The Unit Economics of Boredom

There is a version of the streaming business that exists only on spreadsheets.

In that version, you produce a certain number of hours of content at a certain cost per hour, you distribute those hours across a certain number of subscribers, and you optimize the ratio. Content cost over engaged households. Cost per minute delivered. Cost per minute watched, if you are being rigorous about it. The variables are legible. The finance team can model them. The board can argue about them. Whole decks — entire strategy offsites — are organized around the denominator in that ratio.

And that ratio is, I have come to believe, describing the wrong business.

The scarce input in streaming is not minutes produced. Minutes produced is extraordinarily cheap and getting cheaper. The scarce input is minutes earned — and by earned I mean attention that is arousal-weighted, in the sense that Daniel Kahneman used the word in his 1973 book Attention and Effort. Attention with the pupils a little dilated. Attention that expands the pool rather than draining it. The kind of attention that, when the credits roll, leaves a person wanting to come back tomorrow instead of wanting to cancel before the next billing cycle.

The services that win the next decade will not be the ones that get the cheapest per-hour cost of content. They will be the ones that learn to put a price on mild disappointment — the half-watched episode, the trailer that did not land, the Tuesday night where the user opened the app, scrolled for three minutes, and closed it again. That Tuesday night is where the business is actually being won or lost. And almost no P&L I have ever seen has a line for it.

The Denominator Problem

Every streaming finance team I have worked with or alongside treats content cost as a numerator problem. How much did we spend to produce the hour. How much did we license it for. How does that amortize over the window. Those are real questions and I do not mean to dismiss them.

But the denominator is where the business actually lives, and the denominator is almost always wrong.

The typical denominator is some version of minutes consumed. Total viewing hours. Average daily active minutes. Minutes per paid member. The dashboard metrics that sit in every streaming executive’s Monday morning email. And each of those metrics silently assumes that a minute is a minute is a minute — that the forty-fifth minute of a half-watched procedural carries the same weight as the final minute of a championship game, or the closing minute of a show that made someone text three friends immediately after.

They do not carry the same weight. They do not even carry weights in the same order of magnitude.

I wrote about this in the attention essay — the thing Kahneman observed, more than fifty years ago, was that attention is not a fixed budget. It is a pool whose size fluctuates with arousal. Ten minutes of genuinely riveted attention recruits more cognitive capacity than ten minutes of half-awake scrolling, and it lays down more memory, and it is worth more to an advertiser, and it earns more loyalty from the viewer, and it makes the next subscription decision easier instead of harder.

A minute of riveted attention is not 1.5x a minute of distracted attention. It is something more like 5x or 10x, depending on what you are measuring. And nobody’s P&L knows that.

What the Spreadsheet Misses

Consider two shows. Both cost the same to produce. Both land the same raw minutes-watched number in the first month. The finance team looks at them and concludes they are equivalent investments.

Show A is the one people put on while folding laundry. They hear dialogue. They catch the beats. They do not remember, a week later, a single line of it. They will renew their subscription, probably, but nothing about this show is the reason.

Show B is the one that made someone turn their phone over on the coffee table. That held them still for forty-two minutes. That they then talked about at work the next morning. That they mentioned to a friend at dinner that weekend. That, six weeks later, they remembered to check if a new season was coming.

The denominators are equal. The businesses are not equal. One of them is producing attention that compounds. The other is producing the television equivalent of ambient lighting — which has its uses, but should not be priced as if it is the thing carrying the service.

I do not think the industry has a vocabulary for this gap yet. Or rather, it has one — engagement, retention cohorts, the various loyalty scores — but the vocabulary lives in the marketing organization, and the content-investment decisions live in the finance organization, and they are looking at different spreadsheets.

The unit economics of boredom is the cost of that disconnect. Every show greenlit on the assumption that a minute is a minute is a show that slightly drains the pool. Every show greenlit on the assumption that arousal-weighted attention is the real input is a show that grows it.

The Price of Mild Disappointment

Here is the claim I have been circling toward: the most expensive thing on a streaming P&L is not the content line. It is not the marketing line. It is not the technology line.

It is the user who, on a given Tuesday, opened the app, scrolled for three minutes, did not find something they wanted, and closed it without watching anything.

The most expensive thing on a streaming P&L is the user who opened the app, scrolled for three minutes, did not find something they wanted, and closed it without watching anything.— the thesis

That user has not churned. Not yet. They have not generated a support ticket. They have not triggered any threshold on any dashboard. In most reporting frameworks, that Tuesday does not exist. The minutes consumed is zero, which means it does not show up in the denominator, which means it does not move the cost-per-minute number, which means the finance team does not have a reason to care.

But that Tuesday is the beginning of the churn. That Tuesday is where the mental accounting starts to shift — from “I use this service” to “I pay for this service and I do not really use it that much.” That Tuesday is the one that shows up, six months later, as a cancellation that the retention team will spend a quarter trying to model and will mostly fail to model, because the actual cause is buried in a behavior that was never measured in the first place.

Mild disappointment is expensive. It is the single most expensive thing in the business. And the reason it does not appear on any P&L is that it does not produce an immediate transaction. It produces a slow, invisible draining of the reservoir of goodwill that every subscription product runs on.

The services that win the next decade will be the ones that learn to put a number on that Tuesday. Not a precise number — nobody will ever have a perfect model of mild disappointment — but a directional number. A number that changes the order of the items on the product roadmap. A number that makes the content team and the UX team and the pricing team have a different conversation than they are currently having.

What a Correct Denominator Would Look Like

I do not think the fix is complicated, at the level of concept. It is just unfashionable.

A correct denominator would weight minutes by arousal. It would distinguish between the show that someone watched while half-asleep and the show that someone watched with their phone face-down. It would distinguish between the trailer that played because autoplay fired and the trailer that played because someone clicked on it. It would have a way of saying: this hour of content produced engaged attention, and that hour of content produced ambient attention, and those two hours are not the same input.

The industry has fragments of this. Completion rate is a fragment. Rewatch rate is a fragment. Share rate is a fragment. Whether somebody searched the title by name, versus arrived at it through a recommendation row, is a fragment. The time between sessions is a fragment.

Stitched together, those fragments would make a weighted-minutes metric that actually matched what the business is trying to do. But I have not seen anybody stitch them together publicly. No major streamer has disclosed a weighted-attention metric that goes beyond variations on completion rate, watch time, and the standard engagement scores that anchor every earnings-call talking point. The fragments sit in separate dashboards owned by separate teams who each have local incentives that do not compose.

The companies that stitch it together first — and then actually re-organize their greenlight decisions around the stitched number — will have an advantage that is hard to describe in a deck but will show up, eventually, as lower churn, higher ARPU, and a library that compounds in value instead of eroding.

The Netflix Example, Carefully

I want to use Netflix as an example here, and I want to be careful about what I am and am not claiming.

Netflix launched its ad-supported tier on November 3, 2022. It has adjusted pricing in October 2023, again in January 2025, and again in early 2026 — with the Basic plan quietly phased out for existing subscribers through 2024 in between. It has, along the way, become — on public reporting — the most profitable subscription streamer by a significant margin, with the best combination of engagement, retention, and unit economics in the category.

The standard story about Netflix’s advantage is that it had a head start. That it has the deepest library. That its recommendation system is better. All of that is true, and none of it, I think, is the main thing.

The main thing is that Netflix has spent a decade treating content as if the denominator is arousal-weighted attention and not raw minutes. They do not disclose the internal machinery. They do not publish a weighted-engagement score. But the decisions they make — the shows they cancel that were hitting raw-minute targets but not share-and-talk-about-it targets, the originals they re-up on evidence that goes beyond completion rate, the willingness to let library content carry the low-intensity side of the business while originals carry the high-intensity side — read like the decisions of a company that has quietly built a better denominator.

Other services are reading the same raw-minute numbers and drawing different conclusions, because the numbers are not actually the same numbers. A Netflix minute and a Peacock minute and a Paramount+ minute all get called “a minute” in the trade press. They are not the same unit. They carry different densities of earned attention, which means different retention impact, which means different unit economics, even at identical content spend.

The Paramount-Shaped Version of the Same Problem

Consider the ongoing narrative around Paramount, which was folded into the Skydance structure when that merger closed on August 7, 2025. The public story has been a DTC turnaround — some good quarters, some harder quarters, an ongoing question about whether the combined company has the scale to be a standalone streamer or whether it ends up as a content supplier into somebody else’s bundle.

The framework question underneath the turnaround narrative is the one this essay is about. You can get a DTC P&L to look better in the short term by two very different routes. You can optimize the numerator — cheaper content, less content, more library mix, fewer originals, tighter marketing — or you can optimize the denominator, which means producing fewer but more arousal-weighted hours and measuring them correctly.

Route one looks good for two quarters and then the churn math catches up, because the library is slowly draining the reservoir of earned attention. Route two looks worse for two quarters and then starts to compound, because the content is laying down the kind of memory and loyalty that reduces the cost of the next renewal decision.

Almost every DTC turnaround I have watched from the outside has been built on route one. And almost every one that has actually worked over a multi-year horizon has, under the hood, been built on route two. The first is visible on the P&L immediately. The second becomes visible on the P&L eventually, but it starts on a spreadsheet nobody is looking at yet.

The Ad Tier as a Forcing Function

There is one place where the industry is, almost by accident, being forced to price attention correctly. The ad tier.

When a service sells ad inventory, the advertiser does not care about minutes consumed in the abstract. They care about impressions that are actually impressions — that land on attention that is actually there, that produce brand recall that is actually measurable, that convert at rates that are actually defensible. Advertisers have been doing this kind of math, in television and print and digital, for fifty years. They are not perfect at it. But they are far ahead of the streaming P&L at it.

When a streaming service goes down the ad-tier road — as most of the category has, since 2022 — it imports, whether it realizes it or not, the advertiser’s weighted view of a minute. Live sports minutes are worth more than library drama minutes. Finale minutes are worth more than premiere minutes, sometimes. Morning minutes are worth less than evening minutes. Minutes on a television set in a living room are worth more than minutes on a phone on a commute. The CPMs are not uniform, and the non-uniformity is not a pricing quirk — it is an admission, by a market, that a minute is not a minute.

That admission is the most interesting thing happening in streaming finance right now. It is the beginning of an attention-weighted P&L, and it is arriving through the ad organization rather than the subscription organization, which is why most strategy conversations have not absorbed it yet.

What Happens When The Framework Lands

If the framework I am describing is right — and I think it is, and I think the next decade will prove it — then several things that currently look like mysteries will stop looking like mysteries.

Why does a service that has ten times the raw library of a competitor lose subscribers at a higher rate? Because the library is mostly ambient-attention content, and the competitor’s smaller library is mostly arousal-weighted content, and the attention-weighted denominator favors the competitor even though the minutes-watched dashboard favors the incumbent.

Why does a live-sports package that costs a fortune still pay back, when the per-minute cost looks indefensible on a spreadsheet? Because a minute of live sports is not a minute of library drama. Live is the highest-arousal content in the television ecosystem, which means it is the densest earned-attention inventory available, which means the real per-unit cost — properly weighted — is a fraction of what the naive math suggests.

Why do streaming services keep rediscovering that their biggest retention lever is a small number of shows that people actually love, and not a large number of shows that people will technically watch? Because earned attention compounds and ambient attention does not. The math has been telling them this for a decade. They have been looking at the wrong math.

Why does the “value” tier — the cheapest, most library-heavy tier — churn so much harder than the premium, originals-heavy tier, even after accounting for price sensitivity? Because the cheap tier is almost entirely a low-arousal product. It is optimized for the denominator the P&L measures and not for the denominator the user is actually experiencing. It wins on the spreadsheet and loses in the brain.

The Next Decade

I will not pretend this reframe is simple to operationalize. It requires content teams and data teams and product teams and finance teams to all agree that they have been looking at the wrong number, and then to agree on what the new number should be, and then to stay agreed on it for long enough for the first wave of contrary-looking quarterly results to come in and not cause a panic.

That is a lot of agreement in an industry that is not, historically, famous for it.

But the economics are going to force the reframe regardless. The services that continue to optimize for cheap minutes will keep producing cheap minutes, and their denominators will keep looking good, and their churn will keep being inexplicable, and their content libraries will keep failing to compound. The services that figure out how to price the cost of mild disappointment — the Tuesday night when the user opened the app and found nothing — will keep building libraries that get more valuable over time, not less.

This is not a content-quality argument. It is a measurement argument. You cannot run a business on the wrong denominator for more than about a decade before the business starts to disagree with the spreadsheet, and in this business we are now about a decade in.

The unit economics of boredom is a line item that does not exist on any P&L I have ever seen. It is the most important line item in the category.

Put it on the sheet.