How to assess whether an idea represents a real market gap?

• Did someone pay
• Did someone commit
• Did someone refer another buyer

Step 1 — Start with the pain, not the idea

A real gap is a problem people complain about without being prompted. It surfaces in Slack rants, in Reddit threads at 1 a.m., in the workaround spreadsheet that has 14 tabs and a person whose entire job is maintaining it.

Here's the test I use: if you remove the word "should" from the problem, does it still exist? "People should track their water intake better" is not a gap. "Pharmacy techs are manually reconciling 200 insurance rejections a day on a Friday" is a gap. One is an opinion. The other is a workflow that's already costing someone real money.

CB Insights has been running their post-mortem analysis on failed startups for over a decade. The number one reason startups die, ahead of running out of cash, ahead of bad teams… is "no market need," at around 35–42% of failures depending on the year. The founders weren't lazy. They built well. They just built into a gap that wasn't there.

Step 2 — Look for the substitutes (not the competitors)

If your "gap" has no substitute, no spreadsheet, no consultant, no Zapier hack, no intern doing it manually, that's almost never a sign of a clear runway. It's usually a sign that the pain isn't sharp enough for anyone to bother.

When Airbnb launched, the substitute was Craigslist plus a phone call plus a leap of faith. When Notion launched, the substitute was four different SaaS tools held together with copy-paste. The substitute is the proof that the pain is real. The opportunity is making the substitute look ridiculous.

So when a founder tells me "there's no competition," I get nervous. When they tell me "people are currently solving this with three browser tabs, a WhatsApp group, and a Google Sheet that breaks every Monday", now we're talking.

Step 3 — Talk to buyers, not friends

Twenty conversations with people who would actually swipe a card beats two hundred "that's a great idea" comments from your network. Three questions do most of the work:

• How are you solving this today? (Tells you the substitute, and how painful it is.)
• What is it costing you — in hours, dollars, or lost revenue? (Tells you the size of the wallet.)
• What happens if you don't fix it in the next six months? (Tells you the urgency. If the answer is "nothing really," walk away.)

Notice what's missing: "Would you use this?" That question is worthless. Everyone says yes. Mom Test rules apply; ask about their past behavior, not their future intentions.

Step 4 — Try to sell it before you build it

This is the step that separates the founders who ship from the founders who re-pitch. Before you write a line of code, see if someone will commit. A pre-order. A signed LOI. A paid pilot at 30% of the eventual price. A deposit. Anything that converts an opinion into a transaction.

The reason this works is that money is the only honest signal. Compliments are free. LinkedIn likes are free. Even meetings are mostly free. A wire transfer is not. If you cannot get a single person to put even a small amount of skin in the game before the product exists, the gap probably isn't where you think it is — and you've just saved yourself nine months and a meaningful chunk of your runway by finding out cheaply.

Step 5 — Measure action, never compliments

When you're trying to validate, the only metrics that matter are the ones that cost the user something:

• Payment — they parted with money.
• Commitment — they put their name on something or blocked their calendar.
• Referrals — they sent another buyer your way unprompted.

Everything else is vanity. Page views, signups for a free waitlist, "looks cool!" replies — none of those have ever paid a salary.

Think about Facebook in its first month. It wasn't a clever product. It was a basic directory for Harvard students. But within 24 hours, more than half of Harvard's undergraduate population had signed up, and within a month other universities were asking to be added. That's the signal. Not the feature set, not the design, the unprompted pull from people who weren't asked.