Demand Signals: Prove People Want Your Solution Before Building

You prove demand with evidence people can't fake by being polite: money already committed, workarounds they built before you showed up, and behavior that repeats without a nudge from you. Ranked from strongest to weakest: willingness to pay, existing workarounds, repeat usage, referrals, waitlist conversion, and search volume.

Quick Answer: Rank a demand signal by what it costs the person giving it, not by how good it feels to receive. Money, pre-existing workarounds, and unprompted repeat behavior are close to unfakeable. Waitlist signups and search volume are worth collecting, but treat them as leading indicators, not proof, until something stronger backs them up.

What makes a demand signal real, not a vanity metric

A demand signal is real when it costs the other person something: money, time, reputation, or a habit they have to change. A vanity metric costs them nothing, which is exactly why it's cheap to collect and easy to over-trust.

"I'd definitely use that" costs a stranger nothing to say to a founder who is visibly excited about their idea. A $49 deposit costs something the moment the card is charged. That single difference, sacrifice versus sentiment, is the entire test. Before you log any signal as evidence, run it against three questions:

• Did it cost them something? Time spent, money handed over, or a workaround already abandoned in your favor all count. A verbal opinion does not.
• Did it happen before you asked for it? A workaround someone built on their own, a search they ran without your prompting, or a referral they made unprompted outdate your marketing and can't be an artifact of it.
• Would it survive them being annoyed with you? Compliments are calibrated to your feelings. A canceled Venmo request to a competitor, or a support ticket demanding a feature, is not.

Picture a founder, call her Mara, prototyping a scheduling and payments tool for solo driving instructors. Every signal in this article gets tested against the same business, so you can see how the ranking plays out on one consistent example instead of six disconnected anecdotes. This is one slice of a larger process; the complete guide to startup idea validation walks the full sequence from problem framing through pricing, of which demand signal reading is a single, important stage.

Signal 1: Willingness to pay

Willingness to pay is the strongest demand signal available, full stop, because it's the only one where the other person loses something the instant they hand it over.

It shows up in several forms, ranked by how much it costs the payer: a non-refundable pre-order, a refundable deposit, a paid pilot with a real invoice, and a signed letter of intent with financial terms attached. A verbal "I'd pay for that" is not on this list; it belongs in the compliment category from the previous section. Testing Business Ideas by David Bland and Alexander Osterwalder ranks these experiments among the highest-evidence tests precisely because they force a customer to act against their own short-term interest (parting with money) before you've built anything to justify it.

Running the test doesn't require a product. A one-page description of the outcome, a Stripe checkout link, and an honest note that says "we'll build this in six weeks, refund if we don't ship" is enough. Say Mara puts that page in front of the fifteen instructors she interviewed and asks for a refundable $49 deposit to reserve a spot in the first cohort. Four pay. Eleven decline, several with a version of "sounds great, let me know when it's live." That four-out-of-fifteen number is not a benchmark to hit; it's simply more honest than eleven verbal yeses, because those eleven people just told her, for free, that liking an idea and funding it are different acts.

Signal 2: Existing workarounds

Existing workarounds rank second because they predate you entirely; nobody built a duct-taped process to flatter a founder's pitch.

A workaround is any manual, awkward, or expensive process someone maintains because the real solution doesn't exist yet for them. It's a spreadsheet rebuilt from scratch by three people who've never met. It's a $19-a-month tool bent sideways to do a job it wasn't designed for. It's a Sunday night spent doing by hand what should take five minutes. The very inelegance of the workaround is the proof: nobody keeps investing in a bad process unless the problem underneath it is real and unsolved. This is the same behavioral evidence at the center of The Art of Problem Discovery: a problem worth building for almost always leaves a workaround-shaped fingerprint before you ever show up.

You find workarounds by watching, not asking directly. Ask "walk me through what you did the last time this came up" rather than "do you have a workaround for this," because the second question tips people into inventing an answer that sounds tidier than reality. In Mara's interviews, three instructors independently kept a paper day-planner for lessons synced against a separate string of Venmo payment requests, reconciled by memory at the end of each week. Nobody asked them to build that. That's the signal: a real cost already being paid to cope with a problem she hadn't sold them on yet.

Signal 3: Search volume

Search volume is the weakest signal on this list when read alone, because it tells you a question exists in the world, not who is asking it or whether they'll pay you to answer it.

Free tools do the job: a comparative Google Trends chart shows whether interest in a topic is rising, flat, or seasonal; a keyword research tool's autocomplete and "people also ask" suggestions show the exact phrasing people use when they're stuck. What search volume can't tell you is intent quality. Someone searching "how to schedule driving lessons" might be an instructor hunting for a paid tool, a hobbyist building a spreadsheet for free, or a student trying to book one lesson. The keyword doesn't distinguish them; only a conversation or a landing page does.

Treat search data as a targeting input, not a validation result. It's genuinely useful for choosing which problem language to test first, and for sizing whether a market question is common enough to be worth investigating at all. It becomes dangerous only when a rising search trend gets mistaken for proof that people will pay, which is a completely separate claim search volume was never built to answer.

Signal 4: Waitlist conversion

A waitlist signup is nearly free to give, so the signal worth reading isn't the headcount on the list. It's what fraction of that list actually follows through when you open the door.

An email address costs a visitor almost nothing: one click, no commitment, no cost if they never hear from you again. That's exactly why "500 people on the waitlist" means very little by itself. What matters is the drop-off between joining and acting: how many click through to book an onboarding call, complete a short intake form, or put down a deposit when you email "we're opening ten spots this week." This is precisely the mechanism smoke tests are built to isolate: a landing page or fake-door test that measures the click, not the compliment, and treats the signup itself as only the first, cheapest step in a longer funnel.

Say Mara's waitlist page collects 62 signups over a month. When she emails the list to open ten beta spots, nine reply, six actually schedule the onboarding call, and three complete it and start using the tool. The 62 was never the number that mattered. The six-out-of-sixty-two who acted when asked to do something real is the number that tells her whether the interest on the list converts into use.

Signal 5: Repeat usage

Once something exists to use, even a manual concierge version you run by hand, unprompted repeat usage is close to the hardest signal on this list to fake.

The distinction that matters is prompted versus unprompted. Usage you generate by emailing "hey, don't forget to check out the tool this week" is a marketing artifact, not a demand signal; you caused it. Usage that happens because someone opened the product on their own, three days after you last mentioned it, because they wanted to, is a different category of evidence entirely. Eric Ries frames this in The Lean Startup as the difference between vanity metrics and the actionable metrics that plug into a genuine build-measure-learn loop: a metric only earns trust once it's tied to a repeatable behavior you didn't have to manufacture.

You don't need a finished product to collect this. A concierge MVP, where Mara personally builds each instructor's weekly schedule and sends it as a formatted text message, still generates real usage data: does the instructor reply with changes unprompted? Do they forward it to a student without being asked? Do they message her asking where this week's schedule is when it's a day late? That last one, someone chasing you because your manual process became something they now depend on, is one of the strongest usage signals available before any code has shipped.

Signal 6: Referrals

An early referral is a smaller sample than you'd like, but it's a high-quality one, because nobody risks their own credibility recommending something to spare a founder's feelings.

A referral means an existing user told someone else about your product without you asking them to, and that person showed up because of it. Track this by asking new users directly how they heard about you, and by watching which links actually convert versus which channels generate clicks that go nowhere. Early on, the count will be small enough that a single enthusiastic user can look like a trend when it isn't; treat one referral as a data point, not a pattern, and wait for it to repeat across a few different users before you weight it heavily.

For Mara, the moment worth noticing isn't a number, it's the mechanism: one of her four paying instructors forwards the sign-up link to two other instructors at the same driving school, unprompted, in a group chat she isn't part of. She only learns about it because both of them mention it when they sign up. That's a referral doing what a referral is supposed to do: someone spending their own reputation on your behalf, which is not something people do for a product they're lukewarm on.

Demand signal strength vs. effort to measure

The six signals trade off strength as proof against how much work it takes to collect them, and neither end of that trade-off is automatically the right one to chase first.

Read the table as a sequencing guide, not a menu to pick one item from. Start with the cheap, low-effort signals (search volume, existing workarounds surfaced in interviews) to decide what's worth building a test for, then spend the extra effort on willingness to pay and repeat usage before you commit real build time. A founder who only ever collects the bottom two rows of this table can accumulate a folder full of encouraging data and still be one paid pilot away from finding out nobody will actually buy.

Key Takeaways

• Rank signals by what they cost the giver, not how they feel to receive. Money, pre-existing workarounds, and unprompted repeat usage are close to unfakeable; opinions and waitlist signups are cheap and easy to over-trust.
• Willingness to pay is the single strongest signal. A refundable deposit or a paid pilot forces a real decision that a verbal "I'd use that" never does.
• Workarounds predate your pitch. A spreadsheet, a manual process, or a bent-sideways tool that already exists is evidence the problem is real, independent of anything you've said.
• A waitlist number means little; the conversion rate off it means a lot. Measure what fraction of signups act when you actually ask them to do something.
• Prompted usage and unprompted usage are different metrics. Only count repeat behavior you didn't have to manufacture with a reminder.
• One referral is a data point, not a trend. Wait for the behavior to repeat across a few different users before you weight it heavily.
• Sequence cheap signals before expensive ones. Use search volume and workaround interviews to decide what's worth testing, then spend real effort proving willingness to pay and repeat usage before you build.

Frequently Asked Questions

What is a demand signal in startup validation?

A demand signal is any piece of evidence that costs the person giving it something, money, time, reputation, or a changed habit, rather than just an opinion. Money committed, pre-existing workarounds, and unprompted repeat usage are the strongest examples because none of them can be produced by someone simply being polite to a founder.

Is a waitlist a good demand signal for a startup?

A waitlist is a weak signal on its own because joining costs almost nothing. What matters is the conversion rate when you actually ask waitlist members to do something real, like book a call, complete an intake form, or pay a deposit. Track that follow-through rate, not the raw signup count.

How much money does someone need to pay before I trust the demand signal?

There's no universal dollar threshold; what matters is that the amount is real enough to force a genuine decision rather than an impulsive click. A refundable deposit, a paid pilot invoice, or a signed letter of intent with financial terms all clear that bar. A price so low it doesn't require thought doesn't tell you much.

What's the difference between demand and interest?

Interest is what someone says when a pitch sounds appealing and they want to be encouraging. Demand is what someone does when acting on that interest costs them something, paying, changing a habit, or telling someone else. You validate an idea by measuring demand, not by collecting interest, because interest predicts almost nothing about who will actually become a customer.