Money & sales
Completed nights vs. usual
A venue’s most recent completed night against the median of its own past nights on the same weekday (8-week window, four-night floor). Neutral direction only; a venue is never compared to another venue.
What it is
Completed nights vs. usual is the percentage difference between a venue's most recent completed night and the median of that same venue's own past nights on the same weekday. A Friday is read against recent Fridays, a Tuesday against recent Tuesdays — an eight-week, same-weekday window, with a four-night floor before any comparison renders.
It is deliberately a comparison of one room with its own history and nothing else. No venue is ever compared to another venue's sales. The number carries a direction — higher or lower than usual — and no verdict: it is context for a night, never an automatic judgment that the night was good or bad.
How to use it at your venue
Operators tend to read this as a same-weekday, same-venue signal and stop there — a Friday against the median of its recent Fridays, not against last Tuesday and not against another room in the group. Comparing a unit only against its own comparable history is the long-standing way operators benchmark; the restaurant industry tracks exactly this shape as "same-store sales."[1] Because the number is neutral by design, a grounded use is to pair it with context you already hold — a promotion, the weather, a local event, a holiday — before drawing any conclusion from it.
The median baseline is doing deliberate work. Hospitality revenue is right-skewed by a minority of unusually strong nights, so a mean would overstate the "typical" and chase recent outliers; the median is the more honest picture of a usual night.[2] In practice that means a single freak night — a private buyout, a washed-out Saturday — shouldn't swing the baseline the way an average would. That is intended behaviour, not a gap: the baseline holds steady so tonight is read against a stable usual.
How to read it
Read the percentage and its direction as neutral information about one night against its own recent same-weekday history. Above usual is not a win and below usual is not a failure — both are prompts to ask what was different. The comparison is like-for-like by construction: the night and its priors are built the same way, same weekday, same currency, so the difference reflects the room, not a change in how it was measured. Early in a venue's life, or on a weekday it rarely opens, the baseline is thin and the read is worth less trust.
What it doesn't mean
⚠ What it can't tell you
It cannot say why a night differed — weather, events, staffing and promotions all move sales, and none of them are visible here. It never compares venues to each other: a bigger room taking more money is size, not performance, which is why raw sales across differently sized venues are not rankings. No materiality threshold and no performance judgment is defined — nothing here is a flagged exception. And it is not a purely on-premise figure: orders the feed did not classify stay in under the standing bar-tab rule, so this reads the room plus any unidentified trade. Below the number, the channel share names how much of the night was unclassified.
How it's calculated
Computed once on RollCall's servers the moment a night finalises complete, then frozen — screens display the stored result and nothing recalculates it. The eight calendar dates on the same weekday in the eight weeks before the night are each checked directly; closed or missing nights are simply absent, never filled in. The settled, complete nights among them, in the same currency, form the baseline, and their median is compared with the night's net sales. If a night's figures are later corrected so they no longer match the stored comparison, that comparison is withdrawn rather than shown in two versions.
The exact method
Server-computed and frozen at finalise in (m26FromHistory). The window is the eight same-weekday calendar dates in the prior eight weeks, addressed directly (not a fuzzy recent average). A prior night is eligible only if it is settled AND complete, carries a finite integer net, and matches the latest night's weekday and currency. The baseline is the median of those sorted integer nets; the stored result is the rounded percentage difference between the night's net and that median.
Floor: four same-weekday completed nights before a comparison renders — fewer shows the raw night as a fact, never a judgment. Window: eight weeks. This is the ratified active definition w8.floor4.median.v1 with eligibility eligible.v1 (constants M26_FLOOR=4, M26_WEEKS=8, M26_VERSION="m26.v1"). Channel rule: orders the feed classified as takeaway, delivery, drive-through or shipping are excluded; unclassified orders remain in under the bar-tab rule. Read-time usability (m26UsableForNight) requires the stored anchor-net to still match current figures — a correction withdraws the point. Inputs are POS completed-night documents only; no scans or occupancy estimates are involved.
The research behind it
Compare a bar's Friday to its OWN last 8 Fridays — not to some other bar, not to last Tuesday, and not thrown off by one freak big or small night — because that's the only apples-to-apples comparison that actually tells an operator something real about tonight.
-
Same-Store Sales and Customer Traffic (Economic Indicators) ✓ verified
Read the source ↗ -
Hotel-Industry Averages (published as 'Developing the Full Picture on Hotel Industry Averages') ✓ verified
Read the source ↗
Confidence & caveats
The arithmetic is exact on directly observed, register-auditable POS net — there is no estimate or band. But "exact" is not "judged": only the percentage and its direction are ratified for neutral comparison. No materiality threshold and no good/bad label is defined. The comparison is withheld, honestly, below four same-weekday nights, and the read is thinner early in a venue's life or on a rarely-open weekday, where the baseline is short. A no-comparison state is a state, not a zero — it means there is not yet enough of the venue's own like-for-like history to compare against, nothing about how the night actually went.