The brief that arrived on a Monday
A major sporting event landed on the desk on a Monday afternoon. The host venue would be released for technical access on the Wednesday. The event kicked off on the Saturday. 72 hours from empty stadium to live network supporting 50,000 fans, 2,000 press and production, a broadcast LAN feeding the international rights holders, and the event command-and-control operations centre. The host broadcaster needed the production LAN cleanly handshook 36 hours before the first feed, which compressed the practical build window further.
Our team had done the underlying work many times before. We had not done it this fast. The story of what made the 72-hour window achievable is mostly about what happened in the eight weeks before the trucks arrived, not about heroics during the build week itself.
Why 72 hours is the hard number
Stadium event IT typically wants a 5-to-7-day build window for a deployment this size. The reason is not the install time. Pulling cable, mounting access points, racking switches, configuring the controller plane and validating the network are well-understood activities and the labour scales linearly with team size. The reason is the integration time.
Predictive RF surveys validated on site, broadcast LAN handshakes with the rights holders' production crews, ticketing system integration, security-services integration with venue and police command, accreditation network testing, and the full incident-response rehearsal cycle each take a working day to do properly. They cannot all happen on the same day because they share venue access windows, broadcast crew time and engineering attention. 72 hours collapses what is normally a sequence into a parallel-track delivery.
The 72-hour build only works if the parallel-track sequencing is engineered into the pre-build phase. If you try to engineer it during the build week, the math does not work.
The eight weeks before the trucks arrived
The pre-build was the work that actually made the timeline. Six things shipped during the eight weeks ahead of venue handover.
Predictive RF model. Two seniors took the venue drawings and the seating layout and modelled the WiFi coverage in Ekahau with a representative client-density profile. The model produced the AP count, the AP placement plan and the channel and power allocation. It did not survive contact with the actual venue without adjustment, but it was 90 percent right on landing, which is what matters when the build week starts.
Kit staging. Every access point, every switch, every PtP radio, every cable bundle was unpacked, tested, configured, labelled and re-packed into venue-specific kit boxes at our warehouse. By the time gear left for the venue, each box was tagged with its destination and its install runbook. The team unpacking it on site did not need to think about what went where.
Broadcast LAN handshake. The host broadcaster's production crew flew in two weeks before the event for a technical walkthrough. We had the broadcast LAN topology, the IP plan, the multicast configuration and the latency budget signed off by their lead engineer before we left that meeting. Day-of integration becomes a verification exercise, not a design exercise.
Venue, regulator and security co-ordination. Spectrum allocation with the national telecoms regulator, integration with the venue's permanent CCTV and access control estate, co-ordination with the security command-and-control operations centre. All of these had calendar weeks of meetings, written agreements and rehearsed escalation paths before the trucks left the warehouse.
Runbook authoring. Every recurring task in the 72-hour build had a written runbook. Doors-closed daily checks, broadcast handshake, RF re-validation, dress rehearsal, incident response. The runbooks were short and specific. They were tested with the team during the rehearsal week. Engineers running their twelfth hour of a shift do not improvise well; they execute a runbook well.
Engineering team commitment. The 18 engineers who would deploy were named two months ahead, briefed on the venue and the operating plan, and rehearsed on their specific roles. Three of them flew in from Toronto where our team has international event experience. The team that landed at the venue was not a team meeting for the first time.
The kit
The build at the venue:
220 HPE Aruba AP-635 WiFi 6E access points across the bowl, concourses, hospitality, back-of-house and the broadcast compound. AP placement matched the predictive RF model with around 15 percent adjustment on site for venue features the drawings did not capture.
48 Ubiquiti point-to-point microwave radios providing redundant backhaul to the carrier handoff and to the broadcast compound, plus inter-zone links where fibre runs were impractical inside the 72-hour window.
12 Aruba CX 8325 switches in the venue core with VSX redundancy. A Cisco Catalyst 9500-class production LAN physically segmented for the broadcast crew. An active-passive Palo Alto firewall pair carrying perimeter and tier-aware policy.
Two 10 Gbps microwave uplinks to the broadcast compound with full redundancy. The two paths terminated at different building entries on the broadcast side to defeat single-point failures.
A full temporary network operations centre in a 12U portable rack at the venue, running real-time dashboards, the spectrum monitor, the broadcast LAN visibility stack and the security operations centre toolchain. The NOC was staffed continuously from broadcast crew arrival through final whistle.
The 72-hour build, day by day
T-minus 72 (Wednesday). Venue handover at 06:00. Trucks arrive 06:30. Kit on-site by 09:00. AP and PtP install starts immediately across all zones in parallel, eight crews working from the eight kit areas the warehouse staged. Cabling crew runs Cat6A and fibre from cable trays where they exist and overhead trays where they do not. By 22:00, every AP was mounted and every switch was racked.
T-minus 48 (Thursday). Core configuration. Switch stacks built, VSX configured, firewall pairs synchronised. Controller plane brought up and APs joined the controllers. The RF coverage plan was validated zone by zone with handheld survey tools and the predictive model was adjusted for two areas where the real-world propagation deviated more than expected. By end of day Thursday the venue had working WiFi everywhere.
T-minus 24 (Friday). Broadcast handshake at 10:00. The host broadcaster's production crew walked the broadcast LAN with our lead, verified the IP plan, ran the multicast tests, signed off the broadcast network. Press centre LAN commissioning. Ticketing and accreditation integration. Security command and control integration with the venue and the police command. Full incident-response rehearsal at 18:00 with the full team running drills on every category of failure we had pre-imagined.
Event day (Saturday). Doors-closed checks from 06:00. Final RF re-sweep. Broadcast LAN re-validation. Spectrum monitor running continuously. The 18 engineers were on shift across the venue with the remote NOC backstopping from Abu Dhabi. The broadcast crew was on site by 09:00; the production stack was fully integrated by 11:00. Doors opened at 13:00. First whistle at 18:00. Zero P1 incidents on the live network through the entire event window.
What went right
Pre-staging was the single most consequential decision. Every AP, every switch, every PtP radio left our warehouse already configured, labelled and runbooked. The build crews on site executed against the kit boxes rather than thinking from scratch. The error surface compressed dramatically.
Predictive RF modelling with on-site adjustment. The Ekahau model got us 90 percent of the way; the remaining 10 percent was caught during the Thursday validation walk. A pure on-site survey approach would not have fit inside the 72-hour window.
Eighteen engineers, named and rehearsed. The team that walked into the venue on Wednesday morning had been operating against this venue's plan for eight weeks. They knew their zones, their kit, their runbooks and their escalation paths. The build moved at the speed an experienced crew can move; not the speed a new crew can move.
Runbook discipline. The recurring tasks ran from documented procedures. Doors-closed checks, broadcast handshake, RF re-validation, incident drills. The runbooks compressed decision-making and reduced the variance that comes from individual judgement under time pressure.
The two moments that nearly went wrong
A carrier-side fibre cut at 14:00 on Friday, 28 hours before kickoff. The primary uplink to the venue showed loss patterns that the test traffic profile did not match. We could not pinpoint the cause in the available window. Our pre-staged spare Ubiquiti PtP link went up in 40 minutes and was traffic-balanced into the venue uplink in another 20. The carrier found and fixed the damaged splice overnight. The primary feed never lost a packet during the event itself. Pre-staging is not optional at this scale.
A broadcast LAN configuration drift caught at 23:30 on Friday, six hours before the broadcast crew was due back. The drift scan flagged a stale spanning-tree priority on one of the broadcast switches, left over from a non-production test during the staging week. The fix was straightforward (replace the running config with the validated baseline, re-test the affected switchport, re-validate the broadcast handshake) but the cost of catching it 30 minutes later would have been substantial. Every event we deliver now runs version-controlled broadcast configurations with twice-run drift scans, partly because of this moment.
What we would change for the next 72-hour build
Stage even more spare kit. Every event teaches you that the gear you thought was excessive is the gear that saves the day. The carrier fibre cut consumed the contingency we had pre-positioned; we now stage about 25 percent more contingency than we used to.
Earlier broadcast LAN integration. The 36-hour pre-broadcast handshake window worked, but a 48-hour window is comfortable and worth the additional engineering attention. Standard now for events at this scale.
Rehearse partner-equipment failures, not just our own. The broadcast LAN scenarios we rehearsed all assumed our equipment failed. The 2026 rehearsal cycle includes scenarios where the broadcast crew's equipment fails and our network has to react cleanly to it. Different muscle memory.
Why this matters for stadium IT generally
Stadium-scale event WiFi at 50,000-attendee density on a 72-hour build window sits near the operational ceiling for event IT in the region. The same components, predictive RF, pre-staging, named teams, runbook discipline, broadcast handshake rigour, contingency depth, scale down cleanly to smaller venues and shorter timelines. They do not scale down well if you skip them.
The honest framing: there is no shortcut to 72-hour stadium WiFi at this scale. The work that makes it possible happens eight weeks earlier. The work that fails it happens in the assumption that on-site heroics will close the gap.
Bottom line
Pre-staging, predictive RF, an 18-engineer named team, written runbooks and partner-equipment rehearsal. Five components, none individually glamorous, collectively the reason a 50,000-seat stadium delivered cleanly on a 72-hour build window with zero P1 incidents during the event.
For organisers planning stadium-scale event IT on tight timelines, the lesson is operational rather than architectural. The architecture is well understood; the operating discipline that makes it work under compressed timelines is where the real differentiation lives.
