The failure nobody can see
A fan is in the building, watching the play happen, holding a phone that works for everything else, and the app will not take the bet. The sportsbook's status page says the platform is healthy. It is telling the truth, and that is exactly the problem.
The failure does not live in anyone's cloud. It lives in the radio environment between the fan's phone and the outside world: tens of thousands of devices sharing one finite slice of spectrum inside a steel and concrete bowl, a crowd of bodies that absorb the frequencies everything depends on, and an uplink saturated by twenty thousand people posting video at the same moment a bet confirmation has to get out.
Nobody owns it. Not the sportsbook, not the carrier, not the venue, not the geolocation vendor. Each of them is doing its job correctly. The failure sits in the seams between them, and a failure that never appears on anyone's dashboard never gets an owner, a budget, or a roadmap slot.
The backend is innocent. The building is the crime scene.
Which is why we start with the building, not the software.
Two failures, and the order matters
Everyone's instinct is "stadium Wi-Fi is bad, buy more bandwidth." That instinct is wrong, and being right about the order is the whole method.
Indoors, the first thing to break is not the data connection. It is the proof of where the fan is standing. Before a wager can be accepted the operator has to prove the bettor is physically inside a state where betting is legal, and that check fuses satellite GPS, cellular tower data, a scan of nearby Wi-Fi networks and the public IP address into one verdict. Every one of those inputs is a radio signal, and every one degrades in a packed venue. Satellite signals in particular are deliberately weak and do not penetrate roofs or overhangs; deep in a sealed bowl they are often simply gone. A fan can browse the web fine and still be unable to bet.
The network fails second. A small, latency sensitive payload has to cross a congested uplink, and carriers rationally deprioritize traffic under load to keep the network stable. That degrades the experience, but it does not by itself block the wager. Fix the two in the wrong order and you have fixed nothing.
Why we survey before we design
The location check is not failing because the geolocation vendor is bad at its job. It is failing because the building is not feeding it enough clean signal to do its job. That is a supply problem, and supply is something you can install.
But you cannot install it well without knowing where it is missing. No one has ever measured the in-venue location check failure rate: how many checks fail inside a specific venue, on a specific night, per carrier, per seating tier. It cannot be inferred, modeled credibly, or bought from anyone. It can only be measured in the building. So the measurement is not a preliminary step before the product. In phase one, the measurement is the product.
How the survey works
Every venue engagement is two sessions and one deployment. We place a small fleet of self powered sensor nodes at fixed, photographed, position tagged survey points, ideally the day before the event, and leave them in place across both sessions.
- Empty baselineThe building in its game configuration, with no crowd. Video boards on, lighting on, radios quiet. This is what the structure alone does to the signal.
- Game daySame nodes, same points, with the bowl full. Gates open, pregame, kickoff, halftime, the two minute warning and egress each get their own window, because congestion peaks at specific moments and a session average washes the effect out.
- The deltaThe finding is the difference between the two. If the empty and full distributions overlap, it is the building, and siting fixes it. If they diverge, it is the crowd, and capacity fixes it. Single visit data cannot tell those apart, and any conclusion drawn from it is contestable.
Survey points are the same every visit: the concourse ring, the betting zones and lounges, the lower bowl in the open and under the overhang, the upper deck, club and suite level behind coated glass, the tunnels and vomitories, and a control point outside the gates.
What we measure
GNSS starvation
Per satellite signal quality, fix quality and jamming flags at the receiver, plus any interferer rising above the noise floor in the GPS bands. Attenuation and multipath are invisible on a spectrum screen and equally fatal; the receiver leg catches them.
Band congestion
Channel occupancy and utilization across 2.4, 5 and 6 GHz, and non Wi-Fi emitters sitting on top of them. The location check leans on Wi-Fi scanning, so a saturated or empty scan starves it directly.
Per carrier radio conditions
Signal strength, quality and serving band measured phone side on one instrumented handset per national carrier. A fan's experience depends heavily on which carrier they are on, so the survey reports it that way.
Timed probes and real checks
Round trips sized like a location check payload on every exit path, and manual location check attempts on real accounts at each point. This converts radio conditions into the number that matters: did it go through, and how long did it take.
What it records is spectrum load, not people: Bluetooth and Wi-Fi capture is counted as traffic volume per location, payloads are stripped before anything leaves a node, and every deployment happens under the venue's written authorization.
A survey that cannot lose is not a survey
Kill criteria are written down before collection starts. If the empty and full distributions overlap, the crowd thesis is wrong. If in bowl accuracy sits comfortably inside the permitted boundary at an interior venue, the compliance argument narrows to border venues. If real location checks pass above ninety nine percent in the bowl, there is no commercial problem regardless of what the spectrum says. We would rather learn that in phase one than after two years of building.
From the survey to a venue specific fix
Once the degradation is measured, the fix stops being generic. An open air NFL stadium where satellite positioning half works and the network is the binding constraint needs a different design from an eighteen thousand seat arena under a closed roof where a bettor deep in the bowl has no satellite fix at all. A building with an NBA and an NHL tenant on roughly eighty home dates a year justifies a different install from one that hosts ten. The survey tells us which building we are in.
The hardware itself is three physical elements, and the survey decides where each one goes:
- High capacity access points inside the bowl. Wi-Fi 7 on a clean band, sited against the survey's own dead zones and measured betting zone density rather than a generic coverage design. They do two jobs at once: surveyed in-venue anchors supplying the clean Wi-Fi scan data the location check depends on, and real uplink headroom for the transaction.
- An external antenna with a clean sky view. Multi carrier cellular plus GNSS, sited outside the structure precisely because that is where satellite and cellular reception is uncompromised, giving the venue an unambiguous positional reference.
- A half rack of compute in the venue. Local processing and delivery of the signal at game day scale and latency. The signal has to return at bet time without slowing the wager, and that budget does not tolerate a round trip to a distant data center.
Success criteria are defined against the survey's own baseline, so improvement is provable rather than asserted, and the business case is built on the venue's own numbers from the venue's own game day. The same infrastructure then fixes the second failure almost for free: a clean band offloads the congested cellular uplink and sidesteps carrier deprioritization entirely, because that mechanism does not apply on a band we control.
What this needs from you
Less than you would think. Phase one asks for access and data, not budget.
- Game day access and RF coordination sign off.
- One empty venue window in the same physical configuration as the event, ideally the day before.
- Operator telemetry, if a sportsbook is involved. Location check failure logs for the same venue and window. Our radio data alone shows the environment is bad. Correlated against those logs, it shows exactly what it costs. That correlation costs the operator nothing but an export.
Tell us which building, and we will tell you honestly what we expect to find there before we switch anything on.