Attention Stacking and the cycle nobody wants to break

There’s a song you used to love.

You heard it once and something happened. Then radio got hold of it, and it followed you into shops, into cabs, into the gym, into an advert for a bank. You still know every word. You can sing it- start to finish without thinking. And when it comes on now- you change the station. And you’re not entirely sure when the shift happened.

Now, let’s notice what’s going on in that gap. Your memory of the song went up. Your feeling about it went down. Both things happened at once, from the same cause, and only one of them is visible from the outside.

Now, hold that thought next to a mattress company.

Casper filed its books with a regulator in January 2020. Every marketing dashboard that the company ran was green. Awareness climbing, reach climbing, spend converting.

It went public the next month. It was priced about 30% lower than what private investors had valued it at a year before.

The dashboards were still green.

So here’s the question worth 1,400 words: How does a dashboard stay green all the way down?

Byron Sharp is right, and I want to say so before I argue with anything.

Let’s tackle the strongest objection first. This isn’t a straw man. I’d be ignoring decades of category data if I acted like it was.

Byron Sharp and the Ehrenberg-Bass Institute have spent years establishing that brands grow through broad, continuous reach to light buyers. Mental availability is the chance you pop into someone’s mind when they’re ready to buy. You build it by being present and keep it fresh by showing up again. Brands that stop showing up- fade. Loyalty-led strategies underperform penetration-led ones. This is the most cited empirical position in modern marketing and I think it’s substantially correct.

So reach works. Repetition works. None of what follows is an argument for spending less or hiding.

Where I’d push back is on what the reach is carrying. Sharp’s memory structures get built when distinctive assets attach to a brand that means something. Buy the reach with nothing underneath it and you get exposure that doesn’t encode, and that’s a different thing wearing the same clothes.

That different thing is attention stacking, and it’s what the rest of this is about.

What attention stacking is

Attention stacking is buying more visibility with nothing underneath to hold it. Rented reach layered on rented reach, no owned asset anywhere at the bottom of the pile.

It looks like more ads, then more posts, then another channel, then another format. Then somebody in a meeting says we should probably be on TikTok by now. Usually a person who will not be running the TikTok.

The logic holds up fine while you’re inside it. Enough people see you, some come to trust you, some of those buy. So you pay to be seen, the line goes up, and everyone in the room agrees to call that growth.

Here’s the part nobody says out loud about why attention stacking survives contact with a boardroom.

Attention has a dashboard, and interest doesn’t.

You can see attention rise in real time. Take a screenshot, add it to a slide, and highlight the number that changed since last Monday. It feels like proof. More usefully, it feels like work, which is a different and more addictive thing. Interest gives you nothing to screenshot while it’s happening. It stays quiet for months. In most companies that asymmetry decides the budget on its own, and the thing that compounds gets filed under next quarter, where it lives permanently, like a gym membership.

Back to the mattress company. Casper spent $422.8 million and the dashboards never blinked

Around 2018 and 2019 you couldn’t escape them. Podcasts, television, subway walls papered in branded puzzles. That wasn’t accidental. The whole model depended on buying attention and converting it before it evaporated.

They were good at the buying part. The S-1 shows $113.99 million in sales and marketing for the first nine months of 2019. From January 2016 to September 2019, the total was $422.8 million. That’s attention stacking with an auditor attached.

Every dollar of that had a dashboard behind it. Impressions, reach, cost per acquisition, conversion rate. All visible, all reportable, all pointing the right way.

Then page six of the filing tells you what it bought. More than 16% of direct-to-consumer customers had ever come back for anything. 14% returned within a year. 20% of customers in the first nine months of 2019 were repeat buyers.

Casper raises the obvious defence in the same breath, and it’s a fair one. The traditional replacement cycle is longer than the company has been around. So low repeat rates are almost guaranteed.

David Trainer read the same fact the other way in Forbes. A long replacement cycle hurts the economics instead of excusing them. You invest a lot to win somebody, and then you see minimal repeat business.

He puts the average mattress replacement at seven years. He cites the Better Sleep Council, then points out the Council was created by manufacturers who wanted that cycle shortened.

So let’s take the most generous number. The picture barely moves. $422.8 million against more than 1.4 million customers puts the implied blended acquisition cost near $302. An independent analysis of the filing estimated lifetime value around $428, working from the 16% repeat rate and a 50.7% gross margin, which is the best quarter in the document rather than the year.

A dollar in, a dollar forty out, for a business whose entire model was buying attention.

That’s not a growth engine.

It’s a treadmill with a payroll attached.

The thing your dashboard cannot see

So why did the reporting look fine? Not because anyone was lying.

But, because of the song. Remember the song?

The folk version of advertising wearout says repetition erodes response, full stop. What the research describes is stranger and more useful than that. Craig, Sternthal and Leavitt ran the 1976 experiments that named the effect. In their second experiment, they controlled for inattention and reactance. They found no wearout at all. The highest repetition level held brand name recall as well as or better than lower levels. Four years later Calder and Sternthal ran television commercials and found the opposite. Repetition produced wearout in how viewers evaluated both the ads and the products. And it was not prevented by strategies designed to hold attention.

Both are right, and I think the reconciliation is the whole point of attention stacking. The first study measured recall. The second measured evaluation. Repetition can keep building memory while steadily eroding liking.

Which is your song. Yours!

You know every word. You change the station.

Now let’s look at what a marketing dashboard tracks. Awareness. Recall. Reach. Impressions. Every one of those is the first study. None of them is the second. So the metric climbs while the thing that drives purchase quietly rots underneath it, and your reporting isn’t wrong, it’s measuring the half of the curve that’s still going up.

And that’s how a dashboard stays green all the way down. It isn’t broken and nobody is fiddling with it. It’s pointed at the wrong half.

Attention stacking is what we do when we can only see one half of that curve. You spend more, the visible number responds, and the invisible one keeps sliding.

You’re running faster on the treadmill and reading the speed as distance.

Let’s see it from a different lens. One guy with a camera, same years, no dashboard

Let’s set that against somebody who built the other kind of asset over exactly the same period.

A teenager this time- Marques Brownlee. He started reviewing tech on YouTube in 2008 in New Jersey. No budget, no growth hacks. He made honest reviews and then kept doing that for fifteen years. The key decision was refusing payment for reviews. It cost him real money in the short term. It built something unbuyable over the long one.

For years that choice would have looked terrible on a dashboard. Revenue below what it could have been, growth slower than it needed to be, nothing to screenshot on a Monday.

What it produced is a position where companies reportedly time product launches around his review schedule.

Axios put it about as bluntly as it can be put: if he doesn’t like your product, you’re in trouble.

Two objections here, and I find both of them good.

The first is survivorship. I’ve picked one creator out of millions, and I picked him on the outcome. Thousands of people made honest reviews for fifteen years and reached nobody.

The second is his distribution. It came from YouTube’s recommendation algorithm. By any definition I’ve offered here, that’s rented land.

Both objections land. Neither one touches the mechanism, though, and I’ll say why. Brownlee didn’t escape the platform. He compounded on top of it.

The reach was rented and could have been withdrawn at any point. What built up was an audience that would follow him anywhere. And a name people search for directly. The algorithm gave him the room. It never gave him the trust.

And here’s what matters for the dashboard question. Nothing he was building would have shown up on one. There was no metric that went up when somebody decided to trust him.

Casper’s dashboards were measuring recall while evaluation fell, which is attention stacking with better reporting. Brownlee had no dashboard at all, and evaluation was the only thing he was accumulating.

So why did the numbers stay green? I think it’s simpler than anyone wants it to be.

They were never measuring the thing that was failing. They were measuring the thing that was working.

And that thing kept working right up until the day the business ran out of people who felt anything about it.

And Casper? Well, Casper’s mattress was fine.

That was never the problem, and the dashboards were never going to tell anybody that.