Venue Benchmark · venue-programming-v1
How scoring works
Three signals, each turned into a percentile against the active rooms in the same market, then blended into a 0–100 score. What the score is called depends on how many of the three we could actually measure.
Artist Pull
40%The measured reach of the artists who played the room in the last year — one observation per show, headliner first, weighted toward recent bookings. This is the signal nobody else can compute: it joins a venue to the scores of the acts it actually books. A single superstar night cannot carry a year of programming, and a resident act stops counting fully after a handful of appearances.
Activity Pulse
30%How often the room runs shows (45%), how many separate months it was active in (35%), and how recently (20%). A venue doing eight nights a month all year is a working room; six shows in one week is not the same thing. New rooms are measured against a 90-day minimum window so three shows in an opening week cannot read as twelve a month.
Crowd Draw
30%How full the room gets against its own capacity — so a packed 120-cap listening room is not punished for being smaller than a half-empty theatre. This needs door scans or validated tickets. RSVPs are intent, not attendance, and are never reported as a fill rate.
The name changes with the evidence
A score built from one signal is not the same claim as one built from three, so we do not give them the same name.
- Activity Rank — Activity Pulse only. How hard the room works, ranked against its market.
- Venue Signal — two signals measured.
- Programming Score — all three.
Activity Pulse is required for any of them. A roster with no recent shows tells you about a room's past, not whether it is still running.
Missing data is not a zero — and not a 50 either
If a room has no door data on Runiuni yet, we do not know how full it gets. Earlier drafts of this tool filled that gap with a neutral 50th percentile marked “assumed”. That reads as evidence when it is the absence of evidence — so we stopped. An unmeasured signal is excluded from the maths entirely and the remaining weights are renormalised: with Artist Pull and Activity Pulse measured, they carry 57% and 43% rather than 40% and 30%. The receipt shows the weight each signal actually carried, and every locked row tells you what would unlock it.
Peers, not the world
Rooms are compared against the active rooms in their own city — venues with at least one show in the last year. Dormant rooms are excluded from the comparison entirely rather than sitting at the bottom inflating everyone else. Where both rooms report a capacity we also weight by size similarity, and the cohort label says so; where capacity is unknown, it says only “active music rooms in <city>”, because claiming a size band we did not apply would be a lie about the comparison.
Precision we can support
“Top 12%” means nothing against nine rooms. Below a real benchmark we show no standing at all and say the benchmark is thin. Between roughly 15 and 40 comparable rooms we say “Top third”. Above that we round to the nearest 5%, and only past 100 do we quote an exact figure. We also require a room to have at least three shows on file before we rank it at all.
What we do not use
Third-party review ratings sit outside the score. They are useful context and we may show them beside it, but we do not build a ranked benchmark out of another platform's review content. Nothing in the score is generated by a language model — the arithmetic is deterministic and reproducible from the receipt.
The worldwide line is an estimate — and says so
Your market standing is measured. The worldwide line under it is the one modeled number on the page, built from three inputs we publish outright:
- 19,814 catalogued music venues — the MusicBrainz open census (CC0), counted July 2026. A hard lower bound.
- A coverage correction — of 8 real Jacksonville rooms, that census held 3, so the true worldwide count is roughly 19,814 ÷ 0.375 ≈ 50,000. One market, eight rooms: wide error bars, which is why we round to one digit.
- A modeled activity curve — shows-per-year across venues taken as log-normal, anchored at a median of ~48 shows a year (about one a week) and a top-1% of ~250 (nightly-cadence rooms). These anchors are assumptions, not citations, and the placement moves if they do — which is exactly why it renders as a wide band (“Top ~25%”), never a precise percentile.
The input on your side is the shows on file here, which undercounts most rooms' real calendars — so the copy says “by the N shows on file”, and adding your full schedule raises the line. The estimate is versioned (global-estimate-v1) and frozen like everything else, and it never replaces the measured market standing as the headline.
Versions are frozen
Every score records the formula that produced it. Changing a weight ships a new version rather than editing this one, so a score you saw last month stays reproducible.