It is 11:40 on a Saturday. Someone asks how many people are inside. You answer with a shrug and a word: busy, pretty busy, wall to wall. You have run this room for years, so your shrug is probably closer than a stranger's would be. It is still a shrug. And an hour from now, you will not remember which one you gave.
This is a guide to the actual question, how many people are in my bar right now, and the harder one behind it, how full was I last Saturday. It walks through every way venues count a crowd today, what each way is genuinely good at, and where each one gives out. RollCall makes one of these tools, so the last section is about ours. The other four are useful whether you ever hear from us again.
Counting people takes attention. On a quiet Tuesday you have attention to spare, and you also do not need the number. On the packed Saturday you need it most and have none left. So the accuracy of your headcount runs backwards to its value: your best estimate covers the night nobody cares about, and your worst covers the night everything was decided.
That would be survivable if it were only about tonight. It is not. The busy night is the one you get asked about later. How full were you when the promoter's night ran? Was the extra barback worth it? Did the Thursday special actually pull anyone in, or did it just discount the people who were coming anyway? Every one of those is a question about a number nobody wrote down.
"Nobody counted" is a fine answer on a Tuesday and an expensive one on a Saturday.
There is a second problem underneath. Even when a venue does have a number, it is usually one number for the whole building. That number can be perfectly accurate and still hide the thing you would act on: the patio is jammed, upstairs is quiet, and the two average out to a comfortable-looking figure that tells you to do nothing. Staffing decisions do not happen at the building level. They happen at the bar that is four deep while the other bar has nobody at it.
Every one of these is used by good operators, and every one is genuinely good at something. Here is the credit first, then where it gives out, then what it actually costs you.
Someone at the entrance with a tally counter, clicking people in and, in theory, clicking them back out.
It counts real bodies at the real door, with no model and no inference in between. It costs a few dollars, needs no vendor, no install and no explanation, and everyone from the owner to the fire marshal understands what the number means. On a single-door room with a person whose only job is the clicker, it is very hard to beat.
It needs that person. Re-entries, smokers, staff, delivery drivers and the friend who ducks out to move the car all move through the same door, and the decrements get missed far more often than the increments, so the number drifts upward across the night. And when the door gets genuinely busy, the clicker is the first thing that gets put down.
An hour of paid attention for every hour you are open, and a number nobody can audit afterwards. There is no way to go back and check whether Saturday's count was the careful kind or the put-it-down kind.
The oldest method there is. You walk up, you ask, and you get a number: we are at about two hundred.
Free, instant, and better informed than people give it credit for. Your door person has seen this room at every level there is, and their read on busy carries real information about the room right now. For an in-the-moment call, open the second bar or do not, it is often all you need.
It is a judgment, not a record. Two people give two different numbers for the same room. The same person gives a different number at ten than at one, because busy is relative to what they have just been looking at. You cannot compare Saturday to last Saturday because the ruler changed in between.
Nothing today, and every retrospective question later. When ownership, a landlord or an investor asks what your Saturdays look like, an estimate that lives in someone's memory is not evidence.
Using what the till already knows, covers rung, tabs open, transactions per hour, and treating it as a stand-in for people in the room.
It is real data you are already collecting, it costs nothing extra, and it is tied directly to money. If the question is how a night went financially, this is the right number and no crowd count replaces it. It also keeps its record without anyone remembering to do anything.
It counts transactions, not people. A round for six is one tab. Cover-only entries, comps, the designated driver and the crowd that came for the DJ and drank water never appear. It also lags: sales describe the last hour, not this minute. And it cannot tell you which room anyone is standing in.
An undercount that changes size with the crowd. Bigger groups mean a bigger gap, so you cannot calibrate the ratio once and reuse it. The nights it is most wrong about are the nights you most want to understand.
Fixed sensors or cameras over an entrance that count crossings automatically and keep a running occupancy figure.
It counts continuously without spending anyone's attention, it does not get tired at one in the morning, and on a clean single doorway a well-placed counter is the closest thing there is to a real count that runs by itself. For a venue with one controlled entrance, this is a serious answer.
It is a hardware project, with mounting, power, network, sightlines and a cost per door. Doorways are the easy case; open rooms are much harder. Add a patio, a second entrance and a fire exit people actually use, and the coverage gaps land exactly where the money is. There is also a real privacy conversation to have with staff and guests, and somebody has to own it.
Money and a project, up front, before you know whether the number changes any decision you make. And the more doors and rooms you have, which is to say the more you needed this, the further away the properly covered version gets.
The busy-by-hour bars you have seen on your own map listing, built from an average of past activity rather than from tonight.
Free, zero effort, and genuinely useful for the shape of a week. If you want to know that Thursday builds around ten and Sunday empties out at eleven, an average is a reasonable picture of that, and it is available for venues that have never tracked anything.
An average is not tonight. It cannot know about the game, the weather, the festival two streets over, or the fact that your headliner cancelled at seven. It smooths away precisely the nights worth looking at, the unusually good one and the unusually quiet one, because both are outliers to an average. It is also one figure for the whole building.
Nothing at all, as long as you never use it to make a decision about tonight. The moment you do, you are staffing to a description of your past.
Whatever you end up using, including ours, these are the questions to put to it. They are the difference between a measurement and a claim.
An average is a description of your past. If the number does not move when tonight moves, it cannot tell you to call in another bartender, and it cannot tell you that the last two hours have been unusual. Both things are the reason you wanted a number.
Seventy percent full is useless information if the patio is at capacity and the upstairs bar is empty. Decisions get made per room: which bar to staff, which space to open, where to send the crowd. A single building-level figure averages the decision away.
A reading taken at ten is a fact about ten. If a system is still showing it at one in the morning, it has stopped measuring and started asserting. A number you can trust either refreshes or tells you plainly that it is old.
If the figure is typed in by someone who benefits from it looking good, it is marketing, not measurement. Ask two things: who is able to change this number, and what physically stops them. If there is no answer, treat the number as an advertisement.
That last one applies to us as much as to anyone else, so the next section gives the honest version of how RollCall works, including the part where a venue does have a hand in the reading.
Included with every plan. No hardware, no install, and it runs in a browser.
A staff member opens their shift link, takes a photo or a short clip inside the venue, and the crowd level and estimated headcount are read from that image. Nobody selects a status from a menu. The input is a picture of the room as it stands, which means the number has something behind it that you could go and look at.
You walk the venue once and RollCall learns your layout. After that every scan knows which space it is looking at: main floor, bar, patio, upstairs. Each room carries its own occupancy, and the whole-venue estimate is built from them, with coverage and confidence shown so you can see how much of the building a given scan actually saw. There is a longer write-up of occupancy tracking without hardware, including the rooms and doors where fixed sensors are the better call.
A scan stays live for 90 minutes. When it lapses, the number does not sit there showing Packed all week: it falls back to check-in presence until somebody scans again. A crowd number that never expires is a claim about the past dressed up as the present, and we would rather show you less than show you something that quietly stopped being true.
Real check-ins from the RollCall app are recorded server-side. No client can write that count: not a guest's phone, not a competitor, and not your own venue dashboard. It is the one part of this that is out of everybody's hands, including ours in the ordinary course of running the product.
If an estimate looks slightly off, you can nudge it by a step before you post it. That exists because an operator who knows their room is often right when the estimate is not, and your corrections sharpen future estimates for your venue. So yes, a venue has a hand in the reading. What a venue cannot do is invent one: there has to be a scan of the real room behind it, it expires on its own, and the check-in count underneath is untouchable.
It is an estimate read from an image, refined by your own corrections over time, not a turnstile count of every body through the door. It is not a legal occupancy measurement and you should not file anything with it. If what you need is a certified count for a fire-code obligation, this is the wrong tool and you should say so to whoever is asking. If what you need is to know whether tonight is unusual, which room is full, and what last Saturday actually looked like, that is what it is built for.
Knowing tonight is half of it. The other half is being able to answer a question about a night that has already happened, which is the half almost nobody has. Every scan you post becomes part of your own record, so the history builds by itself as a side effect of running the room.
That gives you crowd over time for tonight, the last seven nights, the last thirty or any range you pick. It gives you a busy heatmap by night and hour in your local time, so the slow slots stop being a feeling. It gives you your best nights over the last thirty days ranked by estimated peak headcount, which is a real answer when ownership or a promoter asks what your room does. And because deals are measured against what that night normally does, you can stop repeating the special that never moved anything.
Both halves come with RollCall Pro, at $199 per venue per month for one venue, $159 per venue per month billed annually: tonight's side, live crowd, room-by-room occupancy, staff scan links, deals, and your listing in the app, and the record, your full history and heatmaps, typical-night comparisons, measured deal lift, capacity and utilisation, and your POS connected read-only. Full pricing →
Straight answers, including the awkward one.
Not with RollCall. It runs in your browser from photos your staff take on the phone already in their pocket. There are no cameras to mount, no sensors, no counter to buy and no terminal at the door. Connecting your POS on Pro is read-only, optional, and takes a couple of minutes.
Keep it. A clicker count and a scan answer different questions. Yours is bodies through the door over the whole night. Ours is how full each room is at this moment. If you already have a count you trust, run both for a week and use yours to sanity-check the scan. When a read looks off you can nudge it before you post, and your correction sharpens future estimates for your room.
Each scan stays live for 90 minutes and then lapses, so about that often while you are open, plus a scan whenever the room genuinely changes: after the doors open, once it fills, once late. Staff get a fresh camera-only link every shift, so for them it is a few taps and no login.
It says Quiet, because that is true. RollCall does not round a quiet Tuesday up to look better, and a quiet night is worth recording: it is what makes a good night measurable later. If you would rather not appear at all during a private event or a renovation, you can go dark from your dashboard and keep every bit of your history.
Here is the exact boundary, because it is a fair thing to ask. A reading has to come from a scan of the actual room. There is no box where you type Packed. What you can do is nudge a read by one step before you post it, which exists so an operator who knows the room can correct an estimate that is slightly off. Underneath that, the check-in count is written by our backend only. No client can write it, including your own dashboard, so nobody can inflate or deflate it.
RollCall is a consumer app too. It opens to the public later this year, and from that day your live crowd level is what someone sees when they look your venue up. That is the point of being listed. The detail behind it stays in your dashboard: estimated headcount, the room-by-room breakdown, your history and heatmaps, and anything from your POS.
Start with one venue on RollCall Pro. No hardware, no install. Your first scan tonight, and your history starts building from it.
Contact
Goes straight to the team — we read everything and reply by email.
prefer email? support@rollcall.bar
Sent — we'll get back to you ✓