Over the passenger-building footprints, the three carriers are close. Of the 9 buildings, 3 are ties. AT&T leads 4, T-Mobile leads Concourse A, and Verizon leads the International Terminal.

This report names a leader only from readings over the buildings. The data is a year of signal readings from Android phones, grouped into tiles: patches of ground about 9 m by 11 m. A tile is healthy when its 5G signal is at or above −95 dBm. The following table shows, for each building, the share of healthy tiles and the median 5G reading.

Each cell: share of tiles with healthy 5G (at or above −95 dBm) · median 5G RSRP in dBm · n = tiles over the footprint. Leader rule (section 4.7): a carrier leads a building only when it is alone at the top on both measures. It must top the healthy share by more than 1 point and the median by more than 1 dB. Otherwise every carrier within those margins of the top on either measure is tied. The rule runs on the unrounded values.

The international side is the weakest place measured. Each carrier's single lowest building is one of two: the International Terminal for AT&T (49%) and T-Mobile (27%), and Concourse F for Verizon (36%). Every carrier is healthy on more than half the tiles over 7 of the 9 buildings.

1. At the gate areas, only AT&T clears the healthy 5G line

   AT&T is healthy at 133 of 201 gate points, on the smallest 5G footprint of the three (5G on 40% of its tiles). T-Mobile (31) and Verizon (12) sit below the line. The two are tied at the median only: Verizon is the weakest carrier by more than 1 dB at 88 gate points, T-Mobile at 10. 59 gate points are weak for all three carriers.

   A gate area is the ground within 80 m of a gate. At the median gate, only 22.4% of that ground lies over a building footprint. The rest is apron, the paved area where aircraft park. This report calls these readings apron-weighted: a gate-area number describes mostly the apron, not the seating area.

2. The airport-owned neutral-host DAS was replaced inside the window

   The airport owns the distributed antenna system (DAS), an in-building system built to be shared by several carriers, called a neutral host. A fact tagged held comes from a source document pinned in this build and checksummed in the verification bundle.

   The Atlanta City Council approved a replacement DAS on May 19, 2025. T-Mobile and AT&T signed, with a January 2026 target date held.

   Verizon's participation in the new system, the integrator, and the per-building cutover dates are not held. This report has no source document for them.

3. What this issue can and cannot evaluate

   This issue is a baseline. Its readings mix the old and new systems, so it cannot evaluate the new system on its own. No capital recommendation follows from this data alone (section 13).

Photo · Domestic Terminal ticketing and check-in hall, May 2023Wikimedia Commons

Domestic Terminal ticketing and check-in hall, with the Main and North Checkpoint signage, May 2023. Photo: Harrison Keely, CC BY 4.0, via Wikimedia Commons.

Each carrier's building readings come first, and its gate-area readings second. Buildings decide the leader. Gate areas are reported as counts and shares, never as a lead.

The following table lists the building verdicts and their margins (section 4.7). Every ranked cell clears the 30-tile floor.

Healthy shares and medians per building are in the Executive summary table. Leader rule (section 4.7): a carrier leads a building only when it is alone at the top on both measures. It must top the healthy share by more than 1 point and the median by more than 1 dB. Otherwise every carrier within those margins of the top on either measure is tied. The rule runs on the unrounded values. Concourse B is a three-way tie because the top share and the top median belong to different carriers.

Strongest and weakest counts: 52 gate points tie at the top and 103 tie at the bottom. The 5G footprint mixes deployment with the panel's handset mix, so it is not compared across carriers (section 16).

AT&T leads over 4 of 9 buildings. Its lowest building is the International Terminal, at 49% of tiles healthy.

At the gate areas it is the only carrier above the healthy line. It is 3.3 dB ahead of T-Mobile and 4.2 dB ahead of Verizon, both above the 1 dB noise floor. It reports 5G on only 40% of its tiles, and 17% of its tiles are 5G present and healthy, against T-Mobile 34% and Verizon 27% (section 3.0). Its LTE layer is the strongest of the three.

T-Mobile leads over Concourse A and ties over the Domestic Terminal and Concourse B. Its lowest building is the International Terminal, at 27%.

At the gate areas it is 1.7 dB below the healthy line. Its median is 0.9 dB from Verizon's, inside the noise floor, so the two are tied at the median. Per gate point, Verizon is the weakest by more than 1 dB at 88 and T-Mobile at 10. 8 of its 11 strongest gate points are in Concourse A. Its SINR median is the lowest of the three. Its LTE median (−95.8 dBm) is below the line and 0.6 dB from Verizon's, a tie.

Verizon leads over the International Terminal and ties over the Domestic Terminal, Concourse T and Concourse B. Its lowest building is Concourse F, at 36%.

At the gate areas it is 2.6 dB below the line and healthy at the fewest points. Its LTE median (−96.4 dBm) is below the line, tied with T-Mobile's. Its 5G is NR3700 on 63% of its 9,964 band-reporting 5G tiles. NR3700 is Ookla's label for 5G on spectrum near 3700 MHz, the mid-band called C-band (section 9.1).

Each cell: gate-area 5G median in dBm · healthy 5G gate points (share).

Hartsfield-Jackson is operated by the City of Atlanta Department of Aviation held. Two terminal complexes are in service.

This report measures 201 gate points, 284 mapped venue points, 44,634 unique tiles (85,763 carrier-tile observations, one per carrier per tile), 44,155 usage tiles and 371 estimated site centroids over the 12 months ending 2026-08-31.

This issue cannot evaluate the replacement DAS. The export is a single 12-month aggregate with no date split, and the per-building cutover dates are not held. The date-range comparison in section 15 is the instrument for that question.

ATL is two terminal complexes, seven concourses and nine passenger buildings. This section lists what was measured and what the public record holds.

Photo · Aerial view over the airfield, May 2026Wikimedia Commons

Aerial view over the airfield, May 2026. Illustrative only. Individual concourses are not distinguishable in this frame. Photo: Zixed11, CC0, via Wikimedia Commons.

Fig. 1 · Venue map · aerodrome, nine OSM passenger buildings, 201 gate pointsSource · Ookla

FIG. 1. Venue map: aerodrome outline, the nine OpenStreetMap passenger buildings, 201 gate points with concourse letters, and the on-airport tile centroids (85,763 carrier-tile observations). North up, scale bar in meters. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Photo · Plane Train in-car display, December 2023Wikimedia Commons

Plane Train in-car display with the concourse sequence F, E, D, C, B, A, T, December 2023. The train runs below the concourses through the transportation mall, which the airport-owned DAS has covered since 2010. Photo: DiscoA340, CC0, via Wikimedia Commons.

The analysis uses 201 measured points. The airport fact sheet publishes 198 gates (155 domestic, 43 international) with no per-concourse count held.

A per-concourse list is held only for Concourse F, from the official 2012 map (F1 to F14). There, 2 published gates have no measured point, and every measured point is on the list.

At the gate points, AT&T is the only carrier above the healthy 5G line. T-Mobile and Verizon are tied below it at the median only. Per gate point, Verizon is the weakest by more than 1 dB at 88 and T-Mobile at 10.

Healthy = gate mean at or above −95 dBm. SINR healthy gate points: AT&T 197, T-Mobile 76, Verizon 132 of 201. LTE healthy gate points: AT&T 182, T-Mobile 68, Verizon 61. Strongest LTE reading: AT&T 174, T-Mobile 4, Verizon 0 gate points.

Over all on-airport tiles the picture changes. Each 5G median below counts only tiles that report 5G. The last column combines presence and level.

Fig. 2 · Venue medians across 201 gate points · locked threshold ladderSource · Ookla

FIG. 2. Venue medians across 201 gate points against the locked threshold ladder (section 16). The two received-power panels draw bars from −115 dBm, so a longer bar is a stronger signal. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Fig. 3 · SINR distribution per carrier at gate-point levelSource · Ookla

FIG. 3. SINR distribution per carrier at gate-point level. Read across carriers with care. SINR depends on the serving band's bandwidth and on neighbor loading. These carriers sit on different bands here (section 9.1 gives each carrier's strongest-serving band mix). So a cross-carrier SINR gap is not a like-for-like comparison. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Fig. 4 · 5G NR RSRP by tile and carrier · locked coverage classSource · Ookla

FIG. 4. Terminal and concourse area map of 5G NR RSRP by tile and carrier, colored by the locked coverage class. Grey tiles carry no 5G reading. Each tile is about 9 m by 11 m, and n is the tile count per panel. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Zones include apron, taxiway and roadway as well as building, so this table is not a leader board. Building shares are in the Executive summary table. That is why Verizon's highest zone is Concourse F (57%) while its lowest building is Concourse F (36%): the zone adds the apron around the pier.

AT&T is first at the 5G gate median, 3.3 dB ahead of T-Mobile, and the only carrier above the line. It is healthy at 133 of 201 gate points, at risk at 68, critical at 0.

It is the strongest 5G reading by more than 1 dB at 138 of 201 gate points. That includes every gate point in Concourses C and D (counts).

52 of the 201 gate points are ties within the 1 dB floor.

Its 5G reports on 40% of its on-airport tiles (10,316 of 25,823). Tiles with LTE but no 5G are 58% of its tiles. Its 5G is NR3700 on 47% of its 2,223 band-reporting 5G tiles. NR3700 is Ookla's label for 5G near 3700 MHz, the C-band mid-band.

Its zone shares run from 59% in Concourse C to 36% in Concourse F. FIG. 4 shows green along the piers from Concourse B eastward and amber over Concourse F and the International Terminal. Its terminal-vs-periphery proxy is −1.0 dB, inside the 1 dB noise floor, so it reads as no difference (section 5.2).

AT&T is the weakest carrier by more than 1 dB at 0 of 201 gate points. 103 points tie at the bottom.

T-Mobile's 5G gate median is −96.7 dBm, below the healthy line. It is tied with Verizon at the median (0.9 dB apart) and 3.3 dB behind AT&T. It is healthy at 31 of 201 gate points, at risk at 170, critical at 0.

It reports 5G on 84% of its on-airport tiles (27,860 of 32,992), the widest footprint. Its 5G is NR2500 on 55% of its 26,021 band-reporting 5G tiles. That mid-band layer reads weaker than sub-1 GHz at the same distance.

Its LTE median is −95.8 dBm, below the line and tied with Verizon's (0.6 dB apart). It is healthy on LTE at 68 of 201 gate points and strongest at 4. Its zone shares run from 42% in Concourse F to 29% in Concourses D and E, a tie within 1 point. It leads over Concourse A (section 4.7). Its terminal-vs-periphery proxy is +1.0 dB, inside the noise floor, so it reads as no difference (section 5.2).

T-Mobile is the weakest carrier by more than 1 dB at 10 of 201 gate points.

Verizon's 5G gate median is −97.6 dBm, tied with T-Mobile at the median and 4.2 dB behind AT&T. It is healthy at 12 of 201 gate points, the fewest of the three, at risk at 189, critical at 0.

It reports 5G on 70% of its on-airport tiles (18,800 of 26,948). Its 5G is NR3700 on 63% of its 9,964 band-reporting 5G tiles. Its LTE median is −96.4 dBm, below the line and tied with T-Mobile's. It is healthy on LTE at 61 of 201 gate points.

Its zone shares run from 57% in Concourse F to 25% in Concourse D. It leads over the International Terminal (section 4.7). Its terminal-vs-periphery proxy is +2.0 dB (section 5.2).

Verizon is the weakest carrier by more than 1 dB at 88 of 201 gate points, the most of the three.

Fig. 5 · LTE percent-low-band per carrier · mean, share above zero, p90Source · Ookla

FIG. 5. LTE percent-low-band per carrier: mean, share of tiles above zero, and p90. Higher means more attachment to sub-1 GHz coverage bands. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Percent-low-band describes LTE attachment to sub-1 GHz, a coverage-grade band. Propagation, idle-mode priorities and site geometry all drive it.

Section 9.1 lists the bands.

Fig. 6 · 5G NR and LTE RSRQ distributions per carrierSource · Ookla

FIG. 6. 5G NR and LTE reference signal received quality (RSRQ) distributions per carrier. The dotted line is a GENERIC industry reference, not a threshold specified for this venue. Medians in this range are ordinary for a loaded network, and this is not a failure call. Read with SINR. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Fig. 7 · LTE CQI distribution and Ookla LTE optimization prioritySource · Ookla

FIG. 7. LTE channel quality indicator (CQI) distribution and Ookla's LTE optimization priority index (vendor layer). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Every carrier's RSRQ median sits below the -10 dB line on both technologies. Medians of −11 to −14 dB are ordinary on a loaded network, and the line is a generic reference, so this is not a failure call.

RSRQ degrades as load rises, so weak RSRQ on strong RSRP is consistent with a loaded serving layer. That is an inference. The export carries no load field (section 16).

CQI is the closest throughput proxy in this export. Ookla's optimization priority flags tiles where signal is strong but quality is poor. The medians are close. The signal is in the distribution tails.

Fig. 8 · Serving-cell volatility per carrier, LTE and 5GSource · Ookla

FIG. 8. Share of tiles where the strongest observed cell differs from the most-frequent serving cell. Proxy for handover instability and dominance conflict. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Where the strongest cell is not the most-frequent serving cell, the network is handing over between competing cells.

Cell fields populate only where Ookla resolved a serving cell, so these n are smaller than the RSRP n (section 16). Churn blends macro-sector overlap with any in-building system's hand-off to the macro layer. The export carries no handover trace.

This section compares the three carriers concourse by concourse. The zone figures show the spread within each concourse area. Leaders are named only in section 4.7, over the building footprints.

Gate points per concourse set how gate results are shown. Four concourses clear the 30-point floor, so their gate shares appear as percentages. The other three appear as counts only and are never ranked.

Tile basis; n = tiles per carrier per zone. Healthy = tile 5G RSRP ≥ −95 dBm.

The carrier-to-carrier gap in healthy tile share is widest in Concourse D, at 30 points. It is narrowest in Concourse A, at 17 points. Each carrier's lowest zone is different:

A zone and a building can point in opposite directions. Verizon's Concourse F zone is its best zone (57% of 1,895 tiles healthy), yet the Concourse F building is its lowest building (36% of 166 tiles). The zone pool is mostly apron, taxiway and roadway outside the pier, so section 4.7 governs any reading of the building.

Fig. 9 · Healthy 5G tile share by zone and carrierSource · Ookla

FIG. 9. Share of zone tiles with healthy 5G RSRP by carrier (tile basis; minimum n per carrier on the axis). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Fig. 10 · Median 5G RSRP by zone and carrierSource · Ookla

FIG. 10. Per-zone median 5G RSRP per carrier (tile basis). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Fig. 11 · LTE RSRP by tile and carrier · terminal and concourse areaSource · Ookla

FIG. 11. Terminal and concourse area map of LTE RSRP by tile and carrier, colored by the locked coverage class. Grey tiles carry no LTE reading. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

No gate point is critical for any carrier. Critical means 5G RSRP below −110 dBm, and the count is 0 of 201 for AT&T, T-Mobile and Verizon. On the tile basis, a small critical share shows up in one zone per carrier:

Fig. 12 · Coverage-class distribution by concourse zone and carrierSource · Ookla

FIG. 12. Coverage-class distribution by concourse zone and carrier (tile basis; A = AT&T, T = T-Mobile, V = Verizon). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

PCI diversity counts the distinct cells (physical cell identities) that serve a zone, macro or in-building. Many cells with healthy SINR means several sectors serve the zone cleanly. Many cells with poor SINR means they compete.

Fig. 13 · 5G PCI diversity per 1,000 cell-resolved tiles, by zone and carrierSource · Ookla

FIG. 13. Unique strongest-serving physical cell identities (PCIs) per 1,000 cell-resolved tiles, per zone per carrier, with the raw unique count in brackets. Normalized because a unique count rises with sample size. Reads in both directions; always pair with SINR (section 16). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

A gate that is weak for all three carriers points to the area around that gate, which at this venue is mostly apron. A gate that is weak for one carrier points to that carrier's network.

Most weak gates are weak for two carriers or for all three. At the gate areas, T-Mobile and Verizon are below the healthy line almost everywhere. AT&T is the carrier that keeps the all-weak count from being higher.

Over the building footprints the picture is much closer, and the leader changes from building to building (section 4.7). So this concordance reads as an apron condition, not a building condition.

Fig. 14 · Carriers below the healthy line per gate point, by concourse zoneSource · Ookla

FIG. 14. Per concourse zone: how many of the three carriers are below the healthy line at each gate point (gate counts). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

The Domestic Terminal is split into North and South halves, with the MARTA Airport Station between the two baggage-claim halls. It feeds Concourses T, A, B, C and D. The airport fact sheet publishes 155 domestic gates across the complex held.

Concourse T hosts American, United and Delta, with other domestic carriers held. The airport places the Admirals Club across from gate T10, the United Club at T11 and a Delta Sky Club near gate T6. Per-airline gate assignments are not held.

Concourse A hosts Delta Air Lines held, with Delta Sky Clubs at the Concourse A centerpoint and near gate A17.

Fig. 15 · Per-gate 5G RSRP · 50 Concourse T and A gate pointsSource · Ookla

FIG. 15. Per-gate 5G RSRP across the 50 Concourse T and A gate points (OpenStreetMap inventory). Each bar is a mean over the tiles within 80 m of the gate. Bars rise from −115 dBm, so a longer bar is a stronger signal. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Photo · Concourse T at gate T10Wikimedia Commons

Concourse T corridor at gate T10, December 2023. Photo: DiscoA340, CC BY-SA 4.0, via Wikimedia Commons.

Photo · Concourse A at gate A19Wikimedia Commons

Concourse A at gate A19, December 2023. Photo: DiscoA340, CC BY-SA 4.0, via Wikimedia Commons.

Concourse B hosts Delta Air Lines held, with a Delta Sky Club near gate B18.

Concourse C hosts Delta Connection and Southwest Airlines held, with a Delta Sky Club near gate C37.

Spirit Airlines held gates C4 and C6 until it ceased operations on May 2, 2026, inside this window. Delta bid for those gates in July 2026 held. Whether that assignment closed inside the window is not held.

Fig. 16 · Per-gate 5G RSRP · 66 Concourse B and C gate pointsSource · Ookla

FIG. 16. Per-gate 5G RSRP across the 66 Concourse B and C gate points (OpenStreetMap inventory). Each bar is a mean over the tiles within 80 m of the gate. Bars rise from −115 dBm, so a longer bar is a stronger signal. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Photo · Concourse B at gate B21Wikimedia Commons

Concourse B, boarding queue at gate B21, December 2023. Photo: DiscoA340, CC BY-SA 4.0, via Wikimedia Commons.

Photo · Concourse C at gate C33Wikimedia Commons

Concourse C at gate C33, December 2023. Photo: DiscoA340, CC BY-SA 4.0, via Wikimedia Commons.

Concourse D hosts Delta Air Lines held. The widening program's new gates were turned over to Delta, and a Delta Sky Club sits at the D centerpoint.

Fig. 17 · Per-gate 5G RSRP · 41 Concourse D gate pointsSource · Ookla

FIG. 17. Per-gate 5G RSRP across the 41 Concourse D gate points (OpenStreetMap inventory). Each bar is a mean over the tiles within 80 m of the gate. Bars rise from −115 dBm, so a longer bar is a stronger signal. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Photo · Concourse D gate D11 hold room, August 2024Wikimedia Commons

Concourse D gate D11 hold room, August 2024, during the widening program. Photo: Vmzp85, CC BY-SA 4.0, via Wikimedia Commons.

The Maynard H. Jackson Jr. International Terminal opened in 2012. Concourse F is attached to it, and Concourse E is reached through the E-F connector. The airport fact sheet publishes 43 international gates held.

Concourse E hosts Delta and international carriers held. It has a Delta Sky Club across from gate E15 and an American Express Centurion Lounge next to gate E11. Per-airline assignments are not held.

Concourse F hosts Delta long-haul and foreign-flag carriers held. The official 2012 map shows gates F1 to F14. A Delta Sky Club and The Club at ATL sit on the mezzanine level. The table adds the International Terminal building itself.

Fig. 18 · Per-gate 5G RSRP · 44 Concourse E and F gate pointsSource · Ookla

FIG. 18. Per-gate 5G RSRP across the 44 Concourse E and F gate points (OpenStreetMap inventory). Each bar is a mean over the tiles within 80 m of the gate. Bars rise from −115 dBm, so a longer bar is a stronger signal. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Photo · Concourse E, gates E1 to E12Wikimedia Commons

Concourse E, gates E1 to E12 signage, December 2023. Photo: DiscoA340, CC BY-SA 4.0, via Wikimedia Commons.

Photo · Concourse F central atriumWikimedia Commons

Concourse F central atrium, May 2023. The International Terminal building is the lowest building reading for T-Mobile and AT&T, and Concourse F is the lowest for Verizon (section 4.7). Photo: Harrison Keely, CC BY 4.0, via Wikimedia Commons.

This is the only basis on which the report names a leader. Over the nine passenger buildings, the three carriers are far closer than at the gate areas.

Three buildings are ties: the Domestic Terminal, Concourse T and Concourse B. AT&T leads over Concourses C, D, E and F. T-Mobile leads over Concourse A, and Verizon over the International Terminal.

Fig. 19 · 5G NR class for tiles over the nine OSM building footprintsSource · Ookla · OSM

FIG. 19. Tiles whose centroid falls inside a pinned OSM passenger-building polygon, colored by 5G NR class, with the per-building per-carrier table. The polygons are the seven concourses, the Domestic Terminal, and the International Terminal. This is our OSM building-polygon mask, not Ookla's Indoor layer. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

5G over the building footprints. Healthy = tile 5G RSRP ≥ −95 dBm; median over the tiles whose centre falls within the polygon. n counts tiles, not scans.

LTE is the fall-back layer when 5G is weak. Over the same footprints it reads as follows.

LTE over the building footprints, the graceful-degradation floor on the leader basis. n counts tiles.

Why the gate areas cannot name a leader. The area within 80 m of each gate reaches the apron on both sides of each pier. Across the 201 gate points, a median of 22.4% of those tiles sit over a building. The highest at any gate is 68%. So a gate-area figure is mostly an outdoor apron reading.

At the gate areas, T-Mobile and Verizon sit below the line at nearly every gate. Over the buildings, the three are tied at three buildings and T-Mobile leads over Concourse A.

Read this as a walk, not a leader board. Every number is a crowdsourced tile aggregate, and the basis changes by stage.

Photo · MARTA Airport Station fare gatesWikimedia Commons

MARTA Airport Station fare gates at the west end of the Domestic Terminal, June 2024. This is the first place an arriving rail passenger's handset meets the airport network. Photo: ElToAn123, CC BY-SA 4.0, via Wikimedia Commons.

Photo · Domestic Terminal baggage claimWikimedia Commons

Domestic Terminal baggage claim, carousel 8, August 2024. Photo: Vmzp85, CC BY-SA 4.0, via Wikimedia Commons.

Fig. 20 · Healthy 5G share per carrier at each stage of the passenger journeySource · Ookla · OSM

FIG. 20. Healthy 5G share per carrier at each stage of a passenger's path. The stages are the Domestic Terminal building, Concourse T, the concourse centres at the Plane Train, the gate areas, and the International Terminal. The basis changes by stage (building mask, 80 m pools around the concourse centres, per-gate means) and is named in the table. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31; OpenStreetMap footprints.

n counts tiles for building and centre rows and gate points for gate-area rows.

Check-in, security and baggage claim are even. In the Domestic Terminal building the three carriers are within 2 points of each other. Every median sits at or near the healthy line, so the experience is middling but similar whatever the handset.

Over the domestic pier footprints, every carrier is above half. The Plane Train delivers passengers to a concourse centre (the journey table in this section). Over the Concourse A to D buildings, the lowest share for any carrier is 60%.

The apron-weighted gate areas are where the carriers separate. Apron-weighted means most of the 80 m pool around each gate is apron, not building. At the A to D gate areas, AT&T is healthy at 74% of gate points, T-Mobile at 21% and Verizon at 7%. So on the apron side of these gates, T-Mobile and Verizon read below the line on 5G more often than not, and AT&T does not. The data does not reach the jet bridge itself.

The international side holds each carrier's lowest building. AT&T and T-Mobile read lowest over the International Terminal (49% and 27%), and Verizon over Concourse F (36%). Verizon is also the one carrier above half over the International Terminal building (61%). At the E and F gate areas, T-Mobile and Verizon are healthy at almost no gate points.

LTE holds over the pier footprints and thins on the apron. Over the building footprints, LTE is healthy on 76 to 88% of tiles for AT&T, 33 to 72% for T-Mobile and 32 to 83% for Verizon. At the gate points, the LTE medians are AT&T −92.4, T-Mobile −95.8 and Verizon −96.4 dBm. Section 9.2 gives service availability.

Tenant readings use 284 mapped points. They come from a Google Places and OpenStreetMap capture pinned 2026-09-21.

Most categories hold 30 points or fewer, so comparisons do not run on points. They run on the tiles within 80 m of each category's points, deduplicated. Point counts appear on every axis.

Seven categories clear the 30-tile floor for every carrier: transport_admin, retail, food, lounge, services, admin and other. The spread between carriers differs by category.

This report does not publish a category ordering. Read the charts as a pattern.

Fig. 21 · Median 5G RSRP by tenant categorySource · Ookla

FIG. 21. Median 5G RSRP by tenant category (tile basis; points and minimum tiles per carrier on the axis). Bars rise from −115 dBm, so a longer bar is a stronger signal. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Fig. 22 · Median 5G SINR by tenant categorySource · Ookla

FIG. 22. Median 5G SINR by tenant category (tile basis). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

This proxy compares tenants over the building footprints with tenants outside them. Terminal categories are the concessions, lounges and services in the concourse and terminal footprints. Periphery categories are the parking decks and lots, the hotel and the fire stations.

Periphery minus terminal median 5G RSRP, tile basis.

A large positive delta is consistent with macro signal spilling in from outside. A negative delta, where the terminal reads stronger, is consistent with an in-building source serving the concourses.

Two of the three deltas sit within the 1 dB noise floor. AT&T at −1.0 dB and T-Mobile at +1.0 dB are ties, so neither is evidence of an in-building source. Only Verizon's +2.0 dB clears the floor, and it is consistent with macro spill from outside.

Distance to macro sites confounds all three readings. Section 9.3 places the estimated macro sites.

Fig. 23 · Median 5G RSRP · terminal vs periphery tenant categoriesSource · Ookla

FIG. 23. Per-carrier median 5G RSRP, terminal tenant categories vs periphery categories, on the tile basis. Proxy: category-based masking, not Ookla's building-polygon indoor mask (queued, section 15). Bars rise from −115 dBm, so a longer bar is a stronger signal. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

An airline sees its exposure through its concourse and its passengers' carriers. The airline column is the set the airport's published concourse guide places in each concourse.

Tile basis; n = tiles per carrier per concourse. Airline sets transcribed from the airport's published concourse guide.

Brands with several locations are a consistency check across zones, not a comparison. Each has 6 points at most, so no carrier ordering is read from this table.

30 of the 45 cells sit below the healthy line. Every cell carries its coverage class, so a sub-healthy value cannot pass unmarked. Each point value is a mean over the tiles within 80 m of the point. Each cell is the median of those point values across the brand's points, and n sums the tiles behind them.

Count-only view (fewer than 30 points per brand); never ranked. n = tiles. Per-point tile n in Appendix A of the report.

The capture holds 16 lounge points. A hygiene pass refiled the airline and card-issuer lounges that the capture had listed as food or restaurant.

Each lounge is a single point, reported for scope and not ranked. Names are printed as the capture spells them.

Counts of single lounge points, not a ranking.

Single-point aggregates over the tiles within 80 m of each point; n = tiles. Reported for scope, not ranked.

A lounge is where a passenger sits for an hour with a laptop. Section 5.6 reads these results together with the Wi-Fi figures.

A concessionaire's question is not which carrier wins. It is whether a customer at the counter, and the payment terminal behind it, has usable signal on whatever network they use.

Concessions sit at the concourse centres and along the pier interiors. The table reads the tiles within 80 m of the food and retail points in each concourse.

Fig. 24 · Healthy 5G share over the concession areas per concourse and carrierSource · Ookla · OSM

FIG. 24. Share of tiles with healthy 5G within 80 m of the food and retail points in each concourse, per carrier. The axis shows the point count per concourse. The pools sit over the concourse centres and the pier interiors, where the concessions are. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31; OpenStreetMap venue points.

Tile basis behind the food and retail points; n = tiles per carrier. Density is the per-tile median of User_Density (unique devices per tile over the window). The device-tile sum counts a device once per tile it crosses, so it is not used to rank.

Concourse T carries the densest concession floor, at a median of 20 devices per tile. Concourse D is where the weakest carrier's share over the concessions is lowest, at 31%.

Three practical readings follow. Each is an inference from crowdsourced tiles, not a site survey, and each uses the building basis of section 4.7.

1. On the international side, plan for Wi-Fi and LTE fallback

   Over the Concourse F footprint, T-Mobile and Verizon 5G is below the healthy line more often than not (38% and 36% of tiles healthy). Payment terminals, order-ahead pickups and loyalty apps on those networks will lean on Wi-Fi or on LTE.

   LTE over the same footprint is healthy on AT&T 80%, T-Mobile 49% and Verizon 32% of tiles. Over the International Terminal building, Verizon is the one carrier above half (61%) and T-Mobile the lowest (27%). There, the same caution applies to T-Mobile and AT&T customers.

   Over the domestic piers, every carrier is above half on the building basis. The offload share in each zone (section 7) shows how much traffic already rides Wi-Fi.

2. Walk-test the lounges first

   The lounges (section 5.5) are where passengers typically sit and work for a while. A lounge customer notices a bad hour in a way a passenger buying a coffee does not.

3. The Wi-Fi layer already exists at the concessions

   Section 7.2 shows Wi-Fi observed at every mapped point in every category. What is not held is which of it is the public network and which is tenant-owned. Controller telemetry (section 15) would answer that.

Photo · Domestic Terminal South departures curb, October 2025Wikimedia Commons

Domestic Terminal South upper-level departures curb, October 2025, inside the window. Photo: Mccunicano, CC BY-SA 4.0, via Wikimedia Commons. Illustrative, resized only.

Device density is this dataset's demand signal. This section reports where devices are seen and where they meet weak signal. It does not rank zones or set re-measurement order, which Recommendation 3 sets on the building basis.

Per-tile medians are low everywhere. The 95th percentile shows where crowds build. The on-airport median is 2 devices per tile across 44,155 usage tiles.

This measure puts weak signal and demand in one number. For each zone, it adds up User_Density over the tiles where every carrier reporting 5G is below −95.0 dBm.

Only tiles where all three carriers report are counted, and that total is shown as the denominator. The usage and radio exports share one tile grid and are joined tile to tile. A coarser join would take the worst reading within about 110 m and saturate.

Device-tile observations: a device counts once per tile it crosses, so the sum scales with the zone's tile count. It is not a count of unique devices and not a passenger count. The weakest-carrier column uses the zone basis. It names the carrier with the largest below-healthy share of the zone's tiles, and a margin within 1 point is a tie.

The zone basis and the building basis can disagree. In the Concourse F zone, AT&T has the lowest healthy share (36% of 1,107 tiles). Yet it leads over the Concourse F building by 17 points (section 4.7). The zone pool is mostly apron, taxiway and roadway, so it is not a building verdict.

Concourse T has the largest sum because its zone also holds the Domestic Terminal landside. That includes the hourly and daily decks, the economy lots and the MARTA station, so it carries the most usage tiles.

Fig. 25 · Device density by concourse zone · device-tile observations where no carrier is healthySource · Ookla

FIG. 25. Left: unique-device density by concourse zone (median and 95th percentile per tile; User_Density = unique devices per tile over the window, deduplicated within the tile). Right: device-tile observations on tiles where every carrier is below the healthy line. The count covers only tiles where all three carriers report 5G, and the chart shows that denominator. A sum over tiles counts a device once per tile it crosses. It scales with the zone's tile count and is not a count of unique devices. Source: Ookla Cell Analytics All Networks usage export, 2025-09-01 to 2026-08-31.

Usage volume shows where bandwidth is consumed, by bearer and by zone. It is sample volume from the crowdsource panel, not total traffic.

Fig. 26 · Observed data volume by concourse zone and bearerSource · Ookla

FIG. 26. Observed data volume by concourse zone and bearer (GB over the 12-month window). Source: Ookla Cell Analytics All Networks usage export, 2025-09-01 to 2026-08-31.

Wi-Fi is already present in every concourse. Wi-Fi usage is observed on 58.4% of the 34,322 on-airport usage tiles that report the Wi-Fi field.

This presence rate is a different figure from the offload share in section 7.1. A tile that does not report the field has no measurement. It is left out of the denominator and is not evidence of no Wi-Fi.

The augmentation opportunity is largest where the worst-served carrier is weakest and Wi-Fi presence is high. At this venue, that is every concourse.

Fig. 27 · Weakest-carrier weak share beside Wi-Fi presence, by concourse zoneSource · Ookla

FIG. 27. By concourse zone: the share of tiles where the WEAKEST carrier is below the healthy line, beside the Wi-Fi presence rate. Both bars point the same way, so a tall red bar with a tall purple bar is a zone where Wi-Fi is already carrying the gap. Zone n counts usage tiles that report the Wi-Fi field, so it is smaller than the density n in FIG. 25. Source: Ookla Cell Analytics All Networks usage export, 2025-09-01 to 2026-08-31.

In the concourse zones, 41.7% of observed data volume moves over Wi-Fi.

Offload share = Σ Wi-Fi MB ÷ Σ all MB on the All Networks panel.

The 200 m buffer around the aerodrome takes in the parking decks, the rental-car and hotel approaches, the cargo aprons and roadway. Those are not concourse tiles, so the concourse-zone figure is the headline.

Fig. 28 · Wi-Fi share of observed data volume by concourse zoneSource · Ookla

FIG. 28. Wi-Fi share of observed data volume by concourse zone, with the venue reference line. Source: Ookla Cell Analytics All Networks usage export, 2025-09-01 to 2026-08-31.

Wi-Fi is observed at every mapped point in every tenant category. Presence does not tell categories apart here. Volume does, and FIG. 29 shows it.

Fig. 29 · Median observed Wi-Fi data volume per tile by tenant categorySource · Ookla

FIG. 29. Median observed Wi-Fi data volume per tile behind each tenant category. The companion presence-rate chart is not drawn because presence is at or near 100% in every category; the rates and point counts are in the text. Source: Ookla Cell Analytics All Networks usage export, 2025-09-01 to 2026-08-31.

The Wi-Fi figures here come only from the Ookla usage layer. Controller telemetry (section 15) would turn presence into capacity.

Hartsfield-Jackson owns its in-building cellular system, and it was being replaced inside this window. The window therefore blends the end-of-life legacy system and the new one.

A distributed antenna system (DAS) carries the carriers' signals indoors over one shared infrastructure. At ATL it is airport-owned and neutral-host, meaning any carrier can ride it.

The reconciled Airport DAS Master (2026-07-28) in the Imagine knowledge graph describes the earlier system.

1. ATL's DAS is self-owned and neutral host

   The record marks DAS, self-owned, and neutral host as yes, and the operator as unknown.

2. The integrator of the 2018 system

   Johnson Controls integrated the 2018 system. Contract term 2018 to 2022 at year precision. A field note records a 50 million dollar project installed by Johnson Controls and SMS that was due to expire in 2022.

3. Later records name the upgrade but no vendor or dollar figure

   A neutral-host DAS deployment row dated 2025 has no vendor recorded. The airport's FY2025 annual comprehensive financial report names a Distributed Antenna System Upgrade as a major capital project, with no dollar figure. The City's FY2026 adopted budget book carries no individual DAS, Wi-Fi or cellular line.

The graph names the integrator of the 2018 system, not of the 2025 replacement. It narrows none of the open items below.

No public record of a Citizens Broadband Radio Service (CBRS), private LTE or private 5G network at the airport was located. Public-safety radio rides a separate layer of the airport-owned DAS, and the 2025 replacement upgrades that layer not independently verified.

The Ookla drop also held a separate five-tile export labelled with a CBRS operator. Five tiles is below the 30-tile floor, so it is kept in the audit dossier as a count and not surfaced here.

The public Ookla layer does not characterize private networks. AT&T's FirstNet Band 14 (LTE700PS) appears on 2,085 AT&T LTE tiles (section 9.1).

The following table is illustrative, not a finding. It lists typical large-hub workloads and the parties a private program would involve. Nothing in the data supports or selects any row.

The band a tile attaches to most often shows which spectrum each carrier puts to work at the venue. Sub-1 GHz attachment implies coverage-grade bands. Mid-band attachment implies strong penetration or an in-building source.

Fig. 30 · Strongest-serving band distribution per carrier, 5G NR and LTESource · Ookla

FIG. 30. Strongest-serving band distribution per carrier, 5G NR and LTE (n on each panel). Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

AT&T (n = 2,223 NR tiles, 24,687 LTE tiles)

T-Mobile (n = 26,021 NR tiles, 21,271 LTE tiles)

Verizon (n = 9,964 NR tiles, 23,342 LTE tiles)

Share = band tiles ÷ the carrier's band-reporting tile footprint (report Appendix B).

A device keeps a usable LTE layer almost everywhere on the campus. That holds even where the LTE level itself is below the healthy line at the piers (section 4.8).

Availability fields are not populated on every tile. The field-population table in section 16 lists every field on the same basis.

Site Finder is Ookla's estimate of where each serving node sits, computed from the same crowdsourced tiles. It is a layer of estimated positions, not a registry.

Fig. 31 · Ookla Site Finder LTE and 5G centroids over the terminal and concourse areaSource · Ookla

FIG. 31. Ookla Site Finder LTE (circles) and 5G (triangles) centroids over the terminal and concourse area, per carrier, with the passenger-building footprints. Centroids are Ookla's ESTIMATED site positions derived from crowdsourced RF, not carrier registry data. Source: Ookla Cell Analytics Site Finder LTE and 5G centroid workbooks (estimated site positions), pulled with the 2026-09-21 drop.

All three carriers show estimated nodes of both technologies inside the building footprints.

Estimated positions, not a registry. 8 centroids labelled with 2 networks other than the three carriers of record are counted in the audit dossier and not surfaced.

Estimated node counts differ by carrier. The export cannot say whether that reflects sectorization, remotes per carrier or panel size. It is recorded as an observation for Recommendation 1.

The export also cannot say what any node carries, such as which bands and at what power. It cannot say whether each carrier's 5G rides on the node or on the macro layer. It cannot show how the hand-off at the jet bridge behaves. Those are the open items in section 8.

The concourse with the most estimated 5G nodes is not by itself the concourse with the highest healthy share (section 4.7). That is a second reason this layer is read as context, not as a ranking.

At 20 gate points, a passenger's 5G experience depends most on which carrier they use. The spread between the strongest and weakest carrier there is 8 to 11 dB.

AT&T is the strongest carrier across those 20 gate points. Each gate point is the area within 80 m of the gate, so these are gate-area counts, not building results.

A gate counts for a carrier only when that carrier trails the next one by more than 1 dB. Anything closer counts as a tie.

Spread is the strongest minus the weakest carrier's 5G RSRP at the gate point. Only gate points where every carrier has at least 30 tiles are included.

These gate points do two jobs. They are the worst-case examples for the stakeholder narrative. They are also the validation points for post-deployment monitoring (section 14).

An in-building system that carries 5G for all three carriers would be expected to narrow these spreads.

Fig. 32 · Top 20 gate points by carrier-to-carrier 5G spreadSource · Ookla

FIG. 32. Top 20 gate points by best-minus-worst carrier 5G RSRP spread, among the 201 gate points where every carrier's per-gate n is at least 30 tiles. Bar color is the worst carrier at that gate. Gate refs follow the OpenStreetMap inventory. Source: Ookla Cell Analytics per-carrier RF tile export, 2025-09-01 to 2026-08-31.

Fig. 33 · Weakest-carrier 5G class per tile, with device densitySource · Ookla

FIG. 33. Top: each terminal-area tile colored by the worst carrier's 5G NR class where all three carriers report. Bottom: User_Density per usage tile (unique devices, deduplicated) on a log scale over the same extent. Source: Ookla Cell Analytics per-carrier RF tile export and All Networks usage export, 2025-09-01 to 2026-08-31.

FIG. 33 answers two questions an operator asks first: where is the weakest carrier weak, and where are the devices?

Amber covers the apron between the piers and both landside terminals. Green runs along the pier interiors. The density panel shows the concourse centres and the Domestic Terminal with the highest device counts per tile.

5 operational-critical points are identifiable by name in the pinned capture: three fire stations and two United States Postal Service points.

No other federal or operational-critical point is identifiable by name. The Customs and Border Protection hall inside the International Terminal is not a mapped point.

AT&T's FirstNet Band 14 (LTE700PS) appears on 2,085 AT&T LTE tiles (section 9.1). Public-safety coverage posture is a section 15 layer.

n = tiles within 80 m of the point. A class is assigned only where a carrier has at least 30 tiles. Smaller cells are shown as readings with n visible.

Photo · Maynard H. Jackson Jr. International Terminal check-in hall, August 2024Wikimedia Commons

Maynard H. Jackson Jr. International Terminal check-in hall, August 2024. Photo: Vmzp85, CC BY-SA 4.0, via Wikimedia Commons. Illustrative, resized only.

In this window, most gate points sit below the healthy 5G line for T-Mobile and Verizon, and about a third do for AT&T.

A gate point is the area within 80 m of the gate. Healthy means 5G reference signal received power (RSRP) at or above −95 dBm. The last column shows how many of those gate points sit within 1 dB of the line, where a small change would move them across it.

These counts are conditional on a tile reporting 5G at all. AT&T reports 5G on 40% of its tiles, T-Mobile on 84% and Verizon on 70% (section 3).

Per-gate passenger volumes are not held. Two annual totals are held, and they measure different things:

The two bases are not interchangeable. Only the FAA enplanement figure belongs in any translation.

Seven asks, ordered by dependency. Each is a record, a re-measurement or a walk test.

1. Obtain the in-building record the public sources do not hold

   Ask for these items:

   The integrator and equipment vendor of the replacement system. The City Council file for the May 19, 2025 carrier agreements.

   Verizon's participation in the new system. Verizon is not named in the 2025 release. It was on the 2010 system.

   The actual per-building cutover dates against the January 2026 target, and which buildings were live on the new system during this window.

   The International Terminal and Concourse F cellular architecture. The airport's technology leadership described it in November 2025 as different from the rest of the campus.

   For each of AT&T, T-Mobile, and Verizon: which bands (LTE, 5G NR) ride on the in-building system in each of the nine passenger buildings. The record also needs the power levels and the hand-off setup to the macro layer at the jet bridge.

   Why it comes first: two carriers sit below the healthy line at nearly every gate area. Site Finder places their nodes inside the piers (section 9.3). More than one explanation fits that pattern. The record decides which.

   Owner · City of Atlanta Department of Aviation with the in-building system operator and the three carriers

2. Adopt this window as the baseline and re-measure per building

   Section 14 lists the per-carrier, per-building criteria. Start with the 5 lowest carrier-building cells over the building footprints (section 4.7):

   Carrier	Building	Healthy 5G	Tiles (n)
   T-Mobile	International Terminal	27%	60
   Verizon	Concourse F	36%	166
   T-Mobile	Concourse F	38%	217
   AT&T	International Terminal	49%	49
   T-Mobile	Concourse E	50%	464

   Four of these cells are below half. T-Mobile over Concourse E prints as 50% and is above half before rounding. That is why Concourse E counts among the 7 of 9 buildings where every carrier is healthy on more than half the tiles.

   The zone basis is weighted toward the apron, so it answers a different question: signal over the apron around the gates. Its lowest cells are re-measured as apron cells, not as buildings:

   Carrier	Zone (apron-weighted)	Healthy 5G	Tiles (n)
   Verizon	Concourse D zone	25%	1,361
   T-Mobile	Concourse D zone	29%	1,548
   T-Mobile	Concourse E zone	29%	2,595

   Owner · Imagine Wireless with the Department of Aviation

3. Re-measure Concourse F and the International Terminal first, for all three carriers

   Each carrier's single lowest building is one of these two. For AT&T it is the International Terminal (49%), for T-Mobile the International Terminal (27%) and for Verizon Concourse F (36%). They are also where an arriving international passenger first turns a handset on.

   Building	AT&T healthy	T-Mobile healthy	Verizon healthy
   Concourse F	55%	38%	36%
   International Terminal	49%	27%	61%

   Re-measure against the section 14 criteria, which are monitoring targets, not acceptance criteria.

   Owner · Department of Aviation with the in-building system operator and AT&T, T-Mobile and Verizon

4. Walk-test T-Mobile and Verizon 5G inside two piers

   Use one pier where all three carriers tie over the footprint (Concourse B) and one where AT&T leads (Concourse C).

   Walk from the concourse centre to the gate hold room and onto the jet bridge. That measures the building-versus-apron gap in section 4.7 instead of inferring it.

   Owner · Imagine Wireless with the two carriers

5. Treat Wi-Fi as a capacity layer that coexists with cellular

   41.7% of observed data volume in the concourse zones rides Wi-Fi (section 7.1). That figure blends the public network with tenant, lounge, hotel and personal Wi-Fi, so it is not attributable to any one network.

   Obtain controller telemetry for the public network to size it. Section 5.6 gives the tenant-side reason.

   Owner · City of Atlanta Department of Aviation, Wi-Fi operator

6. Verify public-safety coverage at the fire stations by walk test

   Atlanta Fire Rescue Station 35, Atlanta Fire Station 40 and Atlanta Station 24 Fire Station are mapped points. The section 11 readings are single-point crowdsourced aggregates, not a coverage assessment.

   Confirm the FirstNet Band 14 posture inside the piers. Emergency responder radio coverage is governed by the authority having jurisdiction and NFPA 1221, not by this instrument.

   Owner · Department of Aviation with Atlanta Fire Rescue, the Atlanta Police Department Airport Section and the authority having jurisdiction

7. Decide the CBRS and private-wireless question with the in-building program

   Decide it alongside the in-building program (section 8), not after it. Keep the Concourse D construction phases in view.

   Owner · City of Atlanta Department of Aviation

This window is the baseline. After any in-building change, re-measure with the same instrument and compare each building against the numbers below.

Criterion for each building and carrier: the median over the footprint at or above −95.0 dBm, and the healthy share compared against this window's baseline. Re-test over a like-for-like window after that building's change.

Each cell: this window's median 5G RSRP over the footprint · healthy share (n = tiles over the footprint). Building basis: tiles whose centroid falls inside a pinned OSM passenger-building footprint.

Criterion: the healthy 5G share of all measured gate points rises above this window's baseline. No target share is set. Re-test over a like-for-like window.

Each zone pools the tiles nearest its gates, which are mostly apron. Compare the zone tile median and healthy share against this window's values after the concourse's in-building change.

Each cell: this window's zone tile median 5G RSRP · healthy share (n = tiles in the zone).

No row is a per-gate apron threshold. The area within 80 m of a gate is mostly apron (section 16), and an in-building system is not built to meet an apron reading.

Use the five widest spreads in section 10 (gates C7, C6, D11, C9 and T1) as the validation gate points. Compare each spread against this window's value. A multi-carrier in-building source should narrow it. Re-test over a like-for-like window.

Gate refs and counts come from the OpenStreetMap inventory. Reconcile gate-named criteria against the Department of Aviation's gate plan before adopting them.

The airport fact sheet publishes 155 domestic and 43 international gates, 198 in all, with no per-concourse count held. The official 2012 International Terminal map publishes gates F1 to F14 for Concourse F.

The Concourse D widening program adds and retires gate positions through 2029. A published per-concourse count for this window is not held.

Use the same instrument as this baseline: Ookla Cell Analytics, with identical thresholds. Match the window: the same 12-month length, or a shorter matched window disclosed with the result.

Ookla's date-range comparison renders the before and after directly. Imagine Wireless re-measures after any in-building change goes live. The crowdsourced tiles give the before-and-after comparison. Acceptance itself is per-carrier commissioning.

Each layer below answers a question this issue cannot. Most of them come from the airport, the carriers or an Ookla portal pull.

Every number on this page comes from Ookla crowdsourced tile data for a 12-month window, joined to mapped gates, tenants and buildings. This section states how, and what the data cannot show.

One locked ladder applies everywhere. No venue-specified threshold was provided.

Noise floor. A margin under 1 dB is not a ranking. Every rank statement prints its margin. The T-Mobile-to-Verizon gate-median margin of 0.9 dB is reported as a tie. The tie is at the median only. Verizon is the weakest carrier by more than 1 dB at 88 gate points, and T-Mobile at 10 (section 10).

Minimum sample. No chart or comparative claim ships on fewer than 30 tiles or points. Concourses B, C, D and E clear the gate floor. Concourse T (21), Concourse A (29) and Concourse F (12) do not. Most tenant categories hold fewer than 30 points. Those tallies appear as counts with n visible and are not ranked. Zone, terminal, building and category comparisons run on the tile basis.

User_Density is the count of unique Android devices seen in a tile over the window, deduplicated. It does not count simultaneous users or time spent. A zone sum of it is device-tile observations.

Wi-Fi offload = Σ Wi-Fi MB ÷ Σ (Wi-Fi + Mobile) MB by volume, at venue and zone level only, on the All Networks panel. It cannot separate the public network from tenant or personal networks. Per-carrier offload is not derivable from this dataset and is not implied.

The panel. Background scans from consumer Android handsets, which Ookla puts at about 3% of devices. iOS devices are not sampled. Panel makeup differs by carrier, so cross-carrier comparisons carry a device-population bias the export cannot correct.

The passenger-journey and tenant readings in sections 4.8 and 5.6 are interpretations of the same aggregates, and are labelled as such.

The data is weighted toward gate areas and public corridors. It does not cover jet-bridge interiors, the Plane Train tunnel or baggage corridors.

PCI diversity, serving-cell volatility, band distributions and Site Finder positions are aggregate proxies or estimates. They are not active traces, registry data or carrier network-management data. The optimization priority index is Ookla's vendor-defined layer. Inferences about cause are labeled as inferences.

One carrier can show several n on this page. This table shows which population each one counts. Cell, PCI and band fields are smaller still (next table).

Cell, PCI and band fields fill only where Ookla resolved a serving cell. Their n is smaller than the RSRP n for the same carrier. The table shows how many on-airport tiles fill each field.

The audit dossier also records capture dates for transcribed values (concourse airline assignments) and the public source behind every entity statement. No value is estimated or invented.

All photographs via Wikimedia Commons, used unmodified except for resizing. Every photograph dates from 2023 or later. Photographs are illustrative material, not measurements.