Figure 3 — Per-impression quality by channel, 2015-2040

Trajectory

Each line shows how much real attention a single ad on each channel produces, indexed to that channel's own baseline (2015, or 2020 for TikTok, which had no US ad business in 2015). Above 1.0x means the channel's per-impression quality has improved. Below 1.0x means each ad is now worth a fraction of what it used to be. Meta has already crossed mid-market (2021) and is approaching bimodal (projected 2026). Linear TV crossed mid-market in 2022 and crosses bimodal around 2028. Out-of-home (OOH), direct mail, and host-read podcasts are the only channels holding their ground — and direct mail is the outlier that has actually improved, reflecting a shift to more targeted lists under rising postage.

Mid-market < 0.5x
Bimodal < 0.25x
Zero-utility < 0.1x
Channel baseline (1.0x)

Structural Resilience

OOH

1.21x

Direct Mail

1.99x

Podcasts HR

0.82x

Volume-Build Masking Decay

CTV

0.08x

Podcasts DAI

0.26x

YouTube

0.42x

Mature Decay

Search

0.11x

Meta

0.05x

LinkedIn

0.30x

Accelerated Collapse

TikTok

0.13x

Linear TV

0.07x

Display

0.08x

Each panel: focal channel in color, all other channels as faint grey backdrop. Lines after 2025 are projected at constant per-factor decay rate. Each channel is indexed to its own baseline year — 2015 for most; 2020 for TikTok, which had no US ad business in 2015 (its line starts at 2020). Newsletters and Retail Media excluded — both have unstable per-impression quality data due to near-zero baselines.

Quality Multiplier

The per-impression quality of a channel: Attention × Density × Legit Share. This is the metric the chart plots, indexed to each channel's own baseline year (1.0x). It deliberately excludes total impression volume — Meta showing 7x more ads doesn't mean each ad is 7x more valuable. The diminishing marginal returns (DMR) thesis is about the substrate getting worse per unit, not total supply.

Why log scale

Per-impression quality across channels spans an order of magnitude — Meta at 0.26x, Linear TV at 0.34x, Search at 0.42x, OOH at 1.08x. A linear axis would compress the bottom into a flat line. Log scale shows proportional change clearly: the distance from 0.5x to 1x is the same as 0.25x to 0.5x.

Mid-market threshold (0.5x)

When per-impression quality drops below half the channel's baseline, the median direct-response brand’s return on ad spend (ROAS) falls below the 3:1 lifetime-value-to-customer-acquisition-cost (LTV:CAC) rule that defined direct-to-consumer (DTC) unit economics from 2012-2020. Mid-market advertisers rationally exit at this point.

Bimodal threshold (0.25x)

When quality drops below a quarter of baseline, only the top 10 advertisers (Walmart, P&G) and scammers can still make the math work. The advertiser base bifurcates into "too big to fail" and "too illegal to leave." Meta is projected to cross this threshold in 2026; Linear TV in 2028.

Primary Sources

Meta, Alphabet 10-K filings (2015-2025)Adelaide Attention Units cross-channel benchmarksPew Research Center platform demographicsBrian Wieser, Madison and Wall — TV inventory declineOAAA annual OOH revenue reports