5G data traffic growth - the devil (FWA) is in the detail

This blog combines two separate, linked LinkedIn articles published in June 2023 on consecutive days. The original posts and comment threads are here and here.

Measuring #mobile data traffic is important for operators, vendors, and policymakers.

As I've said before, we should use *good* #metrics to measure the #telecoms industry, rather than just *easy* metrics. This post is an example of what I mean.

Yesterday, Ericsson released its latest Mobility Report. It's always an interesting trove of statistics on mobile subscribers, networks and usage, with extra topical articles, sometimes written by customers or guests.

While obviously it's very oriented to cellular technologies and has an optimistic pro-3GPP stance, it has a long pedigree and a lot of work goes into it. It's partly informed by private stats from Ericsson's real-world, in-service networks run by MNO customers.

This edition includes extra detail, such as breaking out fixed-wireless access & separating video traffic into VoD #streaming (eg Netflix) vs. social media like TikTok and YouTube.

It had plenty of golden "information nuggets". For instance, traffic density can be 500-1000x higher in dense urban locations than sparse rural areas. I'll come back to that another time.

Global mobile data grew 36% from Q1'22 to Q1'23. The full model online predicts 31% growth in CY2023, falling to just 15% in 2028, despite adding in AR/VR applications towards the end of the decade. That's a fairly rapid s-curve flattening.

For Europe, MBB data growth is predicted at 29% in 2023, falling to only 12% in 2028. That's a *really* important one for all sorts of reasons, and is considerably lower than many other forecasts.

But what really caught my eye was this "#FWA data traffic represented 21% of global mobile data traffic at the end of 2022". Further, it is projected to grow much faster than mobile broadband (MBB) and account for *30%* of total traffic in 2028, mostly #5G. When the famous "5G triangle" of use-cases was developed by ITU, it didn't even mention FWA.

However, the report didn't break out this split by region. So I decided to estimate it myself based on the regional split of FWA subscribers, which was shown in a graphic. I also extended the forecasts out to 2030.

I then added an additional segmentation of my own - an indoor vs outdoor split of MBB data. I've pegged this at 75% indoors, aligning with previous comments from Ericsson and others. Some indoor MBB is served by dedicated in-building wireless systems, and some is outdoor-to-indoor from macro RAN or outdoor small cells.

The result is fascinating. By the 2030, it is possible that over 40% of European 5G data traffic will be from FWA. Just 14% of cellular data will be for outdoor mobile broadband. So what's generating the alleged 5G GDP uplift?

That has massive implications for spectrum policy (eg on #6GHz) and proposed #fairshare traffic fees. It also highlights the broad lack of attention paid to indoor cellular and FWA.

Note: This is a quick, rough estimate, but it's the type of data we need for better decisionmaking. I hope to catalyse others to do similar analysis.

 

A separate second post then looked at the policy aspects of this:

Yesterday's post on mobile data traffic - and contribution from 5G FWA and indoor use - seems to have struck a chord. Some online and offline comments have asked about the policy implications.

There are several conclusions for regulators and telecoms/infrastructure ministries:

- Collect more granular data, or make reasoned estimates, of breakdowns of data traffic in your country & trends over time. As well as #FWA vs #MBB & indoor vs outdoor, there should be a split between rural / urban / dense & ideally between macro #RAN vs outdoor #smallcell vs dedicated indoor system. Break out rail / road transport usage.
- Develop a specific policy (or at least gather data and policy drivers) for FWA & indoor #wireless. That feeds through to many areas including spectrum, competition, consumer protection, #wholesale, rights-of-way / access, #cybersecurity, inclusion, industrial policy, R&D, testbeds and trials etc. Don't treat #mobile as mostly about outdoor or in-vehicle connectivity.
- View demand forecasts of mobile #datatraffic and implied costs for MNO investment / capacity-upgrade through the lens of detailed stats, not headline aggregates. FWA is "discretionary"; operators know it creates 10-20x more traffic per user. In areas with poor fixed #broadband (typically rural) that's potentially good news - but those areas may have spare mobile capacity rather than needing upgrades. Remember 4G-to-5G upgrade CAPEX is needed irrespective of traffic levels. FWA in urban areas likely competes with fibre and is a commercial choice, so complaints about traffic growth are self-serving.
- Indoor & FWA wireless can be more "tech neutral" & "business model neutral" than outdoor mobile access. #WiFi, #satellite and other technologies play more important roles - and may be lower-energy too. Shared / #neutralhost infrastructure is very relevant.
- Think through the impact of detailed data on #spectrum requirements and bands. In particular, the FWA/MBB & indoor splits are yet more evidence that the need for #6GHz for #5G has been hugely overstated. In particular, because FWA is "deterministic" (ie it doesn't move around or cluster in crowds) it's much more tolerant of using different bands - or unlicensed spectrum. Meanwhile indoor MBB can be delivered with low-band macro 5G, dedicated in-building systems (perhaps mmWave), or offloaded to WiFi. Using midband 5G and MIMO to "blast through walls" is not ideal use of either spectrum or energy.
- View 5G traffic data/forecasts used in so-called #fairshare or #costrecovery debates with skepticism. Check if discretionary FWA is inflating the figures. Question any GDP impact claims. Consider how much RAN investment is actually serving indoor users, maybe inefficiently. And be aware that home FWA traffic skews towards TVs and VoD #streaming (Netflix, Prime etc) rather than smartphone- or upload-centric social #video like TikTok & FB/IG.

Telecoms regulation needs good input data, not convenient or dramatic headline stats.

 

UK FTTP: Consolidation and driving uptake

This post originally appeared on June 16 on my LinkedIn feed, which is now my main platform for both short posts and longer-form articles. It can be found here, along with the comment stream. Please follow / connect to me on LinkedIn, to receive regular updates (about 1-3 / week)

Last week I attended the ISPA UK Business Models event, primarily about #FTTP build & adoption.

Two themes dominated:

- Consolidation patterns. The UK has >150 ISPs building #FTTX networks, with a patchwork mix of small/large, urban/rural & vertical/wholesale-only. As interest rates rise & consumer spending is inflation-limited, not all can stay viable.
- How can uptake be accelerated? While many homes are "passed" by fibre, comparatively few are actually signing up for FTTP access services. The lack of revenue for new #AltNets exacerbates the first issue.

Not discussed: data traffic volumes or so-called #fairshare. All the investment is going into initial builds, not capacity upgrades. Streaming and >500GB/mo is actually good news, not a cause for lobbyist handwringing.

The consolidation pathway is complex. There are 3 elements:

- Distress: companies running out of cash, unable to raise fresh capital, and selling assets or the whole business to deeper-pocketed consolidators willing to take a long view of the market.
- Proximity: Mergers or perhaps wholesale/sharing deals between geographic neighbouring ISPs, for scale efficiencies.
- Strategic: larger "mega-mergers" perhaps between wholesalers and integrated telcos, or between B2B and B2C specialists.

There are plenty of challenges. M&A means blending FTTP providers with different vendors, maybe different network engineering qualities, different back-office systems (perhaps proprietary) etc. There may be significant integration costs and practical headaches. Another issue to resolve is competing "overbuilt" fibre grids in urban areas, especially as OpenReach gets to more locations and offers cheap "Equinox2" wholesale.

The uptake question is also thorny. A few speakers pointed out that the UK's FTTC / VDSL broadband mostly proved itself "good enough" during the pandemic, so convincing people they need FTTP or gigabit speeds is a tough sell, especially given cost-of-living issues.

Unless they currently have really terrible connectivity, few people really want to take a day off work to wait for an engineer, risk a day or two without Internet if the switch doesn't work straight away, or pay more and sign up for a new longterm contract.

For some, futureproofing can wait until the future, it seems.

I can think of a number of ways that uptake could be incentivised:

- Trumpet fibre's uses, reliability & maybe impact on property values
- Subsidise an overlap of the old service with the new FTTP, so customers' old connection wouldn't be switched off before it was fully live
- Offer funding to connect homes that are "passed" as long as the connection is fully open-access / wholesale-ready
- Measure, monitor and incentivise B2B use of fibre as well as residential (retail, schools, small offices, home-workers etc)
- Better mapping to find and deal with "exceptions"

All would be enhanced by a consistent view (or scenarios) for the UK #fibre "end state". At the moment that is too amorphous.

Connectivity on trains is hard - but both Wi-Fi and cellular need to be provided for passengers

This post originally appeared on May 24 on my LinkedIn feed, which is now my main platform for both short posts and longer-form articles. It can be found here, along with the comment stream. Please follow / connect to me on LinkedIn, to receive regular updates (about 1-3 / week)

 There have been recent headlines about the possible ending of on-train passenger #WiFi services in the UK. It is deeply controversial.

Apparently the Department for Transport (DfT), United Kingdom has insisted rail WiFi must be "justified financially". It's unclear if that means by extra ticket sales, higher customer satisfaction, or the use of WiFi for #train operational functions like cameras and wireless payment terminals.

I hope it's not referring to so-called "monetisation" by customers paying for WiFi, or being served adverts. On trains, WiFi is a basic amenity, like toilets or power sockets.

That said, train WiFi in the UK is often problematic. It uses clunky captive portals, and often old access points & slow/patchy 4G backhaul. It often fails to work well, or at all. It sometimes blocks video or VPNs. By contrast, in-station WiFi is run separately - and often much better.

Public cellular coverage on the rail network is also poor. Many rail lines run through cuttings and tunnels with limited room for trackside infrastructure & poor lines-of-sight to cell towers. The recent Department for Science, Innovation and Technology Wireless Infrastructure Strategy highlighted poor #railway #wireless coverage & pushed for regular monitoring and access to trackside fibre.
 
What should DfT, DSIT, Network Rail, Train Operating Companies and the future restructured Great British Railways Transition Team (GBRTT) do?
 
- Recognise both cellular & WiFi are essential for passengers, especially on long-distance trains where laptops are common
- Understand that cellular - especially #5G - has problems with signals reaching inside train carriages
- Don't underestimate forecasts for future data use. Add in uplink as well as downlink, and think about latency. Trains may need 1-5 Gbps in the medium term, via a mix of cellular & WiFi.
- Ensure on-train WiFi is easy to use & easily-upgraded. No captive portals, no “monetisation” with ads/data capture & a clear roadmap for regular upgrades. No blocking of any apps, especially VPNs and video. Apply Net Neutrality rules.
- Federation or roaming between on-train & station WiFi systems, extending to smart cities & metro bus/train/tram WiFi over time
- Easier access for MNOs / #neutralhosts to build trackside or near-track infrastructure & use gantries & fibre assets
- Decouple passenger connectivity needs from future critical #FRMCS deployment. They have different timing/cadence & investment cases
- Look at trackside 5G neutral host networks delivered with “excess” spectrum from any future 4-3 merger of MNOs
- Insist on-train gateways are modular & can use a dynamic mix of public 5G, trackside wireless & eventually satellite in remote areas. Ensure they are easily upgradeable without trains being taken out of service
- Upgrade on-train signal repeaters & look at window-etching for better outdoor-to-indoor performance

Note: I wrote this on WiFi on a train back to London from this week’s Wi-Fi NOW conference.

 

Data traffic growth forecasts - AD Little's new report has a lot better methodology than most

This post originally appeared on June 5 on my LinkedIn feed, which is now my main platform for both short posts and longer-form articles. It can be found here, along with the comment stream. Please follow / connect to me on LinkedIn, to receive regular updates (about 1-3 / week)

When I saw that Arthur D. Little had published a report on “The evolution of data growth in Europe”, on behalf of ETNO Association & GSMA, I rolled my eyes.
 
Both organisations have previously published terrible studies by consultants, riddled with flawed assumptions and dodgy multiplier "fiddle factors". I’ve loudly criticised Axon and Coleago reports related to the (un)#fairshare and #6GHz #spectrum debates respectively.
 
So I started the ADL report with trepidation, not helped by a strange typo / editing error in the first paragraph.
 
But actually, the report is pretty good, and I broadly agree with both methodology and conclusions, albeit with one major caveat.
 
It estimates usage of home and mobile broadband on the basis of hours-per-day of active use of heavy applications such as video streaming, gaming and possible metaverse-type experiences.
 
I’ve used GB-per-hour myself, to model passenger data-traffic demand on trains. It makes more sense than the usual Gbps, as most applications are “bursty”. It also fits the typical heuristics of human behaviour. How many seconds a day do you spend on social media?
 
The central prediction of 20% growth in fixed traffic and 25% for mobile usage seems reasonable. I could argue for 25/20 rather than 20/25, but it's fine as a rough estimate.

Importantly these rates for the next few years are well within the bounds of both fixed broadband (moving to #FTTP) and mobile (on #5G) without incremental investments in extra capacity, beyond the main "generational" shift & CAPEX. And that is driven by government policy and competition, not traffic load and congestion. The report convincingly shows that nobody really needs/values more than 100Mbps for current apps, so #gigabit networks have plenty of headroom.

My main criticism is there is no analysis of mobile device traffic carried over fixed networks and #WiFi. Smartphones used at home for video, gaming or social media will be c80% on Wi-Fi, and indoor usage is c80% of the total.

The report also talks about AI pre-emptively downloading content for “infinite scrolling”, but doesn't suggest it could be smart enough to do so mostly over cheap / low-energy fixed connections. (IMO, by 2030, governments may *mandate* cellular offload via neutral-host or Wi-Fi for indoor use).

I agree with the report's assertions that VR is in an indoor/fixed application, that most #IoT traffic is a rounding-error and that #Web3 is probably irrelevant. The #metaverse scenarios seem mostly plausible.
 
One area I think ADL underestimates is fixed broadband for video streaming. While Netflix and YouTube are “active” viewing, historically, many people just leave broadcast TV switched on, even if nobody is in the room except the cat.

If TV really goes online-only, then that becomes a genuine “waste” of capacity, unless you can advertise to pets.

Overall - really quite good analysis, which (ironically, given the sponsors) fatally undermines the #InternetTrafficTax rhetoric.

 

Spectrum: The shifting tone of the satellite industry

This post originally appeared on June 7 on my LinkedIn feed, which is now my main platform for both short posts and longer-form articles. It can be found here, along with the comment stream. Please follow / connect to me on LinkedIn, to receive regular updates (about 1-3 / week)

I'm in Brussels this week at the Forum Europe European Spectrum Management Conference.

There's a lot to discuss, especially around #6GHz and 3.8-4.2GHz and the role of unlicensed and local/shared bands, as well as the upcoming World Radio Conference WRC-23.

I'll have more to say, but here I just want to highlight one particular theme that has been evident over the last couple of days: the tone of the satellite sector, which is here in force, especially with GSOA and Intelsat.

In the past at these #spectrum events, the #satellite industry has turned up with a familiar script:

"Hi, we're from the satellite industry. Please don't take our spectrum. We help with defence, aviation & connecting the unconnected. Please don't take our spectrum. We work tightly with the mobile industry, doing backhaul & IoT and timing sync. They're our friends & vice versa. Oh, and did we mention our spectrum? Please don't take any more of it"

But this time, it's different. The message is now closer to:

"We're doing all ths cool new stuff, including for wireless broadband, direct to device and defence. So actually, we want to keep all our spectrum. And maybe give back the old #mmWave spectrum you took years ago, that the mobile industry hasn't even used. Seriously, you want *more* spectrum to be taken from us and pre-allocated to 6G now? Are you having a laugh?"

There was a whole panel on direct-to-device, and satellite has fought its corner on the upper 6GHz (it can coexist with low/medium power WiFi, but not high power 5G) and fixed satellite links in 4GHz band. The future-looking 6G panel started a fierce debate on 7-24GHz, which covers various of the satellite incumbent bands.

There's been a few references to South Korea's regulator reclaiming unused 28GHz licenses from MNOs that haven't used the band. And there's a broad opinion that mobile/IMT is not a friendly partner for spectrum-sharing, at least for national MNO macro networks at full power. (Local private networks are OK-ish, it seems).

"An IMT identification is an eviction notice - the incumbents must leave".

"It's disingenuous to discuss coexistence studies - we've been here before and know how it ends. It's not our first rodeo with the mobile industry"

Now clearly this year, in the last few months before WRC23, is when arguments get more vigorous. But some of the stuff at the #EUspectrum event has been seriously punchy - Intelsat asked whether Europe should be focused on primacy in an amorphous "race to 6G" or a more geopolitically-crucial "space race".

My view is that the #5G industry is seeing some chickens coming home to roost at the moment. It overpromised Release 18 features with Release 15 timelines, got mmWave spectrum years before it could be exploited, and have left politicians and regulators with egg on their faces.

Meanwhile, the satellite sector is positioning itself as super-cool and important. It has a swagger that is being noticed by policymakers, and for good reason.

 

Private 5G: Two different approaches at the Coronation

This post originally appeared on June 9 on my LinkedIn feed, which is now my main platform for both short posts and longer-form articles. It can be found here, along with the comment stream. Please follow / connect to me on LinkedIn, to receive regular updates (about 1-3 / week)

 A month ago, the UK and much of the world watched King Charles' #Coronation in London.

They were able to watch it partly because of the immense efforts of the various #broadcasters involved. Since then, two separate stories have emerged about the role of dedicated #5G connectivity in the TV coverage:

1) A dedicated private 5G network supplied by Neutral Wireless and BBC R&D, used by several broadcasters
2) A slice of the Vodafone public 5G network, enabled for ITN, based on Ericsson gear

In the comments I've linked to various articles and a great interview on Ericsson's Voice of 5G podcast show. They have details of the other partners involved too. In the BBC blog post they also mention a 3rd network on a separate cell, working alongside Sony, for low-latency (I think) remote-controlled cameras.

The #Private 5G network used 8 radios along The Mall (the tree-lined road between Buckingham Palace to Trafalgar Sq). It used 2x 40MHz channels in the UK's shared-licence band between 3.8-4.2GHz, with 1Gbps capacity (mostly for uplink). It was used by around 60 devices - I guess mostly cameras and test equipment via gateways, plus the BBC's onsite radio studio. They also used LiveU bonding systems to add capacity from public MNO networks. I'm not sure about the vendors of the radios or standalone core.

The 5G SA #networkslicing solution was apparently used for a single sector at a 3.5GHz temporary base station aimed at the Palace balcony. It also worked with LiveU. On the podcast, Andrea Donà (VF's head of network in the UK) talks about "dedicating bandwidth to one sector for the slice" and carving out some of the uplink capacity.

One thing that is unclear to me is how many other users were sharing the VF standalone 5G network hosting the slice - SA hasn't been fully launched commercially in the UK, although in January VF said it had invited selected users to trial it. I also don't know whether the 5G NSA and SA networks were sharing the radio resource, or if they use separate channels.

The public 4G / 5G networks (and also Wi-Fi bands) in the area were pretty overloaded, despite additional mobile towers adding capacity. The Vodafone / Ericsson podcast notes that VF uses "all the bands" at major events (although there's no #mmWave 5G in the UK yet) - so including 4G at 2.1GHz and 2.6GHz, and some lower bands for 2G/3G.

My take from this is that #private5G is considerably more mature than #5Gslicing, but that both are interesting for broadcasters. Both need quite a lot of specialist engineering, but TV is a sector with lots of very clever specialists and great ability to set up temporary networks. Of course, both networks were *outdoors* which meant that the thick stonework of the palace and Westminster Abbey weren't relevant.

One last note - the huge bulk of broadcast audiovisual output at the coronation would have used dedicated #PMSE wireless for cameras and microphones. But the #UHF spectrum debate is for another post.

 

CAPEX in telecoms - beware of headline numbers

This post originally appeared on June 12 on my LinkedIn feed, which is now my main platform for both short posts and longer-form articles. It can be found here, along with the comment stream. Please follow / connect to me on LinkedIn, to receive regular updates (about 1-3 / week)

CAPEX numbers are important in #telecoms. But they're also often collected and analysed in a haphazard fashion, or sometimes twisted and misinterpreted. There are examples that wrongly imply casual links or are carefully selected to drive specific policy choices.

- Telco execs watch CAPEX stats as they're important elements of cashflow & also signify key strategies and technology transitions
- Vendors watch #CAPEX stats to understand demand for new products
- Investors watch CAPEX as inputs to their valuation models, and as a barometer for company/industry health and prospects
- Policymakers watch CAPEX as it gets captured in "investment" statistics, and as an indicator for potential regulatory changes (or as a metric of success of previous policies)

Various ratios are commonplace, for both companies and the industry:
- CAPEX vs. revenues
- CAPEX vs. EBITDA
- CAPEX of telecoms vs. tech/hyperscalers
- CAPEX vs. R&D spending
- Fixed vs. Mobile CAPEX
... and so on

The problem is that "telco CAPEX" is also a very vague and malleable concept. Digging into it reveals many more questions - and problems with the methodologies and conclusions drawn, especially where headline numbers are concerned.

Some of the questions I'm currently looking at include:

- What counts as a "telco"? Are you including towercos, subsea fibre operators, municipalities building networks, MVNOs and many others?
- Are historic CAPEX numbers restated when telcos sell or acquire other businesses, especially tower spin-outs?
- Is it meaningful to compare CAPEX for 10 / 30 / 50 year assets such as #FTTP, which will generate decades of new revenue, with last year's figures?
- How do you separate CAPEX for basic coverage vs. incremental capacity vs. "generational" upgrades to fibre or #5G? A lot of CAPEX occurs even if usage is low
- How do you deal with leasing or other financing models? If CAPEX shifts to OPEX, how is it captured in the stats?
- What happens with "cloudified" networks? Firstly they rely on shared (often 3rd-party) assets, and secondly they are *supposed* to lower costs / investments. But will the lower CAPEX be viewed as a sign of distress, not modernisation?
- Is non-network CAPEX broken out (eg retail sites, central offices, datacentres etc)?
- Is "adjacent capex" included and if so, how?, eg in-building #wireless, #spectrum licenses, software development

I hear many commentators and lobbyists claim "#NetNeutrality led to lower CAPEX!" or "Streaming traffic leads to higher CAPEX!" or "There's an investment gap!". Without detailed data - and an analysis of causality - you have to question the veracity & meaningfulness of such rhetoric.

In summary - CAPEX is indeed important. But in fact it's so important, that headline numbers are often useless or misleading.

Ask for details on segmentation, methodology and definitions - if they aren't available, treat the numbers with deep skepticism.

#FTTX #telcos #regulations #networks #fairshare