The registration desk opened at 8 a.m. By 8:14, the badge scanners had stopped talking to the check-in server, a line of 300 people had formed, and the event’s technical lead was on the phone with a venue rep who kept repeating that the Wi-Fi was working fine.
The Wi-Fi was working fine. That was true and completely beside the point.
Anyone who has run a conference, expo, or large meeting has some version of this memory. And it usually traces back to the same root cause: someone treated a temporary network like a home router problem when it’s closer to an air traffic control problem. Here’s what actually goes into building one that survives contact with a crowd.
Start with a device count, not a speed
The instinct is to ask how many megabits you need. Wrong first question. Ask how many radios will associate to your access points at peak.
Rule of thumb that holds up well: attendees bring 2.2 to 2.7 connected devices each. Phone, laptop, watch, maybe earbuds that hold their own connection. So an 800-person conference isn’t 800 clients. It’s closer to 1,900, and they don’t arrive evenly — they arrive in a spike at registration and again when the keynote ends and everyone checks email at once.
A decent enterprise access point handles 50 to 70 concurrent clients well. Push past that and you’re not slow, you’re unstable: association timeouts, DHCP failures, devices that show full bars and move no data. The fix is more access points at lower power, not fewer at higher power. Counterintuitive, and it’s the mistake almost everyone makes on their first deployment.
Then do the bandwidth math honestly
Not every device is streaming 4K. Real conference traffic is mostly email, messaging, web, and a long tail of cloud sync. Budget 0.5 to 1.5 Mbps per active user for general sessions, and concurrency is usually 30 to 50 percent of devices, not 100.
Where the number explodes is anything involving upstream video. A single livestreamed session at 1080p wants 8 to 12 Mbps up, held steady. Twenty exhibitors running product demos over video calls will together want more upstream than the entire attendee population. Hybrid events shifted this permanently — upstream used to be an afterthought on asymmetric connections, and now it’s frequently the binding constraint.
“I ask clients for their run of show before I ask for anything technical,” said Marcus Whitfield, a network consultant who builds temporary infrastructure for trade associations. “If there’s a livestream, a hybrid panel, or a hundred exhibitors doing demos, that changes the circuit I order. If it’s eight hundred people checking Slack, that’s a completely different build and a much cheaper one.”
Choosing how bandwidth gets to the room
Four practical paths, each with real tradeoffs.
Venue fiber or house internet. Cheapest when it exists and is actually provisioned for events. Verify the committed rate, not the advertised one, and ask whether it’s shared with the hotel’s guest network. It usually is.
Dedicated temporary circuit. A GigE drop ordered from a carrier gives clean, predictable capacity. Lead time is the catch — two to six weeks is normal, sometimes longer. This is a decision you make when you sign the venue contract, not during load-in.
Point-to-point microwave. Where line of sight exists to a nearby fiber-fed building, a microwave shot can deliver serious capacity on days of notice rather than weeks. Common in dense downtowns and around convention districts.
Bonded cellular. Multiple carrier SIMs combined into one link. Fast to deploy, portable, and now capable of supporting a few hundred devices from a single road case. The caveat is honest: your attendees are congesting the same cell sectors you’re relying on, which is why multi-carrier bonding rather than a single hotspot matters. For outdoor sites, festival grounds, and pop-up activations, this is often the only sane option, sometimes paired with satellite where terrestrial coverage is genuinely absent.
Providers focused on this niche publish practical walkthroughs of the whole sequence — Trade Show Internet’s guide on how to set up temporary WiFi for an event covers the planning order in more depth than most vendor material bothers with.
The unglamorous settings that decide everything
A short list of things that quietly ruin deployments:
DHCP scope sizing. A /24 gives you 254 usable addresses. Your 1,900 devices will exhaust that before the keynote. Size for a /21 or larger and shorten lease times to 30 or 60 minutes so addresses recycle as people leave the ballroom.
2.4 GHz discipline. There are three non-overlapping channels. Three. In a hall with neighboring exhibitor hotspots, 2.4 GHz is a swamp. Many deployments now disable it entirely on most access points, keeping a few for legacy scanners and IoT devices that can’t do anything else.
Channel width. The temptation is to run wide 80 MHz channels for headline speeds. In high-density rooms that’s backwards — narrower 20 or 40 MHz channels give you more non-overlapping options and better aggregate throughput across many clients.
Rogue hotspots. Exhibitors bring their own routers. Dozens of them. Every one adds interference. Getting a no-personal-hotspot clause into the exhibitor manual does more for network quality than most hardware upgrades, though enforcement is its own diplomatic exercise.
Captive portals. If you’re collecting emails through a splash page, test it on current iOS and Android before the doors open. Captive portal detection breaks in creative new ways with every OS release, and a portal that won’t load reads to attendees as broken internet.
Survey with people in the room
An empty ballroom and a full one are different radio environments. Bodies absorb 2.4 and 5 GHz energy, so coverage that measured clean at setup can develop holes at capacity. Walk the space during the first session with a survey tool running and note where signal collapses.
“We do a load test with synthetic clients before doors, then a physical walk during the opening session,” said Elena Vasquez, an event operations director who runs about fourteen conferences a year. “Twice now that walk caught a dead zone in a back corner where we’d put the press table. Both times we fixed it during a coffee break instead of hearing about it in the post-event survey.”
Have a failure plan
Networks break. The question is what happens next.
Keep badge scanning able to run offline with local caching and later sync — registration is the one system where a five-minute outage becomes a two-hour line. Have a cellular backup for payment terminals. Know which single person is authorized to make the call to fail over, because deployments where three people are debating while a keynote streams to nobody go badly.
And put a real number on what an outage costs before deciding the network budget is too high. For an event where sponsors paid for livestream visibility and exhibitors paid for lead capture, a failed network isn’t a technical inconvenience. It’s a refund conversation.
The best temporary networks are the ones nobody at the event ever thinks about. That invisibility is expensive, deliberate, and worth planning for months ahead of the doors opening.
