Why this needs Qloo

The incumbent method, built properly

A booker deciding who plays a room today works from two things: popularity (whose numbers are big) and genre bucket (is this act in this venue's category). That is the whole heuristic, and we gave it the best form its logic allows — same candidate pool as Roomtone, same entity data, no taste-graph signal anywhere. Popularity only.

Qloo does return each entity's popularity percentile, so the baseline is cheap to compute honestly. What it has no access to is the one thing Qloo actually sells: who listens to what.

Every measured room, side by side

ρ is Spearman rank correlation between Roomtone's ordering and the incumbent method's. The last column is the popularity↔affinity correlation inside the room — how much of Qloo's raw affinity is really just fame. That is the bias the lift metric removes.

How to read the disagreement

Where ρ is high, popularity happens to be a good proxy — a room whose crowd is the mainstream of its genre. Where ρ collapses, the two methods are ranking different acts, and one of them is wrong. The useful case is a room where popularity is actively misleading: the acts that draw a Webster Hall crowd are not the acts with the biggest numbers.

What this does not prove. A low correlation is not a box-office result. It says the incumbent heuristic and the taste-graph read disagree about who belongs in a room — the disagreement itself is the evidence that Qloo is doing work, and the room read is checkable by anyone who knows the venue. We are not forecasting ticket sales anywhere on this site.