Where Streaming Data Fails as a Concert Forecast

A stream costs the listener nothing at the margin. It requires no date, no travel, no companion, no childcare and no commitment made weeks in advance. A concert ticket requires all of those. Two measurements built on such different cost structures cannot be expected to track each other, so divergence is expected.

Four systematic reasons for divergence

• Cost asymmetry: streaming measures attention at zero marginal cost; attendance measures a discretionary purchase competing against every other use of the same money and evening.
• Geography: streams are national or global; a concert is a decision about one building on one date, and demand is a metropolitan quantity.
• Passive consumption: a large share of streaming volume comes from playlists and background listening in which the listener is not choosing the artist and may not be able to name them.
• Catalog depth: an audience returning to older recordings behaves differently from one following new releases, and it converts to attendance at a different rate.

The catalogue point is the most useful in practice. Luminate reports the share of U.S. on-demand audio consumption accounted for by deep catalog, and that share varies widely by genre — rock, for instance, is heavily weighted toward older recordings.¹ A high deep-catalog share identifies listening concentrated in older recordings. It does not establish ticket conversion. Genres weighted toward newer releases produce a different catalog-age profile, and event-level sales are required to compare concert demand.

Streaming share and live share are not the same share

The accompanying table sets a streaming share against a live-activity share for the same broad genre family, drawn from two different sources with two different definitions. It is deliberately not presented as a comparison of like quantities. One is an independent measurement of national listening volume; the other is a promoter's statement about the composition of its own promoted arena and stadium events.² Read together they establish only that the ordering produced by streaming does not reproduce the ordering produced by live activity — which is the argument. Read as a single metric they would be nonsense.

What would actually forecast demand

The inputs that predict a market's response to a date are local and behavioural: prior attendance in that metropolitan area, sell-through and yield at the last comparable date, the size of the artist's owned channels, whether the audience has previously paid for anything, and the venue inventory available at the right capacity band. Streaming volume is a weak national signal about a local decision. It is used because it is available, cheap and updated daily, which is a description of its convenience and not of its predictive validity.