The 141-minute marketplace

Doom scrolling is not an accident of design. It is a manufacturing process, and what it manufactures is advertising inventory.

The unit of attention has collapsed. Mobile users give a single piece of feed content roughly 1.7 seconds before scrolling on. Average Facebook session length fell from 2.7 minutes in 2013 to about 54 seconds today. The average user now scrolls the equivalent of 300 feet of content daily and encounters over 5,000 pieces of content, up from 1,400 in 2012.

Inventory quality degrades as volume rises. More impressions per session means each carries less attention. This is why CPMs can look stable or even fall while your cost per acquired customer climbs — you are buying more, cheaper, thinner impressions. Social CPM peaked at $5.48 in Q4 2024 before easing to $4.37 the following quarter. Your CAC did not follow it down.

The result is a scarcity market disguised as an abundance market. Dashboards show unlimited reach at low CPMs. The P&L shows acquisition costs rising 8–16% a year. Both are true. The abundance is in impressions; the scarcity is in attention — and attention is what converts.

Before examining what brands pay, it is worth establishing what is being bought. The numbers are startling in both directions.

The cost side — what users give up

Three profiles, valuing each person's social time at their own earning rate. The teenager loses the most time. The senior professional loses the most money.

The price side — what that attention sells for

Meta's ARPU hit $57.03 globally in 2025 — the highest it has ever recorded, on revenue of $200.1 billion. US & Canada ARPU sits near $233, roughly 16x the Rest of World figure of $14.

Cross-checking against the total market: US social ad spend is projected at $126 billion in 2026 across roughly 250 million US social users — about $500 per user per year. Divide by ~900 annual hours and both paths converge on the same number.

The asymmetry

Set what attention costs the person against what it sells for, and the gap is not marginal. A senior professional's attention is sold for about 0.8% of what their own time is worth: $52,560 of foregone value annually, monetised at $233.

Most marketing teams believe they are buying media. They are not. They are entering a real-time prediction market, thousands of times per second, on terms the platform sets.

Every time a user opens Facebook, Instagram or TikTok, an auction runs for each available placement. Tens of millions of ads are eligible; the field narrows to a few hundred before it resolves. And crucially, the highest bidder does not win.

Total value = Bid × Estimated Action Rate × Ad Quality

Only the first term is yours. Estimated Action Rate is the platform's prediction that this specific user takes your specific conversion event. Ad Quality is its judgment of your creative. Two advertisers bidding an identical $2 CPC do not get identical outcomes — the one with the higher predicted action rate wins, and the loser must bid $4 or $5 to buy the same impression.

Three consequences, each with a line-item cost

You pay a tax on weak creative. Quality score functions as a direct cost multiplier — poor creative does not merely underperform, it raises the clearing price on every impression you do win. Motion's analysis of over 550,000 ads found only 5–8% become genuine winners, while roughly half never receive meaningful spend.

You may be bidding against yourself. Audience overlap — multiple ad sets, or multiple brands in one portfolio, targeting the same users — creates internal competition inside your own account. You inflate your own clearing price and pay for the privilege. For multi-brand D2C groups this is frequently the single largest source of invisible waste, and no platform dashboard will surface it, because from the platform's perspective nothing is wrong.

The platform's objective is not your objective. Its auction maximises revenue per impression while keeping users scrolling. Yours is contribution margin. These align often enough to be confusing and diverge often enough to be expensive.

Beneath the channel-level view sits the allocation finance rarely sees, and that determines profitability.

By channel. At the $500K–$3M stage, most D2C brands put 40–50% into Meta and 20–30% into Google. Meta buys discovery; Google harvests demand that already exists. The mature portfolio pattern runs roughly 70% proven channels, 20% growth, 10% experiments.

By funnel stage. A common working split at scale is 30% top-of-funnel, 35% mid, 35% bottom. Over-weight the bottom and you harvest a shrinking retargeting pool until frequency caps bite and CPMs climb. Under-weight it and you pay to build demand a competitor converts.

By SKU — where it usually breaks. Hero SKUs attract budget because they convert. But conversion rate is not contribution margin. One DTC apparel brand had $380K of $1.5M in working capital locked in a single hero SKU's safety stock and receivables — 25% of available capital in one product that may not have been generating free cash flow at all. The same discipline applies to media: a SKU with a 40% return rate is not a marketing problem, it is a product problem being funded by the marketing budget.

By creative volume. Creative production is now a real line item — roughly 25% creative to 75% media at scale — because in an auction that prices quality, creative velocity is media efficiency.

Audience composition varies far more across platforms than most media plans assume, and mis-mapping it is the fastest way to overpay. The global social user base runs roughly 54.6% male to 45.4% female — but that average conceals almost everything that matters.

Age tells a second story. The 25–34 bracket dominates every single platform, so brands targeting it have genuine choice — and should optimise on cost, not habit. Outside that bracket the differences become decisive: among US adults 18–29, YouTube reaches 95%, Instagram 80%, Facebook 68%, TikTok 63%.

The structural warning is simple: the audience you bought two years ago is not the audience you are buying today.

Most D2C scorecards measure the wrong layer. Three tiers matter, and they are not interchangeable.

Average ROAS tells you what a channel has returned. Marginal ROAS tells you what the next dollar returns. They are different numbers, and only the second should inform a budget decision. A brand can be your best performer on average and still deserve less budget — because "best" is an average, while budget decisions happen at the margin.

Fospha's analysis found 83% of revenue impact goes untracked when D2C brands rely on last-click or siloed platform reporting. You cannot allocate against a number that misses five-sixths of the effect.

Part Two established that attention costs the user a fortune and sells for 55 cents an hour. So the advertiser holds the cheap end of the trade. What does that cheap input actually return?

Take $1 of D2C ad spend at a realistic 3.0x ROAS — paid social typically runs 2.5x to 4x, and Facebook ROAS reached 2.79 during peak Q4 2025.

Now apply the waste. Roughly 30.6% of digital ad spend is lost to mistargeted audiences, low-quality placements and tracking errors, per audits of enterprise accounts. The average Google Ads account loses another 12% to undetected anomalies, and some $63 billion globally is absorbed by invalid traffic.

If only $0.694 of every dollar is doing work, the productive portion is running at 4.32x, not 3.0x. Recover just half that waste and the picture changes materially.

Roughly 64:1 against the person doing the scrolling. The attention economy is not efficient at converting human time into economic value — it is merely very good at capturing it cheaply. For advertisers, the implication is uncomfortable and clarifying at once: you are buying the cheapest input in the entire chain, and throwing away a third of it.

The allocation question — how much should each brand, each SKU, each channel, each region get next quarter — is currently answered by negotiation. Last year's number, adjusted in the room by whoever argues best. Ask what happens if you cut a line 15% and the honest answer is that nobody knows until the quarter closes and the damage is visible.

That is not a failure of anyone's judgment. It is a problem of scale. A portfolio with six brands across search, social, retail media, influencer and four geographies generates hundreds of interacting response curves under budget and capacity constraints. No human solves that in a planning meeting. People are being asked to perform arithmetic that is not humanly doable — then held accountable when the guess is wrong.

Operations solved this class of problem decades ago. Production scheduling, fleet routing and inventory replenishment are all constrained optimisation, and no serious manufacturer runs a line on gut feel. Marketing is the last major cost centre still allocated by argument.

The bottom line

The attention market will not loosen. 141 minutes is the ceiling. The money chasing those minutes compounds at double digits. The platforms have every incentive to keep manufacturing thinner inventory to sell into it.

You cannot outspend that. You can only out-allocate it.

You are paying 55 cents an hour for the scarcest resource in your business — and throwing away a third of it.

Halving your waste is worth more than doubling your budget, because doubling your budget does not create a single new minute of human attention.

Your competitors can outbid you. Nobody can outbid the clock.

The brands that win the next three years will stop treating budget allocation as a quarterly negotiation and start treating it as what it is: a solvable optimisation problem with an objective, decision variables and constraints — and an answer that can be calculated, forecast, and checked against reality when the quarter closes.