The Linkedin Demand Test

I originally wrote this post as a lengthy GTM playbook: ICPs, wedges, sequencing, channels, sales motions, the works. On reflection, I’ve cut almost all of it. At the earliest stages, none of it matters.

Customer discovery does not start with interviews. It starts with seeing whether the right person will stop, reply, ask, click, forward, or take the call.

AI has made software cheap to build. A small team can prototype, ship, and iterate in days. So the bottleneck has moved: when everyone can build the app, the hard part is knowing what is worth building, and the fastest way to find out is to put your idea in front of the exact right person and watch what they do.

That is the test. Not a survey. Not a meeting. A message.

Bring the insight. Let the market vote.

Start with a strong view of the problem and the solution. I agree with Keith Rabois here: the founder’s job is to bring the insight, not to crowdsource it.

Customer feedback can mislead. People rationalize. They are polite. They describe symptoms instead of causes. Ask them what to build and you will often get answers that sound useful and point the wrong way.

So do not ask the market for strategy. Ask it for behavior. Hold your perspective strongly, but treat it as a testable hypothesis. You are not testing whether your story sounds clever. You are testing whether you have a painkiller, not a vitamin: a problem urgent enough that someone will spend time, attention, or money to make it go away.

Action: Write your wedge as one sentence a stranger could react to: [These people] have [this urgent problem], and [this] makes it stop. If you cannot, you are not ready to test.

Separate signal from noise

Here is the rule that governs everything below: a polite reply is not demand. A like is not demand. “Looks great, keep me posted” is not demand.

Demand is someone moving toward you because the pain is real and you might make it stop. Everything people say is noise. What they do is signal.

A rough bar: send 20–30 well-targeted messages. If almost no one engages, do not blame the copy. Your thesis or your list is wrong. If even a handful reply with questions, ask for more, or forward you on, you have found a thread worth pulling.

Action: Keep a simple sheet. Log every message and score only behavior: accepted, replied, asked, forwarded, met, returned, paid. Those numbers become proof for investors later.

Find the exact people

Specificity is everything. “Sales teams” is not a market. “Healthcare” is not a market. “AI for operations” is not a market.

If you do not know precisely who should care, you cannot tell whether your thesis is wrong or you are just talking to the wrong people.

I use Happenstance.ai to search my network in plain English: “operators running concierge or DPC practices,” “medspa founders who have cracked retention,” “investors focused on vertical AI.” It surfaces specific people who match, plus the warm path to them if there is one. That turns an ICP hypothesis into a real list to test against.

Action: Build a list of 30 named people who should have your problem, not 30 companies. If you cannot name 30, your ICP is still too vague to test.

The Linkedin Demand Test

Before building more product, test whether your message earns attention.

Just 250–300 character LinkedIn messages, where every word has a job.

The goal is not to pitch. It is to see whether a specific problem, positioning, or ICP is compelling enough that busy people will engage.

You are not trying to cram everything in. You are hitting three signals: I chose you intentionally, I am worth answering, and you may get something useful from the conversation.

That middle signal is where most cold outreach dies. A stranger has no reason to be trusted, so the fastest fix is borrowed proof: a mutual connection, a shared room (an accelerator, a school, a prior employer), a recognizable backer, or a real result. A mutual connection is strongest. It is the same warm path worth chasing above. One anchor, relevant to them, stated plainly. More than one reads as insecurity, and proof you have to inflate is worse than none.

Three ways to open

The formula stays fixed. What moves is the opening and the proof you borrow. Three that work, each carrying a different kind of validation:

1. Warm path: a mutual connection. The strongest proof there is. Someone they trust vouches, even loosely, and the message is no longer cold.

Hi Nick. [Sarah Chen] pointed me to you. I’ve been studying clinics that deliver great care beyond the visit, and Commons keeps coming up. I advise a company in this space and would value 15 minutes to compare notes. Happy to share what we’re seeing.

2. Credential + specific praise. A relevant credential, paired with a real observation so it reads as context, not a flex.

Hi Lenore. SkinSpirit seems to have cracked retention in a way most medspas haven’t. I spent [years running growth at Ro] and now advise a company on exactly that. Would love 15 minutes to compare notes. Happy to share the patterns we’re seeing.

3. Give first: when you have no proof to borrow. No mutual connection, no credential yet? Lead with value. The gift is the credibility: effort and insight a random seller would never bother with.

Hi Clint. I’m building a tool for AI-enabled care teams, and I put together a short read on how the best ones handle intake. A few patterns I rarely see discussed. Would love to send it your way. If it’s useful, could I get 5 minutes of your time?

Different doors, same house: each signals I chose you, I’m worth answering, and there’s something in this for you. The bracketed bits only work if they are true. Borrowed proof you have to fake is the one thing worse than no proof at all.

Action: Pick your real proof anchor before you write a word: the name you can drop, the room you shared, the result you can point to. If you do not have one yet, get the warm intro first; it is worth more than any line of copy.

Use AI to aim, not to hallucinate demand

AI is useful here, but not because it can tell you whether the market cares.

It can help you find the right people. It can reframe the same wedge for different audiences. It can personalize outreach and surface patterns across replies and calls.

What it cannot do is manufacture the only signal that matters: whether the right person actually gave you their attention. Use it to aim, then track what happened. Do not let it summarize weak signal into false confidence.

Action: Let AI draft and sort; never let it grade. A reply is a reply because a human sent it, not because a model called it promising.

Then talk. Listen for behavior.

Once someone engages, the conversation gets real. You are no longer asking abstract questions; you are talking about something they already reacted to.

Ask how they handle the problem today. What breaks? What have they tried? Who owns it? What happens if they do nothing? What would need to be true for them to change? Listen for repeated pain from a narrow set of people. The question is never “did this seem like a good idea?” It is “what did they actually do?”

The same test applies to fundraising. For customers, the question is whether the pain is real. For investors, it is whether the story travels: whether one credible person will spend social capital to pass it to another. A warm intro is that vote.

Action: After each call, write one line: what did they do next? If the answer is “nothing,” the pain is not real yet, no matter how good the conversation felt.

Build around the reaction that repeats

Only then do you build. The MVP should prove one core behavior: upload the list, run the search, send the care plan, approve the workflow, invite the teammate. Ugly is fine. Manual is fine. Narrow is good.

Early traction is not a polished product. It is the right customer repeating the core behavior because it solves a real problem.

Action: Name your one core action and instrument it. If your best customers are not repeating it, more features will not save you. The value prop is off.

Why it matters

Building always feels like progress, but early on it can be a trap. When AI makes software cheap to create, the scarce thing is not product. It is judgment. The edge is knowing which idea, message, and customer can earn attention before you spend weeks polishing something nobody asked for.