1. Executive readout
Across LAX, observed traffic splits almost evenly between Wi-Fi and cellular — Wi-Fi carries 54% of total observed data. That airport-wide parity hides real per-terminal spread: some terminals lean on Wi-Fi while others, with heavy passenger dwell, see Wi-Fi under-absorbing demand — the candidate gaps below.
54%Airport-wide Wi-Fi offload share
10Terminals analyzed (gate-proximal)
21,923 GBObserved Wi-Fi usage (gate zones)
T1Top Wi-Fi gap candidate
Top gap candidates (high passenger dwell where Wi-Fi is under-absorbing demand):
- Terminal 1 — offload 49.8%, user-density median 24.0 (gap score 12.0) · 342 APs, Wi-Fi 5, mgr Boingo, LAWA · Boingo — discretionary
- Terminal 2 — offload 36.7%, user-density median 16.5 (gap score 10.4) · 72 APs, Wi-Fi 6 (some 5), mgr LAWA · RFP-Lite? — discretionary
- Terminal 8 — offload 47.0%, user-density median 11.0 (gap score 5.8) · 26 APs, Wi-Fi 5, mgr Boingo · Boingo — discretionary
What is and isn't justified now: Terminals inside TDIP capital projects
(T4, T5, TBIT) get Wi-Fi upgrades through that funded work — keep them out of discretionary
roadmap spend. Discretionary attention belongs on high-dwell, low-offload terminals not
capital-funded — which lines up with the Boingo end-of-life renewal (T1, T7, T8, American
Eagle; the 232 Wi-Fi 5 APs) and the RFP-Lite terminals (T2, T3, T6, West Gates). Where cellular/DAS
already carries the use case (§6), a Wi-Fi upgrade is not warranted at this stage.
⚠️ Honesty up front: this is crowdsourced, Android-weighted usage
data — a proxy for where Wi-Fi is used, not a measurement of signal strength, capacity,
or interference, and it cannot isolate LAWA-managed Wi-Fi from carrier/Passpoint Wi-Fi. Findings
are indicative. See §8.
2. Venue overview
LAX: nine terminals (1, 2, 3, 4, 5, 6, 7, 8, TBIT/Bradley) plus the Midfield Satellite Concourse (West Gates) and regional/landside facilities.
4,363Total APs (campus, LAWA inventory 2026-05-12)
232Boingo-managed (end-of-life)
4,131LAWA-managed
1,116Wi-Fi 5 APs (upgrade backlog)
AP inventory by terminal (the as-is baseline). Upgrade path: TDIP capital-funded · RFP-Lite? candidate for the lighter next RFP · Boingo end-of-life, renewal-driven.
| Terminal/facility | APs | Manager | Majority generation | Wi-Fi upgrade path |
|---|---|---|---|---|
| TBIT (Bradley) | 782 | LAWA | Wi-Fi 6 (some 7) | TDIP |
| West Gates (MSC) | 655 | LAWA | Wi-Fi 6 (some 6E) | RFP-Lite? |
| Terminal 3 | 430 | LAWA | Wi-Fi 6 | RFP-Lite? |
| Terminal 1 | 342 | Boingo, LAWA | Wi-Fi 5 | Boingo |
| Terminal 4 | 316 | LAWA | Wi-Fi 6/6E (some 5) | TDIP |
| Terminal 6 | 206 | LAWA | Wi-Fi 5/6/6E | RFP-Lite? |
| Terminal 2 | 72 | LAWA | Wi-Fi 6 (some 5) | RFP-Lite? |
| Terminal 7 | 66 | Boingo, LAWA | Wi-Fi 5 | Boingo |
| Terminal 5 | 34 | LAWA | TBD (new build) | TDIP |
| Terminal 8 | 26 | Boingo | Wi-Fi 5 | Boingo |
| Terminal 4 (Regional) | 11 | Boingo | Wi-Fi 5 | Boingo |
Generation mix (campus): Wi-Fi 5 = 1,116 · Wi-Fi 6 = 2,865 · Wi-Fi 6E = 235 · Wi-Fi 7 = 147. The 232 Boingo APs (Terminals 1, 7, 8 and American Eagle) are the end-of-life Wi-Fi 5 estate the Boingo Wi-Fi 7 proposal targets.
Managed-Wi-Fi operational snapshot (Boingo Insights)
Boingo's own RADIUS/controller telemetry for the managed estate — a partial, managed-network view that complements the crowdsourced Ookla layer (and previews what full Cisco Catalyst data unlocks, §9):
100.0%Service availability (30-day)
9,370Current active clients
503.2 MbpsMax speed-test result
Other graph-sourced context: Vendors — Boingo Wireless, Cisco, HPE Aruba Networking, Verizon Wireless. Carriers/spectrum —
AT&T Mobility, Verizon Wireless, T-Mobile US, Charter (CBRS), FirstNet (Band 14). Boingo 7-year Wi-Fi agreement (BOAC award), ~1 yr remaining, ~14-mo upgrade; likely transition to LAWA-owned/managed post-upgrade. DAS master (Feb 2026): Verizon-led consortium, started 2018, expires 2028, still building 4G layers; open question on terminal-specific RFP coverage. Construction — T4, T5 (demolished/rebuild) and TBIT refresh in flight under TDIP, June 2028 LA28 deadline — Wi-Fi upgrades funded through those capital projects. All
terminals carry Passpoint (Boingo-managed carrier offload).
3. Wi-Fi coverage by terminal
The core view. We segment observed Wi-Fi usage, offload share, and passenger-dwell density to gate-proximal zones (each on-airport tile attributed to its nearest gate within 120 m; gate clusters roll up to terminals). Green/amber/red below is keyed to offload share — Wi-Fi-dominant (green) vs mobile-dominant (red).
▶ Open interactive coverage map (toggle Wi-Fi usage / offload / density; per-gate popups)
| Terminal | Gates | Tiles | Wi-Fi GB | Offload | Dwell med. | State |
|---|---|---|---|---|---|---|
| Terminal 2 | 13 | 1,776 | 1,653.3 | 36.7% | 16.5 | Mobile-dominant |
| Terminal 6 | 16 | 1,932 | 1,604.2 | 40.5% | 7.5 | Mobile-dominant |
| Terminal 3 | 14 | 1,810 | 1,592.8 | 44.5% | 6.0 | Mobile-dominant |
| Terminal 8 | 7 | 1,608 | 1,306.5 | 47.0% | 11.0 | Near parity |
| West Gates (MSC) | 23 | 3,456 | 1,810.8 | 47.2% | 4.0 | Near parity |
| Terminal 1 | 13 | 2,272 | 2,539.6 | 49.8% | 24.0 | Near parity |
| Terminal 7 | 13 | 1,704 | 1,931.5 | 55.7% | 11.0 | Wi-Fi-dominant |
| Terminal 5 TDIP-funded | 9 | 735 | 269.4 | 56.7% | 2.0 | Wi-Fi-dominant |
| TBIT (Bradley) TDIP-funded | 19 | 4,206 | 6,774.4 | 59.7% | 4.0 | Wi-Fi-dominant |
| Terminal 4 TDIP-funded | 11 | 2,068 | 2,435.1 | 60.1% | 8.0 | Wi-Fi-dominant |
Method note: gate-proximal tiles (~28% of on-airport tiles fall within
120 m of a gate). Terminal interiors, headhouse/check-in, apron and landside are outside this
radius and are not represented in the per-terminal rollup — gate hold-rooms are the dwell zones
the gap analysis targets. As-builts / Ekahau (pending) would anchor these to floor plans.
4. Noise & interference overlay
Co-channel interference and noise are exactly what crowdsourced usage data cannot show.
⏳ Awaiting LAWA input. Ookla Cell Analytics has no native Wi-Fi RSSI or noise/interference layer. A true noise/interference overlay needs Cisco Catalyst controller data (being escalated by Danielle). A WiGLE crowdsourced AP-signal pull (SSID / BSSID / best RSSI by location) is wired and ready in the pipeline — it activates the moment the WiGLE account email is verified, and will add an indicative AP-signal layer ahead of controller data.
Until then, treat any apparent “gap” as a usage/offload signal, not a
proven RF coverage hole. Firm interference conclusions wait on §9.
5. Gap identification, prioritized
Gaps ranked by passenger dwell × (1 − offload share): most passengers where Wi-Fi is under-absorbing demand. Capital-funded terminals are flagged so the roadmap doesn't double-count.
Strongest discretionary cases (low offload + high dwell + legacy Wi-Fi 5,
and NOT inside a TDIP capital project — so a real spend decision, not already funded):
Terminal 1 (Boingo, 46% Wi-Fi 5, offload 49.8%, user-density 24.0); Terminal 8 (Boingo, 100% Wi-Fi 5, offload 47.0%, user-density 11.0).
These line up with the Boingo end-of-life renewal: Terminals 1, 7, 8 (and American Eagle) are Boingo-managed Wi-Fi 5 — the 232-AP estate behind the pending Wi-Fi 7 proposal. Where those same terminals show high dwell and Wi-Fi under-absorbing demand, the renewal is the lever.
Terminal 1 (Boingo, 46% Wi-Fi 5, offload 49.8%, user-density 24.0); Terminal 8 (Boingo, 100% Wi-Fi 5, offload 47.0%, user-density 11.0).
These line up with the Boingo end-of-life renewal: Terminals 1, 7, 8 (and American Eagle) are Boingo-managed Wi-Fi 5 — the 232-AP estate behind the pending Wi-Fi 7 proposal. Where those same terminals show high dwell and Wi-Fi under-absorbing demand, the renewal is the lever.
Do not double-count: T4, T5 and TBIT are TDIP capital projects (Q2 2028) —
their Wi-Fi is funded through that work. Correction from the AP inventory: West Gates (MSC)
is on the lighter RFP-Lite path, not TDIP — so it stays a discretionary candidate.
⏳ Awaiting LAWA input. As-builts / Ekahau floor plans will anchor these gate-proximal findings to physical zones; TDIP master schedule confirms exact funded scope and dates.
6. DAS & cellular context
The augmentation layer — does cellular already carry the use case where Wi-Fi looks thin? From the completed LAX Verizon-offload analysis (Ookla Building KPIs, Apr 2026):
- Verizon 5G beats Wi-Fi on download by ~2.5× (199 vs 81 Mbps), but Wi-Fi wins upload ~4.4× (100 vs 22 Mbps) and latency ~5.2× (7.5 vs 39 ms) — the two networks are complementary, not substitutable. Wi-Fi carries ~51% of total observed data, consistent with the 54% airport-wide offload measured here.
- If Passpoint offload were disabled, degradation concentrates in ~12 of 25 buildings (48%) where Verizon cellular cannot absorb the load; one flagship building sits at −98 dBm 5G RSRP.
- DAS master: DAS master (Feb 2026): Verizon-led consortium, started 2018, expires 2028, still building 4G layers; open question on terminal-specific RFP coverage.
Implication for the roadmap: where strong 5G/DAS already serves the use case, a
saturated Wi-Fi upgrade is not justified now. Where Wi-Fi uniquely carries upload/latency-sensitive
traffic (most interactive passenger apps), it is not a candidate for de-investment.
7. ASQ correlation
Tie measured coverage to passenger satisfaction by terminal to support the spend narrative.
⏳ Awaiting LAWA input. ASQ scores by terminal and the LAX 'Performance by Terminal / Wi-Fi Service Quality' images (Q4'25 vs Q3'25 deltas) are pending from Danielle / available in the Imagine graph. Once in hand, each terminal's gap score (§5) overlays against its ASQ trend — low offload + high dwell + falling ASQ is the strongest spend case.
8. How this analysis works — method, definitions, provenance
Primary quantitative input is Ookla Cell Analytics aggregate tiles for the LAX campus, a single 12-month window (June 2025 – May 2026). The sample is crowdsourced and Android-weighted — roughly 3% of the airport's Android handset population; iOS is largely excluded. Tiles carry observed Wi-Fi and mobile DATA USAGE and a user-density count per ~100 m tile; the feed is not signal strength, capacity, or interference, and it cannot separate LAWA-managed Wi-Fi from carrier/Passpoint Wi-Fi on the same device.
Pipeline — every number in this report is produced by these five steps:
- Clip — Raw Ookla GeoJSON tiles are clipped to the KLAX aerodrome polygon (OSM way 649268639) plus a 200 m buffer, retaining 79,545 on-airport tiles and discarding the ~40% metro spillover.
- Segment — Each tile is assigned to a terminal. Segmentation v2 (this revision): if the tile centroid falls inside a real OSM terminal building footprint (point-in-polygon, pinned snapshot), it is assigned to that terminal; tiles outside every footprint fall back to the nearest terminal's 350 m gate-envelope; the rest are airfield/landside. A tighter gate-proximal cut (nearest gate within 120 m) isolates hold-room zones.
- Aggregate — Per terminal (both cuts): Wi-Fi usage, mobile usage, offload share, and the user-density median.
- Join — 342 Google Places venues are joined to tiles within 80 m of the venue point, clipped so tenant and tile in different building footprints never join; LAWA's AP inventory (counts, generation, 6 GHz) is joined per terminal; Boingo Insights telemetry and the LAWA ASQ tracker are layered as transcribed/cited context.
- Score — Each terminal receives a composite saturation-risk flag from the 4-point rubric below; gaps are ranked by the gap score.
Metric definitions (with units):
| Metric | Definition | Units | Source |
|---|---|---|---|
| Wi-Fi offload share | Σ WiFi_Data_Usage ÷ (Σ WiFi_Data_Usage + Σ Mobile_Data_Usage) × 100 over the zone's tiles | % of data-usage volume (not subscribers/connections) | build_lax_building_summary.py / build_lax_wifi_terminal_summary.py |
| User-density index | median of Ookla User_Density across the zone's tiles | concurrent unique users per ~100 m tile (dimensionless count — NOT minutes, NOT dwell time) | build_lax_building_summary.py |
| Gap score | user_density_median × (100 − offload_share_pct) ÷ 100 | dimensionless composite (many concurrent users where Wi-Fi absorbs least demand) | validate_lax_wifi_report.py |
| Saturation risk | +1 each: offload < 50 · user-density ≥ 6 · ≥1 AP in Boingo peak-load top list · Wi-Fi 5 share ≥ 40%. HIGH ≥ 3 · MEDIUM = 2 · low ≤ 1 | ordinal index (HIGH/MEDIUM/low) — a prioritization signal, not a measured channel utilization | build_lax_signal_layer.py |
| Session duration (Boingo) | median session length from Boingo Insights (transcribed) | minutes — the only genuine time metric in this analysis | lax_telemetry.json (Tier B) |
Thresholds, filters and windows (all disclosed):
| Parameter | Value | Role |
|---|---|---|
| Airport clip boundary | OSM aerodrome polygon + 200 m | tile inclusion |
| Terminal segmentation | OSM building footprint (point-in-polygon); 350 m gate-envelope fallback | full-zone terminal assignment (segmentation v2) |
| Gate-proximal radius | nearest gate within 120 m | hold-room cut |
| Gate table filter | ≥ 8 tiles per gate | suppress sparse gates |
| Tenant join | tiles within 80 m of venue point, same-footprint clip | venue attribution |
| Risk cut-points | offload<50 · density≥6 · peak-load≥1 · Wi-Fi5≥40%; HIGH≥3 | saturation rubric |
| Analysis window | Jun 2025 – May 2026 (single 12-month aggregate) | all Ookla metrics |
Evidence provenance tiers:
- A — Reproducible — Computed by us from data we hold; reproduce.sh regenerates it and the validator independently recomputes it (offload, density, gap, risk).
- B — Transcribed — From a source we hold but not machine-readable — LAWA AP-inventory xlsx, Boingo Insights screenshots (captured 2026-06-25). Values pinned and drift-guarded; originals retained in source/.
- C — Named source, not independently verified — Named findings we cannot produce: LAWA's Cisco Catalyst '>85% channel utilization' result (the saturation linchpin) and the LAWA ASQ tracker. Stated as LAWA's findings, corroborated — never presented as our measurement.
Uncertainty & inference limits:
- Point estimates from a ~3% Android-weighted crowdsourced sample; no confidence intervals are published because the sample cannot support them. Read figures as indicative.
- Terminal-level is the reliable inference unit; gate- and venue-level figures are directional.
- Granularity limits: Google Places returns venue POINTS (no footprints or floor level); the ~100 m Ookla tile is the resolution floor; there is no Z-axis, so multi-level terminals collapse into one plane. Adjacent venues can share tiles even after the same-footprint clip. LAWA GIS/BIM floorplans would unlock room-level attribution and are a named acquisition item.
- The 6 GHz recommendation's binding-constraint evidence (>85% channel utilization) is Tier C — LAWA's Catalyst finding, cited through the Imagine knowledge graph; we do not hold the export.
What is indicative vs definitive, and what this data cannot claim:
- Crowdsourced and Android-weighted — roughly 3% of the airport Android handset population; iOS largely excluded.
- Passive collection via partner apps (e.g. weather.com, AccuWeather) plus active speed tests.
- Data aggregates into three-month bins; event-level timestamps sometimes lost on export.
- Feed captures both Wi-Fi and cellular; we filter to Wi-Fi, but the holistic view is more honest.
- Ookla cannot reliably separate LAWA-managed Wi-Fi from MNO, roaming, or Passpoint-offload Wi-Fi on the same device.
- Signal strength / usage alone does not prove usable capacity or throughput — capacity and interference need noise and controller (Cisco Catalyst) data.
- Wi-Fi crowdsourced density is thinner than cellular, so the Wi-Fi view is coarser.
- This is an indicative network assessment, not an AP-placement remediation design.
Audit trail: claims ledger, provenance manifest, segmentation delta and
reproduction instructions in
LAX_Wifi/build/audit/; verification bundle guarantees
same inputs in → same numbers out. This section renders from lax_methodology.py,
the same single source as the docx deliverables and audit/METHODOLOGY.md.9. Next layer
What lifts this from indicative to definitive:
- Cisco Catalyst controller data — true AP-level health, client counts, channel utilization, and real co-channel interference. Converts §3/§4 from usage proxy to measured RF.
- WiGLE AP-signal pull — crowdsourced SSID/BSSID/RSSI, an interim signal layer (wired, pending email verification).
- As-builts / Ekahau + AP counts + ASQ — anchor heat to floor plans, baseline the as-is estate, and correlate to satisfaction.
Deeper multi-dimensional analysis (Platinum) and the full roadmap with
funding strategy (Diamond) remain separate, standalone-priced engagements — named here,
not folded into this interim deliverable.