Figure 4 — Why each channel is degrading

Quality Decomposition

The diminishing marginal returns (DMR) thesis says per-impression quality is collapsing on most channels — and not for the same reason on any two of them. This chart breaks down each channel's quality change into its three components: how much human attention each ad gets, what fraction of the audience can actually afford the product, and how much of the inventory is real vs bots and scams. The archetype A channels (out-of-home or OOH, direct mail, host-read podcasts, newsletters) are the exceptions; everywhere else, pick a channel to see which factor is dragging it down.

A — Attention
D — Density
L — Legit Share
Click a channel for the full breakdown

Meta

Mature Decay

Net quality

-74%

Volume

+600%

Attention
-42%

Active human attention per ad

0.55 → 0.32

Density
-42%

Share of audience in upper-income tier

1.00 → 0.58

Legit Share
-24%

(1 − bot fraud) × (1 − scam-ad share)

0.95 → 0.72

Density is the dominant drag on Meta — down 42% over the decade.

Reference channel for the broader DMR thesis. Per-impression US blended CPM $5.50 → $8 → $11 over the decade. 2015 triangulated from Nanigans Q4 2015 benchmark coverage (eMarketer reporting Nanigans Q4 ecommerce CPM $7.35 +49% YoY, blended full-year US retail and gaming inventory landing around $5-6). 2020 triangulated from Tinuiti Q4 2020 benchmark + Meta earnings price-per-ad YoY (gated vendor data; enterprise midpoint ~$8). 2025 anchored on Varos Q1 2025 disclosure ($10.88 blended Meta CPM; $10.48/$10.35/$11.86 across Jan/Feb/Mar). Quality multiplier A × D × L fell 74% (0.523 → 0.134): Nelson-Field measures 2s active attention on the feed; Pew documents upper-income daily use inverting 78% → 45%; Reuters (Nov 2025) surfaced Meta's internal projection that ~10% of 2024 ad revenue came from scam ads. Nominal ARPU grew 6.5× ($41.65 → $270) while per-impression CPM grew ~2×; the rest of ARPU growth came from 7× more impressions per user. The "Running to Stand Still" per-user framework lives on its own page and uses Meta's audited ARPU directly — the CPM here is the apples-to-apples per-impression number for cross-channel comparison.

Each panel shows the percent change from 2015 to 2025 in three quality factors. The headline number (top right) is the net per-impression quality change (Attention × Density × Legit Share). The “Volume” stat is the impression growth that disguises the per-unit decay — most decay channels show massive volume growth alongside quality collapse.

Attention

How many seconds of active human attention each ad gets, anchored to OOH = 1.00 using Nelson-Field biometric data, Adelaide Attention Units (AUs), and Lumen eye-tracking. Premium video gets ~13.5s, Facebook feed gets ~2s, TikTok ~3.5s, OOH ~12s.

Density

The fraction of the channel's audience that's in the high-value income tier where most consumer spending happens. Pew documented Facebook's daily upper-income usage falling from 78% in 2015 to 45% in 2025 — the platform inverted from "Harvard exclusive" to downmarket.

Legit Share

The fraction of impressions that are real humans seeing real ads — calculated as (1 minus bot fraud) times (1 minus scam-ad share). fraud0 measured 21% bot share on paid social. Reuters revealed Meta internal docs projecting ~10% of 2024 ad revenue came from scam ads.

What's not in this chart

Impressions (the volume term) are deliberately excluded. They live in the stat box below the chart instead, because most degrading channels show massive impression growth — Meta has 7x more impressions in 2025 than 2015. Volume disguises decay; the volume story belongs on Running to Stand Still, not here. This chart asks: "of each impression you actually buy, how much real attention does it deliver?"

Primary Sources

Nelson-Field biometric attention data (Amplified Intelligence)Pew Research Center platform demographicsfraud0 / DoubleVerify / Adalytics invalid traffic dataReuters — Meta scam-ad internal documents