Loudness is not a measurement

The Argument · Ins7ghts Daily, Thursday 15 October 2026

For about fifteen years records got mastered louder every year, because loud sold. Engineers squashed the dynamics until the meters pinned and the quiet parts disappeared. Everyone was optimising one number, and that number stopped describing the music. Then the streaming platforms normalised playback volume and the whole arms race was worth nothing overnight.

The evidence. The Economist says an epidemic of unreliable numbers now threatens the boom: related-party revenue, usage counted three ways, booked capacity sold as proven demand. A wealth desk reading the central bank finds limited signs of broad transformation in output and labour data. Manulife put $1 billion of value on 2027. Fewer than one in four organisations can tie generative AI to bottom-line profit, and 59% do not measure their own data quality at all. Every one of those numbers is loud. Only one of them has a published scale behind it, and it is the one about not measuring.

The best case against: it is too early. Electrification took twenty years to show up in the productivity statistics, and demanding an audited metric from a two-year-old deployment would strangle work that is genuinely compounding. Teams measure what they can while the thing is still moving, and a crude number beats paralysis. That is a fair reading, and for an engineering team mid-build it is probably the right one.