other way.
5. The smallest possible MVP
Cost: $5k–$15k. Time: 4–8 weeks.
If, and only if, the previous four steps gave you signal, build the smallest possible product that delivers the validated value. The cut-list framework from "The MVP Mistake" applies here. If your scope is more than 30–40% of what you originally imagined, cut more.
Anti-patterns to avoid
- "My friends and family love it." They love you. Different signal.
- "This big company would obviously want this." Selling to enterprise from a deck is a multi-year journey. Validate with smaller buyers first.
- "I just need to build a demo to show people." A landing page is the demo. Save the engineering for the real build.
- "My validation is that competitors exist." Competitors = market exists. Doesn't validate that you can win.
- "I can't validate without the product working." You can. You're rationalising the part that's fun (building) over the part that's awkward (selling).
The point
Validation isn't about being cautious. It's about spending your build budget on the right idea. The first idea you have is rarely it. Run the cheap experiments first; save the engineering for the validated version.
The bottom line
Talk to 20 customers, run a fake-door test, pre-sell to your warmest five, deliver concierge MVPs, then build the smallest possible product. In that order. Most ideas don't survive step one, and that's the point.
Once you've validated, FindDevs gets you three Australian developers who specialise in shipping MVPs in 6–8 weeks. Free, in 24 hours.