Why Indoor Navigation Fails When an Exhibition Fills Up: Offline Positioning at Computex

The Enemy of Exhibition Positioning Is Not Walls, It Is People

Indoor positioning normally fights walls: partitions, metal structures, elevator shafts. An exhibition hall is by comparison an easy site, with high ceilings, open floors and clean sight lines. On paper it is the best venue you could ask for.

Then the doors open.

A major trade show draws tens of thousands of visitors a day. Tens of thousands of phones and countless mobile hotspots crowd into the same space, all fighting over the 2.4GHz band, and human bodies absorb radio signals as well. The clean coverage map you measured in an empty hall stops being true the moment the crowd arrives.

What decides the outcome is connectivity. The on-site record from Computex 2014 is blunt: At peak hours the on-site WiFi and 3G could not connect reliably, and users on our app kept navigating normally.

Losing the network is fatal to any positioning architecture that puts the computation in the cloud. The cloud pipeline runs like this: the phone collects nearby signals, uploads them, the cloud matches them against a fingerprint database to work out the position, and the result comes back. Every step needs the network. That design works well in offices, hospitals and other venues with stable connectivity. The model can be updated at any time, the back office sees live status, and the client on the phone stays light. An exhibition is not that kind of venue. The show lasts a few days, the crowd is extreme, and the network is least available exactly when it is needed most, so every advantage of the cloud architecture collapses into the same single point of failure. Users do not distinguish between a broken network and broken positioning. They simply decide the app does not work.

Moving the Computation Back to the Phone: How Offline Positioning Works

Turn the problem around: keep the entire positioning process off the network.

A Bluetooth beacon does one thing: it broadcasts its own identifier continuously and never opens a connection to any device. The app stores the map and the beacon coordinates locally in advance, then simply listens on site. Which beacons it hears and how strong each signal is are enough to compute the position on the phone itself. Not one step in the chain needs uplink traffic. Bluetooth also runs on 2.4GHz, but it does not have to hold a connection open: the beacon broadcasts one way, so a few dropped packets do not matter and the phone picks them up again a second later. A congested band drags down connections that carry data back and forth, not the act of listening to a broadcast.

Aspect	Cloud-computed positioning	Offline positioning
Position computation	Phone uploads the signals, the cloud computes and returns the result	Done entirely on the phone
When the network drops	Service stops	Keeps working
Map data and model updates	Can be pushed at any time	Must be downloaded in advance while a network is available
Live crowd data in the back office	Yes	No (can be designed to upload afterwards)
Suitable venues	Permanent venues with stable networks	Event sites with extreme crowds and unreliable networks

This approach went through its exam at Computex, June 3 to 7, 2014. We installed 5,000 beacons across Taipei World Trade Center Hall 1, Hall 3, and the first and fourth floors of the Nangang Exhibition Center, and released Android and iOS apps at the same time. Over the five days of the show there were 8,500 unique users. At peak hours the on-site WiFi and 3G could not connect reliably, and navigation kept working. Positioning accuracy came in at 3 meters, exactly the width of a standard booth, which is just enough to walk someone to the front of a stand.

Deployment density is the other useful comparison. In 2017 we deployed more than 700 beacons at Hong Kong International Airport, covering positioning and navigation across the whole site; the three exhibition halls took 5,000. The difference is what the job demands. Booths are measured in 3-meter units and human bodies attenuate the signal badly, so the phone has to hear enough beacons at once to intersect them, which means placing them densely. As far as we know, 5,000 units is still the largest beacon-based indoor navigation deployment at a single event anywhere in the world. Beacons run on batteries that last two to three years, installation needs no cabling, and they are simply collected at teardown, so a five-day run asks nothing of them.

Want to know how many units your venue needs and how much installation time to allow? Send us the floor plan and we will estimate it for you.

Book a Site Assessment

The Price of Going Offline, Stated Up Front

The benefits of running offline are paid for with something else. Four things to know before you evaluate it:

Map data has to be downloaded first. Users need to load the map data somewhere they still have a network. Anyone opening the app for the first time on site gets stuck at that download. Pre-event publicity and the check-in flow have to make room for it.
Conditions on the phone are out of your hands. With Bluetooth switched off or location permission denied, navigation will not run. That is true of any Bluetooth positioning solution, but in an offline architecture there is no back office to diagnose the user remotely, so the in-app prompts have to be more direct.
The organizer's live dashboard is limited. If phones send nothing back, the back office has no live position stream. Crowd statistics have to be designed to upload once a network is available, which costs you immediacy. Where the network is down, you cannot have both "visitors find their way" and "organizers watch everyone in real time"; immediacy is what gives way.
Last-minute changes cannot be pushed. If a booth moves or a route is closed off at short notice, editing the map data still will not reach phones that are already offline. The freeze date for map data has to be agreed with the organizer before the show.

Our judgment is that at an exhibition where the network dies as soon as the crowd arrives, all four are affordable, because the alternative is a service that does not work at all. At a permanent venue with a stable network, the same trade is not necessarily worth making. Site conditions decide the architecture, not the other way around.

What to Settle Before You Deploy

Cost and schedule depend heavily on the venue, so we do not publish generic numbers in an article. There are three things worth deciding before the first conversation, and they make it go much faster:

Where the app comes from. For Computex in 2014 we released standalone Android and iOS apps for the event. Will your event live inside an existing official app, or ship its own? That decides the installation burden on visitors and how the pre-event publicity has to work.
The installation and teardown window. Beacons need no cabling, so installation is fast: at Computex, 5,000 units took three days from the start of installation to the end of signal collection. When your venue lets crews in directly determines the scale and schedule that are feasible.
The map data freeze. When the booth layout is final, and how later changes are handled, is a timeline to set together with the organizer.

The answers to all three are already in your venue and the shape of your event. Work them out first and the opening meeting can go straight to scale and schedule.

Which Venues Should Consider Offline Positioning

A few criteria; the more of them you meet, the more offline positioning belongs on your shortlist:

Daily attendance during the event runs into the tens of thousands
The venue network was designed for exhibitor operations, not for visitors
Half an hour of navigation downtime turns straight into complaints or press coverage
The run is short, leaving no time to expand network capacity for the event

What trade shows and concerts have in common is that crowd density is treated as an exception when network capacity is designed and is the norm on the day itself. A positioning choice should start from the worst conditions on event day, not from the empty-hall signal map taken during the site survey.

Frequently Asked Questions

Is offline positioning accurate enough to walk someone to a booth?

Bluetooth beacon positioning lands between 1 and 3 meters. Measured at Computex it came in at 3 meters, and a standard booth is exactly 3 meters wide, so the error is roughly the size of one stand. That is enough to bring someone to the front of a booth. Finer positioning inside a booth calls for a technology such as UWB, accurate to 10 to 30 centimeters, with an entirely different cost structure.

Can organizers still get crowd data?

Yes, but not in real time. The app can be designed to upload anonymous route records once the phone reaches a network, which is fine for analyzing hot zones and dwell time afterwards; a live wall display during the event is not possible. If real-time crowd data is a hard requirement, pair it with gate counting or another measurement that does not depend on phones, and let the two data sets complement each other.