Source-linked AI summary
A Characterization of the COVID-19 Pandemic Impact on a Mobile Network Operator Traffic
Andra Lutu, Diego Perino, Marcelo Bagnulo, Enrique Frias-Martinez, Javad Khangosstar
TL;DR
Lockdown changed mobility and telecommunication use. The paper analyzes O2 signalling and network-performance data, finding reduced mobility and download traffic alongside a surge in voice traffic.
Problem
The paper investigates how lockdown-related changes in people’s mobility affected a UK mobile network operator’s traffic patterns.
Source: §1 · p. 1Method
The authors analyze cellular signalling and network-performance statistics to assess user mobility and network usage across the UK and specific regions.
Source: §1 · p. 1Results
Mobility fell 50%, while voice traffic rose up to 150% and overall download traffic declined 20%, with especially large decreases in densely populated urban areas.
Source: §7 · p. 13Takeaways & Limitations
Changes were geographically uneven, including sharper usage decreases in dense urban areas and increased uplink traffic in suburban areas.
Source: §7 · p. 13Home-location estimates depend on the MNO’s market share and how well it reflects the general population, although comparison with official population estimates showed a linear relationship (r2 = .955).
Source: §2.3 · p. 4
Abstract
from arXiv · showhide
During early 2020, the SARS-CoV-2 virus rapidly spread worldwide, forcing many governments to impose strict lockdown measures to tackle the pandemic. This significantly changed people's mobility and habits, subsequently impacting how they use telecommunication networks. In this paper, we investigate the effects of the COVID-19 emergency on a UK Mobile Network Operator (MNO). We quantify the changes in users' mobility and investigate how this impacted the cellular network usage and performance. Our analysis spans from the entire country to specific regions, and geodemographic area clusters. We also provide a detailed analysis for London. Our findings bring insights at different geo-temporal granularity on the status of the cellular network, from the decrease in data traffic volume in the cellular network and lower load on the radio network, counterposed to a surge in the conversational voice traffic volume.
1 INTRODUCTION
The study examines how UK lockdown measures affected O2 users’ mobility and cellular-network traffic. It reports lower data traffic and radio load alongside a surge in voice traffic, with substantial geographic differences.
1 INTRODUCTION
- UK mobility fell 50% during lockdown, and the authors report no correlation with confirmed COVID-19 cases, attributing the reduction to enforced government orders.The reduction was more significant in densely populated urban areas than in rural areas.Source: §1 · p. 1
- The authors report a 50% UK mobility decrease during lockdown, followed by regional relaxation with increases in London and West Yorkshire in weeks 18–19; about 10% of London residents temporarily relocated.Source: §1 · pp. 1–2
- Downlink data traffic fell by up to 25% and radio load by approximately 15% across the UK, returning MNO traffic load to levels similar to March 2019.The paper relates the usage decrease to reduced mobility and people likely relying more on residential broadband for download-intensive applications.Source: §1 · p. 2
- VoLTE traffic peaked at 150% above the pre-pandemic national median, offsetting near-flat uplink data traffic; voice packet-loss error rates rose by more than 100% in weeks 10 and 11.The authors attribute the packet-loss increase to congestion in voice interconnection infrastructure and report that error rates returned to previous levels in following weeks.Source: §1 · p. 2
- Mobile-network usage decreased more in densely populated urban areas than in rural areas, likely reflecting central London districts’ seasonal residents and business and commercial activity.The urban-rural difference includes downlink and uplink traffic volumes, radio load, and active users.Source: §1 · p. 2
2 DATASET
The dataset combines passive signalling and radio-performance feeds from a large UK mobile network operator with device, location, administrative, and geodemographic data. The authors derive aggregated mobility measures and daily 4G performance statistics for analysis at multiple geographic scales.
2.1 Measurement Infrastructure
- The passive measurement infrastructure monitors MME, MSC, SGSN/SGW, and cell sites, collecting control-plane voice and data information and cell-site performance indicators.The network includes 2G, 3G, and 4G technologies, and the MNO had more than 25% UK market share in 2019.Source: §2.1 · p. 2, §2 · p. 2
2.2 Data Feeds
- Data feeds combine control-plane event logs, device-catalog and network-topology records, hourly radio-performance KPIs, and postcode-linked administrative and geodemographic classifications.The geodemographic classification is based on the 2011 Output Area Classification.Source: §2.2 · p. 3
2.3 Mobility Statistics
- Signalling data locates smartphone users at radio towers; mobility statistics are calculated in six four-hour bins and daily, with roughly 22 million native users retained after top-20-tower filtering.Devices are filtered to smartphones and native MNO users before tower locations are attached using network topology.Source: §2.3 · p. 3
- Home postcodes use the tower with the longest nighttime connection over at least 14 February days; estimates cover about 16 million users and track entropy and radius of gyration at postcode-or-larger scale.Entropy represents movement repeatability, while radius of gyration indicates distance travelled; the inferred home distribution is influenced by MNO market share.Source: §2.3 · p. 4
2.4 Network Performance Statistics
- Network analysis centers on 4G cells, where users spend 75% of their time, and tracks data, radio-load, throughput, active-user, and conversational-voice KPIs aggregated to daily cell medians.Voice metrics include traffic volume, simultaneous active users, and uplink and downlink packet-loss rates.Source: §2.4 · pp. 4–5
3 MOBILITY
The analysis tracks daily mobility for approximately 16 million users over weeks 10–19 of 2020, comparing values with week 9. Mobility fell sharply after stay-at-home measures, with regional variation in later relaxation and a sustained decline in Inner London residents present at home.
3.1 National Mobility
- Gyration fell 50% by late week 13 versus week 9, after a 20% decline in week 12; entropy also dropped, but less than gyration.Mobility metrics showed a slight increase from week 15 despite the lockdown remaining in force.Source: §3.1 · pp. 5–6, Figure 3 · p. 5
- Mobility reduction was not correlated with reported COVID-19 cases; entropy decline began after the pandemic declaration and became significant after lockdown.The paper relates the reduction to public announcements and lockdown measures.Source: §3.1 · pp. 5–6, Figure 4 · p. 5
3.2 Regional Mobility
- All five studied regions showed sharp mobility declines in weeks 13–14, but London and West Yorkshire increased mobility in weeks 18–19 while Greater Manchester and West Midlands stayed low.London’s gyration was 20% below the national average and its entropy 20% above it.Source: §3.2 · p. 6, Figure 5 · p. 6
3.3 Geodemographic Mobility
- Across geodemographic groups, gyration fell by more than 50% versus the week-9 national average, with a transition in week 12 and steep decline from week 13.The central group had the largest gyration reduction but the smallest entropy reduction; distributions varied little across clusters.Source: §3.3 · pp. 6–7, Figure 6 · p. 7
3.4 Temporary Relocation from London
- From week 13 onward, 10% fewer Inner London residents were present in their inferred home area; Hampshire received more London residents during much of lockdown.The analysis identifies relocations when residents’ visited locations did not include their home county that day.Source: §3.4 · p. 7, Figure 7 · p. 8
- Before stay-at-home orders, Londoners increasingly spent weekends in other counties; that pattern disappeared in weeks 11–12, while travel to East Sussex rose just before the order.Weekend trips to Hampshire also increased by late April, with a smaller increase to Kent.Source: §3.4 · pp. 7–8, Figure 7 · p. 8
4 MOBILE NETWORK PERFORMANCE
Across the UK, lockdown coincided with lower downlink traffic and radio load but sharply higher conversational voice traffic; the size and direction of network changes varied across regions and geodemographic clusters.
4.1 Data Traffic Evolution
- Downlink data traffic reached 24% below week 9 in week 17, while total network data traffic fell roughly 20%.Downlink volume rose 8% in week 10 before declining; uplink volume changed little, with a median daily decrease under 5% in week 18.Source: §4.1 · pp. 8–9
- Downlink user throughput fell by up to 10% despite lower traffic and a 15.1% radio-load decrease, which the authors conjecture reflects application limits.They also report 28.6% fewer active downlink users at the week 19 minimum.Source: §4.1 · pp. 8–9
4.2 Voice Traffic
- Voice traffic volume spiked 140% in week 12, but the increase did not offset decreases in other traffic classes; excess inter-MNO interconnection demand raised downlink packet loss in weeks 10–12.Network operators and service providers quickly restored downlink error rates below normal values.Source: §4.2 · pp. 9–10, Figure 9 · p. 9
4.3 High Density Areas
- Downlink traffic fell in all five high-density counties, most sharply in Inner London at 41%, versus about 20% elsewhere and 15% in Outer London.Inner London also saw fewer active downlink users; the authors identify temporary resident relocation as a contributing factor, while uplink trends differed between Inner and Outer London.Source: §4.3 · p. 10
4.4 Geodemographic Clusters
- Most geodemographic clusters followed national trends, but rural-resident areas kept downlink volume largely stable while Cosmopolitan areas saw sharp declines after week 13.Connected users in Cosmopolitan areas fell by up to 50%; their user-count and downlink-volume correlation was +0.973, compared with 0.299 for Rural residents.Source: §4.4 · p. 11, Figure 10 · p. 11
5 LONDON-CENTRIC ANALYSIS
The London analysis finds especially large traffic declines in central postal districts and Cosmopolitan areas, while other districts and clusters show distinct user and traffic patterns.
5.1 Network Performance
- Downlink and uplink traffic in Western Central fell over 80% between weeks 14 and 19; Eastern Central declined more than 70% relative to week 9.These central-district declines coincided with fewer radio users and lower resource occupancy.Source: §5.1 · p. 12, Figure 11 · p. 12
- Northern London had 10–23% more downlink users in weeks 10–14 while downlink traffic stayed stable, unlike the declining central districts.The authors suggest this may indicate mobile-network hotspots shifting from central areas toward the north.Source: §5.1 · p. 12, Figure 11 · p. 12
5.2 Geodemographic Clusters
- Cosmopolitan areas had over 50% lower uplink and downlink traffic in week 13, while Multicultural-cluster areas had 40% higher uplink traffic and over 20% more active downlink users.All three London geodemographic clusters followed the same user downlink-throughput trend.Source: §5.2 · pp. 12–13, Figure 12 · p. 12
6 RELATED WORK
Related work documents pandemic-era mobility and network changes using phone traces and other aggregated sources. The paper contrasts its UK mobile-network findings with reports from other operators and internet providers.
6 RELATED WORK
- Studies used cell-phone traces, Google place-visit counts [31][19][23], Facebook maps [16], Cuebiq data [27], and Baidu mobility indexes to examine mobility changes and interventions.For the UK, TomTom reported 60–80% fewer trips during lockdown, consistent with this paper; another study found mobility began falling before lockdown, with the sharpest drop after closure.Source: §6 · p. 13
- U.S. reports described higher home Wi-Fi use and lower mobile LTE traffic, while application reports recorded increases in VoIP, videoconferencing, VPN, and video use.Other studies reported increased outgoing educational traffic [10], fixed-internet traffic growth of 15–20% [11], and regional variation in edge performance.Source: §6 · p. 13
- U.S. operators reported voice-minute increases up to 24% and data-minute increases around 20% [5], whereas this paper found no data-traffic surge; BT reported a 5% mobile-data decline and voice growth.The authors suggest different prevalence of flat-fee contracts may explain the U.S.–UK data-traffic difference.Source: §6 · p. 13
7 CONCLUSIONS
The paper concludes that UK lockdown coincided with a 50% mobility decrease and substantial, geographically uneven traffic changes. The MNO maintained service quality overall, though voice packet loss from interconnection-capacity pressure required intervention.
7 CONCLUSIONS
- Mobility decreased 50% overall, with nonuniform changes across geographical areas and social backgrounds.The paper reports voice traffic increases up to 150%, overall download-traffic decreases of 20%, especially in densely populated urban areas (-60%), and a 10% uplink increase in suburbanities.Source: §7 · p. 13
- Despite substantial pattern changes, the MNO maintained service quality standards, with radio load below common values and per-user throughput likely application-limited.Source: §7 · p. 13
- Voice-traffic packet loss followed interconnection infrastructure capacity being exceeded, and network operations teams fixed the issue.Source: §7 · p. 13
A ETHICAL CONSIDERATIONS
Data collection and retention followed the MNO’s terms and local regulations, with operator-reviewed privacy safeguards and restrictions on data sharing.
A ETHICAL CONSIDERATIONS
- Network data collection and retention complied with the MNO’s terms and conditions and local regulations.The datasets were covered by NDAs prohibiting third-party re-sharing, including for research purposes.Source: A ETHICAL CONSIDERATIONS · p. 14
- Operator-reviewed processing extracted only aggregated postcode-level user information, with no person-identifiable, personal, or contract information available.The operator reviewed and validated raw data for GDPR compliance.Source: A ETHICAL CONSIDERATIONS · p. 14
- The paper’s authors did not participate in extracting or encrypting the raw data.Source: A ETHICAL CONSIDERATIONS · p. 14