Demand research: what do people actually want?

The question in the title is the wrong question. Let me fix it before answering it.

"What do people want?" is free to answer. Wanting costs nothing, so everyone wants generously. The thing that predicts a sale is different: the money people are already spending on a bad solution.

I take that distinction seriously for a measured reason, not a hunch.

The measured problem with asking "would you buy this?"

The gap between intention and behaviour has been measured in social psychology for forty years. A review pulling together ten separate meta-analyses found that across 422 studies, intentions accounted for about 28 percent of the variance in behaviour. In health behaviours, people convert good intentions into action roughly 53 percent of the time. A coin flip.

The experimental side is sharper. Interventions that succeed in shifting someone's intention by a medium-to-large amount move their actual behaviour by only a small-to-medium amount. Raising intent does not raise behaviour proportionally.

Translate that into product terms. You run a survey asking "would you use a tool like this?" and 70 percent say yes. That 70 percent is a stated intention. And your survey measures something weaker than the intentions in those studies: the person is being polite, they are saying yes to something free, and nobody asked them for a card number.

These studies are about exercise and health, not shopping. They do not say surveys lie. What they say is narrower and more useful: stated intention is a weak predictor of what someone actually does. And a survey response is weaker than intention.

The Intention–Behavior Gap, Sheeran & Webb

What I look for instead: evidence of money already spent

A complaint is free. A payment is not. In demand research the thing I am hunting for is proof that someone has already paid a price to solve this problem.

The price doesn't have to be money. Five signals:

• A sentence with a number in it. "I pay 30 a month for this and it still doesn't do X." That person is a customer, just an unhappy one.
• A cancelled subscription. Someone writing what they dropped and what they moved to is handing you both the budget and the gap.
• A hired freelancer. If someone is paying a human for a repeating task, the task happens often enough to become a product.
• A hand-built workaround. Copy-paste between three tabs, a spreadsheet they maintain themselves, a Monday morning ritual. Time is a form of payment and usually the most expensive one.
• A three-star review on a paid product. One star is anger, five stars is politeness. Three stars is someone who paid and can describe exactly what is missing.

All five share one property: the person had already made a buying decision before they started talking. You are reading behaviour that happened, not intention that might.

Where to look

Reddit's own search is weak. I'm writing that as something you can check in ten seconds rather than something you should take from me: run the same phrase inside Reddit and then as a Google site:reddit.com query, and compare what comes back. I do my searching through Google.

Freelance marketplaces are the least used and most honest source. The job posts on Upwork or Fiverr are a live list of things people are paying for right now. Not a survey, a payment. If the same job keeps getting posted, there is a product sitting in it.

Then: reviews of competing paid products, their "most requested feature" threads, and their changelogs. A competitor's three-star reviews are your roadmap.

Facebook groups and Discord servers hold most of the traffic that used to live on forums. They search badly, which is exactly why the signal in them is still cheap.

Three things that look like demand and aren't

Upvotes. A thousand upvotes on a complaint is not a thousand people who will pay. Agreeing is free, the same way wanting is free.

"I would 100% buy this." This is precisely what the gap above measures. One comment is one person's intention, and intention is 28 percent of behaviour.

Your own excitement. The most dangerous of the three, because it selects the other two. Start the search holding an idea and you will find the evidence that supports it. I try to start with "what did these people last pay for" instead of with an idea. I don't always manage it.

Where this method stops

It only finds problems people have articulated. The ones they have learned to live with, the ones they no longer bother complaining about, appear in no search. Some of the largest opportunities sit there, and this method is blind to them.

Second, everything you find is past tense. It shows what people paid for last year. It does not tell you whether the market is growing.

Third, and this one bit me: searching in one language shows you one world. I write for a Turkish-speaking market, and Turkish forum traffic is thin. Searching in Turkish, finding little, and concluding "no demand" is easy, when what you actually have is no data. I search both languages and look at the difference before deciding anything.

What I don't know

I have not carried this method through to a sale. There is no revenue I can point at, so I can't tell you it works. What I've written here is the set of criteria I use when I do my own research, and the evidence sitting behind those criteria.

There is a second thing I don't know: how much of that spending evidence actually converts into a product. Plenty of freelance work may never productise, because every instance is a little different. I don't have a test that separates the two yet.