Trust What Customers Do, Not What They Say

One of the most common pieces of startup advice is to talk to potential customers before you build anything. It is good advice. But there is a problem with it: people are remarkably bad at predicting what they will actually do.

Ask someone whether your idea sounds useful and they might genuinely tell you it does. Show them a prototype and they might say they would love to try it. Describe a problem you want to solve and they may enthusiastically agree that the problem is frustrating.

Then you launch.

And they don’t buy it.

This doesn’t necessarily mean they lied to you. It means what customers say and what customers actually do are two different kinds of evidence.

For anyone trying to validate a business idea, understanding that distinction matters.

##“That sounds like a great idea”##

Imagine you tell ten people about a new product you’re thinking of building. Seven say it sounds interesting. Four say they could see themselves using it. Two tell you to let them know when it launches.

That feels like validation.

But what exactly have you validated?

Mostly, you’ve established that people don’t dislike the idea enough to tell you so. There are all sorts of reasons people give positive feedback. They want to be encouraging. They don’t want to criticise something you’re clearly enthusiastic about. The idea sounds useful in the abstract. They can imagine some hypothetical situation where they might use it. And saying: “Yes, I’d probably use something like that.” costs nothing.

Paying £10 for it is a completely different decision.

This is one reason startup idea validation can produce false confidence. If you ask people whether they like your proposed solution, you can easily end up measuring politeness, curiosity or hypothetical interest rather than genuine demand.

A problem isn’t necessarily an opportunity

There is another trap.

Customers complain.

A lot.

Search Reddit, forums, product reviews or social media and you can find complaints about almost anything. Existing software is too complicated. A service is too expensive. A particular process is annoying. A product is missing a feature.

Those complaints are valuable market signals, but they aren’t automatically evidence that somebody needs a new solution.

There is an enormous difference between: “This annoys me.” and “This annoys me enough that I will change what I currently do.”

That gap matters. Someone might complain every month about how cumbersome their accounting software is while continuing to use it for the next five years. They might dislike a feature of an existing product but consider switching to a competitor even more inconvenient. They might complain about the price of a service but reject every cheaper alternative because they trust the incumbent. Or they might have created a simple workaround that is good enough.

The problem exists. The frustration is genuine. The commercial opportunity may still be weak.

Look for behaviour, not just opinions

This is where behavioural evidence becomes much more interesting.

Instead of only asking whether people experience a problem, look at what they are already doing because of it.

Are they building spreadsheets to work around existing software?

Are they combining three different tools because no single product does what they need?

Are they paying freelancers to perform something manually?

Are they repeatedly switching between competing products?

Are they searching for alternatives?

Are businesses dedicating staff time to solving the problem internally?

Are customers paying significant amounts for an imperfect existing solution?

These signals have something that a survey response doesn’t: cost. The customer is already sacrificing money, time or convenience. That makes the problem considerably more credible.

A person saying “I’d love an easier way to do this” is interesting.

A person spending four hours every Friday doing it manually is much more interesting.

Workarounds can be powerful demand signals

One of the strongest signals of an unmet need is often an ugly workaround. People rarely enjoy creating complicated spreadsheets, manually copying information between systems or stitching several products together.

They do it because the outcome matters enough to justify the effort.

That doesn’t guarantee they will buy your product. Price, switching costs, trust and dozens of other factors still matter. But it tells you something important: the problem is already causing behaviour.

That is much stronger evidence than simply asking whether the problem exists.

The same applies to money.

If customers are already paying for imperfect alternatives, you don’t need to ask whether there is theoretical willingness to pay. There is at least some evidence of it already.

The more interesting questions become how much they’re paying, what they dislike about the existing solution and what would persuade them to switch.

Be careful how you conduct customer interviews

None of this means customer interviews are useless. They’re extremely valuable.

The problem is often the questions being asked.

Compare these:

“Would you use an app that automatically did this for you?”

and:

“How do you do this today?”

The first asks someone to predict their future behaviour. The second asks them to describe actual behaviour.

Likewise:

“Would you pay £20 a month for this?”

is generally weaker than:

“What are you currently spending to solve this problem?”

And:

“Do you find this frustrating?”

is weaker than:

“What happened the last time you encountered this problem?”

Good customer research doesn’t just ask people what they want; it investigates what they already do.

Actions have different strengths as market signals

Behavioural evidence isn’t binary either. A Google search is behaviour, but it requires very little commitment. Joining a waiting list is stronger. Booking a demonstration is stronger again. Entering payment details is stronger still. Actually paying is considerably stronger. Continuing to pay six months later may be stronger than all of them.

The closer the behaviour gets to the commercial action your business ultimately requires, the more seriously you should take it.

This is particularly important when evaluating early traction.

Five hundred people clicking an advert might tell you that your proposition attracts attention.

Fifty people creating accounts tells you something else.

Ten people paying tells you considerably more.

And eight of those ten remaining customers six months later tells you something different again.

They are all signals. They just shouldn’t carry equal weight.

This is why simple idea validation is difficult There is a temptation when assessing a startup idea to reduce everything to a simple question:

Is there demand for this?

Unfortunately, markets rarely give such clean answers. People can experience a genuine problem without wanting another product. They can dislike existing competitors without being willing to switch. They can express strong interest without being willing to pay. And they can sometimes demonstrate demand without ever explicitly saying they want a solution at all.

The job isn’t simply to collect positive signals. It’s to work out what those signals actually mean.

That requires looking across multiple forms of evidence: customer discussions, existing behaviour, workarounds, competitor adoption, reviews, search behaviour, pricing, willingness to pay and the alternatives customers already have.

No individual signal proves that a business will succeed. But some signals are much harder to fake than others.

Evidence before encouragement

When you’re excited about an idea, positive feedback feels good. Ten people telling you they’d use your product can feel like ten future customers.

They aren’t.

Treat those conversations as a starting point rather than proof. Ask what people currently do. Look for evidence that the problem changes their behaviour. Find out what they already spend. Understand their workarounds. Investigate why they haven’t adopted the alternatives that already exist.

Most importantly, distinguish between interest, frustration and actual demand.

That’s also the principle behind SignalCraft. Rather than trying to tell founders whether an idea is “good” or “bad”, SignalCraft looks for the market evidence surrounding it: customer behaviour, pain points, workarounds, competing solutions, willingness-to-pay signals and gaps in the available evidence.

Because the question isn’t whether people say your idea sounds good. It’s whether the evidence suggests they’ll actually do something about it.