Fascination Is Not Demand: Meta's $85B Metaverse Lesson

People can be fascinated by an idea without ever wanting it enough to change their behaviour. That distinction is easy to miss. It is also the most expensive thing you can miss.

McKinsey, or another MBB, reportedly charged around $10M to Meta.

Meta then burned roughly $85B on the metaverse.

Those figures are the ones in circulation, not audited numbers from my desk. Take them as reported. The ratio survives any correction you want to make to them: the fee is a rounding error, and the write-down is not.

At the height of the hype, McKinsey estimated the metaverse could become a $5 trillion market by 2030. Meta invested as if consumer adoption would naturally follow.

It didn't.

A $5 trillion number is not a measurement

Here is what nobody says out loud in the room where that slide gets presented. A market-size forecast for a category that does not exist yet is an assumption wearing a suit.

Somebody takes an estimate of how many people might be interested. Applies an adoption curve borrowed from smartphones, or from broadband, or from whatever last decade's success story was. Multiplies by a spend-per-user figure that also came from somewhere else. Compounds it forward eight years.

Nothing in that chain measured behaviour. Not one step of it tested whether a single human being would give up something they currently do in order to do this instead.

And then the number travels. It leaves the deck, enters the capital plan, and by the third quarter it is being cited as if somebody had gone out and observed it.

Wrong remains wrong. Everything built on top of it inherits the error, and it compounds quietly until an accounting line makes it visible. That is what a wrong market research insight actually costs, and it is never booked against the budget that produced it.

People found it fascinating. That was the trap.

Ask someone in 2021 what they think of a persistent virtual world where they work, socialise and own things. Watch what happens.

They lean in. It is a genuinely interesting idea. It is futuristic. It has that quality of sounding like the future arriving, and people enjoy that feeling. Ask them to rate their interest and you will get a good number back, and nobody is lying to you.

Now ask the same person to put on a headset for two hours a day instead of scrolling on their phone, to move their friendships there, to spend real money on a virtual jacket. Different question. Different machinery entirely.

Fascination is free. It costs the respondent nothing to feel it and nothing to report it. Demand always costs something — money, time, effort, or a habit somebody has to break. That is the whole difference, and conventional research is built in a way that cannot see it.

People can be fascinated by an idea without ever wanting it enough to change their behaviour. Billions get lost in the gap between what customers say and what actually drives them. That gap has a name: the say-do gap.

Why the questionnaire cannot catch it

Two reasons, and neither of them is sample size.

First, a survey never charges a price. Every question is answered in a world where the answer is free. There is no headset on the respondent's face, no €499 on the screen, no colleague looking at them oddly. Interest measured without a trade-off is not a weak signal about demand. It is a signal about something else.

The moment you force a real trade-off into the measurement, the ranking changes. Not shifts — changes. Things that scored high on interest fall through the floor, and something dull and practical climbs to the top. Every practitioner who has run both has seen this happen, and most of them still report the interest number, because it is the one that survived the last review.

Second, people answer with a constructed opinion. Buying decisions form intuitively and get justified afterwards. Decades of neuroscience say so, and the standard questionnaire ignores it by design. For a brand-new category there is no habit to read the answer off, so the respondent assembles one on the spot out of what they have read, what sounded plausible, and what they would like to be true about their own future self.

That construction is sincere. It is also disconnected from the mechanism that will govern their behaviour in three years. It is why customers can't tell you why they buy.

The research industry then hands the construction back with a confidence interval attached.

Attention was never the metric

The metaverse had extraordinary attention. Coverage everywhere, a company renamed after it, every consultancy publishing a point of view within about six weeks. If attention converted to demand, it would have worked.

It doesn't convert. Awareness of a category tells you people have heard of it. It tells you nothing about whether the category is bound to anything they want. Same reason awareness is a vanity metric at the brand level, one scale up.

And notice the second amplifier. The people who answer surveys about new technology with the most enthusiasm are the same people who adopt first. So the earliest usage numbers confirm the forecast. The curve looks validated at exactly the moment it is being validated by the only cohort that was ever going to behave that way. Ford ran into the identical mechanism with EVs, which cost them $19.5B and a question asked too late.

What you measure instead

Stop asking people for reasons. Reaction-time based implicit measurement gets at what a category is actually linked to in someone's head before they have time to assemble an explanation for you. Fast responses signal intuitive certainty. Slow ones signal deliberation, and purchases do not run on deliberation. That is the Frame, Measure, Infer sequence.

For a virtual world that means you stop measuring "interest in the metaverse" and start measuring what the thing is bound to. Escape or isolation. Presence or absurdity. Belonging or performance. Those bindings predict what happens when the headset is in the box on the shelf. The interest score predicts what happens in the next survey.

Then separate the drivers from the passengers. In any real market, curiosity, tech affinity, income, age and social circle all move together, and standard statistical models get biased by that. You end up with a clean-looking driver ranking that is wrong. Causal AI is what tells you which factor causes adoption and which merely travels alongside it.

The distinction is not academic. If the real bottleneck is that nobody wants to be seen wearing the thing, no amount of resolution improvement fixes it. If it is that the virtual world is boring after week two, content spending is the answer and hardware spending is money set on fire. Both hypotheses fit the same correlation data. Only one of them was worth $85B.

This is what testing whether a product will sell before launch is supposed to do, and it is a different exercise from a concept test with a top-two-box score stapled to it.

The $10M was never the risk

I don't think the consultancy is the villain here, and I say that as someone who spends a lot of time arguing with consultancies.

Somebody asked for a market-size number, and a market-size number was delivered. That is the transaction. The failure is that a forecast built on interest was allowed to function as evidence of demand, and nobody in the chain was responsible for saying out loud what it actually was.

Roughly 5% of brands grow. Roughly 5% of new products survive. Roughly 5% of campaigns work. The same 5% keeps showing up, and what they share is not a better forecast. They found out what actually drives behaviour before they committed the capital. Everyone else decorates a guess. The full argument is in THE TOP 5%.

To change this is my very mission.

So: is your next big bet backed by demand, or by fascination? And do you have a way to tell the difference?

Fascination vs demand: frequently asked questions

What is the difference between fascination and demand?

Fascination is interest that costs nothing. A person can find an idea futuristic, clever and exciting while never giving up money, time or an existing habit for it. Demand only exists where somebody trades something away to get the thing. Every metric that captures attention, excitement or stated appeal is measuring the first and being read as the second.

Why do market-size forecasts fail so badly for new categories?

Because a market-size number is an extrapolation of an assumption, not a measurement of behaviour. It usually starts from stated interest, applies an adoption curve borrowed from a different category, and compounds both across several years. Nothing in that chain tests whether anyone will change what they currently do. The number then travels into a capital plan as if it were evidence.

Can you measure whether people will actually change their behaviour?

Yes, but not by asking. Implicit, reaction-time based measurement captures what a category is linked to before a respondent builds an explanation, and it separates warm interest from the associations that actually move behaviour. You then need causal modelling to tell which of those associations drives adoption and which merely travels alongside it.

What should a board ask before committing capital to a new category?

Ask what customers would have to give up to adopt, and whether anyone has measured their willingness to give it up. Ask whether the demand number comes from observed trade-offs or from stated interest. And ask what result would have stopped the investment. If no result could have stopped it, the research was decoration.

Dr. Frank Buckler is the founder of SUPRA and a pioneer in Causal AI for marketing. He has applied implicit research methods across FMCG, pharma, financial services, and insurance for over 25 years. His current book is THE TOP 5%.