How do you run on-site check-in for a conference across multiple hotels?

The bottleneck is the handover, not the scan

Vendors sell scan speed. Scan speed is not your problem.

Watch a working lane and the sequence is: the attendee shows a code, the scanner reads it, the printer produces a name label, someone applies it to a badge, threads a lanyard through the badge, adds a bag or a gift depending on who the person is, and points them at the right door.

The scan is two seconds. The rest is sixteen.

This has three consequences that shape everything else:

• Doubling scanner speed changes nothing. Halving the handover changes everything. Pre-threading lanyards during quiet periods, pre-assembling bags by category, and putting the printer within arm's reach of the person handing over the badge each buy more throughput than any hardware upgrade.
• Two people per lane beats one person with a faster device. One operates the system and prints; one assembles and hands over. They work in parallel on different attendees.
• Anything that adds a decision at the desk costs far more than it looks. “Which gift does this person get?” is a two-second question that turns eighteen seconds into thirty. Encode the answer somewhere the volunteer can read without thinking — see badge colours below.

Sizing lanes

Size on the busiest ninety minutes, not on the headcount.

Arrivals are never flat. For a morning-start conference, somewhere between half and two-thirds of same-day arrivals land in the ninety minutes before the opening session. Registration desks that are comfortable across a four-hour window collapse in that window.

The arithmetic:

A worked example. A 1,500-person conference, doors at 09:00. If 900 people arrive between 07:30 and 09:00, that is 600 per hour, which is three lanes running flat out. Three lanes running flat out is a lane with no slack, so build four. Plus one enquiry lane, which does not check anyone in.

That arithmetic assumes a morning with room to arrive. Two other shapes are common, and running events on BagEvent they have settled into their own ratios:

• Check-in opens a full or half day before. Large conferences — academic congresses of 3,000 and more especially — hand out badges the day before. Arrivals spread across the day, so plan one lane per 300–400 attendees collecting that day, with no extra lane on top.
• Check-in opens only one or two hours before the start. Then 60–70% of people arrive in the last thirty minutes. Plan one lane per 150 attendees. The two sets of numbers agree: 150 attendees at 65% is about a hundred in half an hour — two hundred an hour, one lane's pace.

The enquiry lane is not optional and it is not a check-in lane. Every unusual case — wrong name on a badge, not in the system, registration pending approval, someone else attending instead — goes there. Without it, one complicated attendee stops a lane for four minutes and the queue behind them never recovers.

Ratios that hold up in practice:

Lanes

Peak-hour arrivals ÷ 200, plus one

Lanes, check-in the day before

One per 300–400 attendees collecting that day

Lanes, check-in 1–2 hours before the start

One per 150 attendees

Enquiry lanes

One per check-in point, minimum

Lanes from day two

Two, plus the enquiry lane — day one checks in 80% or more

VIP and speaker lanes

One each, or one shared, with the role printed on the badge

Staff per lane

Two — one operating, one handing over

Guides per point

Two — one fixed at the queue head, one floating

Desk width

1.0–1.5 m per lane

Power

Three independent sockets per lane

Network

Symmetric dedicated line, ~10 Mbps per hotel point, ~100 Mbps at a large venue

Symmetric matters more than the headline number. Check-in is a write-heavy workload — every scan is an upload. Venue Wi-Fi with a fat download and a thin upload will look fine in a test and fail at 08:30.

Keep the main lanes for the main case

Most of what slows a lane is someone who is not a standard attendee. Give each of them their own route, decided before the doors open:

• VIPs and speakers get their own lane — one each, or one shared. Print the role on the badge, so the volunteer at the lane and the door steward inside can both see it without asking.
• Exhibitors collect before the event, away from the desk. Each one takes about five minutes — their own staff badges plus the people they bring — and more for a complicated stand. One exhibitor at a check-in lane stops it for the length of five ordinary attendees' check-ins.
• Walk-ups register on their own phone. A QR code sign next to the queue takes them off the lanes entirely; they come back to the enquiry lane once they are confirmed. On most professional events there are few of them — which is exactly why they should not get a lane.
• Materials go out with the badge, at the same lane. A separate collection point for bags looks tidier and costs floor space, a second queue, and people who collect twice.

Distribute the peak instead of absorbing it

The most effective throughput improvement is not a faster lane. It is fewer people arriving at the same place at the same time.

If attendees are staying in hotels, run check-in at the hotels the evening before and hand over the badge, the lanyard and the bag there. Those people never join the venue queue. On the morning of the conference they walk in already wearing a badge.

At scale this means many small check-in points rather than one large one — a large multi-day conference may run check-in across dozens of hotels, each with one or two lanes, plus a main desk at the venue. Distributed check-in is more staff in total but far less risk: a queue of forty people in a hotel lobby at 20:00 is an inconvenience, and a queue of four hundred outside a venue at 08:45 is a story about your conference.

Three things make distributed check-in work:

• Every point runs the same procedure. Same desk layout, same script, same device configuration, same materials. Volunteers are not transferred between points once trained, because a point's procedure is muscle memory by the second hour.
• Every point has one named technician and one named organiser contact, both reachable in a messaging group for that point specifically. Not a general group — one per point, so a printer jam at hotel eleven does not surface in a channel with two hundred people in it.
• Check-in follows the person, not the building. Someone who is booked into hotel A but shows up at hotel B should be able to check in at B. What they cannot do at B is collect a room key. Keep the two flows separate and say so at the desk: check-in is one system, accommodation is another.

The countdown

Dates relative to the first check-in day.

The 06:00 handover is the item most often missed and most expensive when it slips. “The venue will set up in the morning” means the venue sets up at 07:40, you discover the power strips do not reach the desks at 07:45, and you open late. Put the handover time in the venue contract.

The T-3 test has to be on the real network. A check-in system that works in the office and fails at the venue is almost always a network problem, and you want to find it three days out, not on the morning.

Volunteers

Most check-in staff are casual hires who have never seen the system. Design for that.

• Train on the day before, not the morning of. Thirty minutes covering the script, the materials, the badge colours and the gifts. One hour on the device: scan, print, search by name or phone, the three error screens they will actually see. Then have each of them do a full check-in while you watch.
• Split the roles explicitly. Scan and print. Materials and handover. Queue guidance. A volunteer who is doing all three is doing none of them well.
• No substitutions after training. If someone cannot come, train the replacement, even briefly. An untrained volunteer at a lane is slower than an empty lane, because the queue still forms.
• Teach them how to describe a fault, not how to fix it. “The scan works but nothing prints” and “the printer light is off” send a technician to different places. “It's broken” sends them nowhere. Explicitly forbid opening or re-cabling equipment.
• Tell them what they may not do with the data. Attendee names, phone numbers and employers are on their screen all day. No photographs of the screen, no sharing, no exceptions. This belongs in the training, not in a policy document nobody reads.

Give them the exact words for the four things they say two hundred times:

Good morning — could I see your QR code?

Here is your badge, and here is your conference bag.

Not registered? Please scan this code and fill in the form. Come back to this desk once it is approved.

If we cannot answer something, we will take you to someone who can — we will not guess.

That last line matters more than it looks. A volunteer guessing at an answer creates a problem that surfaces two hours later at a different desk.

Badge colours are an operating system

Colour-code lanyards and badges by attendee category — VIP, guest, media, analyst, staff, crew, exhibitor — and tie every entitlement to the colour.

This is not decoration. It removes decisions from the desk. The volunteer does not need to know who gets which gift; they need to know that red gets the VIP set and green gets the small one. It also makes room control visible: a glance across a session tells you whether the people in the room belong there.

Print the name onto a label applied to a pre-cut badge rather than printing whole badges. Label printers are faster, cheaper and jam less than card printers, and a mis-spelled name costs a label rather than a card.

The second and third check-ins nobody plans for

Registration is the check-in everyone budgets for. Two more happen.

Session access. Parallel tracks, closed-door sessions and workshops each need their own scan at the door, both to enforce access and to know who was actually in the room. This is a different device, a different operator and a different failure mode from registration — and it happens while the registration desk is quiet, so the same staff can often cover it. What breaks it is unclear naming. Label each collection point with date, venue, session and start time so an operator holding a device knows which point they are, and put that label on the login code itself.

Meals. If meals are ticketed, redemption must be deduplicated across every location at once, not per restaurant. Someone who has eaten at venue A must be refused at venue B. Time-box the redemption window as well — a voucher valid from 17:00 to 21:00 and nowhere else in the day.

The operational consequence is that “already redeemed” must not read as “already redeemed here.” Volunteers need to be told, in training, that the message means the voucher was used somewhere — possibly across town — and that the answer is not to try again.

Device custody

Dozens of tablets, scanners and printers move across many locations and many hands. Treat it as a logistics problem with an owner.

• One named person issues and receives every device. Not a team. A person.
• Verify at handover, in front of the recipient: battery, screen intact, charger and cable present, system responsive, camera working.
• Devices are issued already signed in, on the single screen they are meant to be on. No password on the device, because a volunteer locked out at 08:15 is a closed lane.
• Written rules: do not sign out, do not open other apps, do not delete anything. Say it plainly and say it twice.
• State the replacement policy before handing anything over. Lost or damaged equipment is charged at replacement cost. Saying this at issue is what makes it true at return.
• Return to the same named person, at a stated time and place.

The rules read as officious. They exist because every one of them is a failure that has happened.

Roving

Assign someone whose only job is to walk the points — twice in the morning, twice in the afternoon, plus on demand.

They are not there to help at a lane. They are there to see what nobody is reporting: the point that is quietly two hundred badges behind, the printer that has been jamming every fifth print, the volunteer who has been improvising an answer since 09:00.

At distributed events, the roving pass before doors open is the highest-value hour of the entire operation. Every point tested, every volunteer confirmed present, every device online — before anyone is queueing.

What actually goes wrong

Eight situations account for nearly all of it. Write the response down, put it on the desk, and rehearse it.

Two principles run through all eight. The desk never adjudicates — anything requiring a judgement goes to the enquiry lane or the organiser. And nothing blocks entry that does not have to: give people the badge, take the details, fix it afterwards.

Where software actually helps

Most of the above is operations, and no product replaces it. Software helps in four specific places:

• One attendee record across every touchpoint. Registration, hotel check-in, session access and meal redemption reading and writing the same record — which is what makes cross-location deduplication possible at all.
• Search that works when the code does not. Name, phone or email, fast enough to use in front of a queue.
• Live counts per point. Knowing at 08:20 that one point is falling behind is the difference between moving a volunteer and reading about it in the debrief.
• Offline tolerance. Venue networks fail. A check-in client that keeps working through a dropout and reconciles afterwards turns an outage into a delay instead of a stop.

Everything else on this page is people, timing and arithmetic. Get those right with a spreadsheet and you will outperform a team with better software and no plan.