VoIPo3G becoming a reality. Remember, you read it here first....

Fascinating post by Andy Abramson about various experiences with Skype over 3G running on the iPhone.

Coupled with efforts by various other VoIP players - and the evolution of assorted voice-on-LTE advocates, it certainly looks as though VoIPo3G (or VoIPo4G if you buy into the marketing guff around LTE) is becoming much more feasible and important.

I've noticed a few of my rival analyst firms putting out forecasts and comments recently, as if this was somehow new and unexpected.

Readers of this blog, and customers of Disruptive Analysis' reports will have seen this coming more than two years ago (I'm pretty sure that I coined the term VoIPo3G myself - see the Google search results here)

I published a research report at the end of 2007 which included a full analysis of what was likely to occur - as well as the probable use cases and partnerships that could/should arise. I predicted up to 255m active users of VoIPo3G by the end of 2012. Not all of these would be for "primary telephony" - the majority would be using VoIP as an adjunct to ordinary full operator voice service.

Now, some things I expected in the report haven't unfolded quite the way I expected at the time I wrote the report:

- HSUPA has been slower to be rolled out in devices than I predicted, which has limited VoIP quality
- Apple and Google have de-railed the 2007 smartphone dominance of Symbian and Microsoft, sending some of the VoIP plans to the drawing board, especially as Apple's deals with telcos prohibited VoIPo3G
- Laptop mobile broadband has grown very swiftly, often used as a direct substitute for fixed broadband, and used with various apps including VoIP
- LTE and WiMAX have rolled out more slowly, voice-over-LTE has been standardised more slowly than anticipated.
- UMB disappeared from radar
- Operators have (generally) not followed my recommendation to experiment with VoIP on HSPA/HSPA+ before committing to it with LTE. This is a strategic error in my view.

Nevertheless, the idea of partnership between operators and Internet VoIP providers has seemed prescient - especially given the Verizon/Skype and Telefonica/Jajah tie-ups. It will be interesting to hear if Andy's theory about close Apple/Skype collaboration is true - and if AT&T and other operators have some indirect involvement as well.

It's about time I did a full reassessment of the mobile VoIP space - I've been covering it since 2001 and have been pretty on-track thus far.

Watch this space.

(If any historians of telecoms strategy and forecasting would like to get the original report, let me know & I'll cut you a very good deal)

Read More

Is it just me, or is 3G either really good or really bad, but rarely "OK"

I've started noticing that my experience of mobile broadband (iPhone 3GS on Voda UK) is much more polarised than that of fixed broadband.

With ADSL, performance seems to vary over a fairly wide spectrum of possibilities, because of the vagaries of contention at my local exchange's DSLAM, the connection of the server at the other end, and especially WiFi in different parts of my house. Overall, it's generally OK - I've had one major outage in 5 years (after a massive fire/flood at a BT Exchange that took out half of central-west London's phones).

But on mobile, it seems to be all or nothing. Either very good... or frustratingly bad.

I've been playing around with the Speedtest app on my iPhone (3GS, on Voda UK). And the results are really polarised. The downlink speeds (in Kbit/s) for my last 20 or so tests:

1940
1238
1 (yes, really - and I got 236kbit/s uplink!)
2638
13
2249
36
1554
33
690
1125
75
1088
717
3465

In other words, I either get 1-2Mbit/s+, or less than 100kbit/s (and not on EDGE, either, according to the indicator). I've only got two "middling" results around 700kbit/s, and none at all in the 200-500kbit/s range. Several times I've had results that make GPRS look poor. On several occasions recently I've had to switch the radio on/off via the "flight mode" switch to see if it helped to re-register. A couple of times I've done a full on/off of the phone.

Now, I recognise that I probably tend to do speed tests when I'm most frustrated, so there's a bias there. There could also be some bugs in the phone's firmware. But it's really conspicuous that I almost never get a "sub-par, but sort-of OK" performance. It's essentially all or nothing, even when I supposedly actually have coverage.

That's a very different user experience to fixed broadband.

I'm sure there are commenters who have much more test/measurement expertise than my amateur armchair efforts. But to my uneducated eye, it looks like there are problems beyond raw air-interface capacity. I'd expect a much more graceful decline in performance if it was just due to 2, 5, 10 or 100 other people in the same cell at the same time checking Facebook at the same time as me.

Is it a glitch somewhere in the RNC or GGSN? Who knows? Am I being policy-managed in response to my usage (middling, maybe 300-500MB/month) or my coruscating blog post on Voda's roaming practices a few weeks ago? [No conspiracy stories here to be fair - the variability was there long before].

Either way, it can't be long before we get some massive crowd-sourced programme to track network quality and deconstruct the root causes - the open question in my mind is whether this happens before or after the operators themselves get on top of it.

The theme of using DPI, probes and data-mining to analyse network and service-assurance problems is one of the top 10 technologies I introduce in my recent research paper on Mobile Broadband Traffic Management.

Read More

What are the side-effects of speed/vol tiering for mobile broadband?

I'm at the Open Mobile Summit again today. I've still got a load of things to write up from yesterday as well (there are some too-brief comments up on my experimental and much-hated Twitter feed @disruptivedean).

I've just seen an interesting presentation from TeliaSonera, talking about its early deployment of LTE in the Nordic market and its pricing/tiering for mobile broadband.

It's worth noting a few things first - Sweden and Norway are quite "special" markets: affluent people, initial introduction of cheap and "very flatrate" USB modems for PCs in 2007 (€20 for all-you-can eat, I think was stated), lots of fast fixed broadband, and quite a lot of quite cheap spectrum, split between relatively few operators (mostly fixed/mobile integrated).

The speaker highlighted an interesting tiering scheme being introduced in the summer, as it extends its LTE rollout beyond its current (very) early-adopter phase of a "couple of thousand" users.

It is going to charge LTE access ("4G") for 30GB at €36 per month, at "the highest speeds feasible on the network", including dropping down to 3G where necessary.

It will charge lower prices of €4, €23 and €32 for 2GB, 10GB and 20GB at 3G-only speeds, dropping down to 2G when the cap is exceeded. It's not clear if this is a throttled "3G speed on a 4G" network, or whether it just involves an non-LTE modem.

So basically, they are using 3G vs 4G to provide price discrimination on a proxy for speed performance - and then combine it with different caps to enable hybrid speed/volume tiers.

But when asked, they are not doing speed discrimination within 3G (eg 3Mbit/s vs. 7Mbit/s vs. HSPA+ at 21Mbit/s) because it is too difficult to guarantee real-world perception of tiering. It's much easier to discriminate between HSPA at "1-10Mbit/s" vs. LTE at "10+".

This all makes sense - trying to have more than 2 or 3 tiers is difficult even in fixed broadband, where there isn't the variable of changing radio conditions involved.

In future, they're planning to offer "top ups" for your cap, if you run short at (say) week 3 of a month. They will also do zero-rated traffic for specific applications supplied by themselves (eg a deal with Spotify will mean that streamed music may not count towards the quota).

There are prepaid plans for 3G but not LTE at the moment.

I'm curious about a few things, though. The first is what the performance of the LTE network will look like when it starts loading up - it's all very well to claim customer-achievable speeds of 20-40Mbit's when you've got maybe 2000 users and 500 base stations (1500 sectors) - most of the time, any active user is going to be alone or perhaps sharing capacity 1-3 other people. What's going to happen when there's a conference room with 50 people with laptops and dongles, or even just a branch of Starbucks with 5 users?

The other question is around unexpected side-effects of price discrimination.

Ultimately, it sounds like we are getting to a position of differential pricing for what is a (relatively) commodity service. The big question is how "tradeable" that commodity is - and how arbitrage might work.

In prepaid voice markets, it is common for users to have multiple SIMs, and swap them over frequently to take account of pricing oddities. Many have multiple phones, to enable them to always call "on net" and benefit from the lowest prices rather than pay extra for cross-network fees.

It seems likely that while me may see more clever operator pricing, we'll also see more clever user behaviour to circumvent this. Given that data services don't need fixed phone numbers, I see much greater opportunity for dual-SIM, connection-sharing, tethering and MiFi-style products to facilitate arbitrage. There is also a lot of potential for new forms of least-cost routing and collaborative applications that pool different users' allowances.

So for example, how long before we see an application that can suggest:

"If I'm in week 4 of the month, and still have >30% of my quota left, then share or trade that capacity with my Facebook friends automatically, when we're within WiFi range of each other".

I think we're on for some fascinating device/web-based techniques for "gaming" the new data pricing structures.

Read More

NEW research report on Mobile data: Offload vs. Compression vs. Policy

Disruptive Analysis has published a short "thought leadership" paper on technologies for managing mobile broadband traffic.

While future deployments of LTE or HSPA+ will add more capacity for mobile operators, it is still critical to examine ways to reduce congestion on a shorter-term basis. And even with future 3G / 4G capacity additions, the hunger from new applications - especially those based around video or other rich media - will demand strong discipline for optimisation.

This is a complex area - there are many ways to control traffic, or minimise its impact on the most expensive parts of the cellular network. There are arguments for offload to WiFi or femtocells, compression in the core network, or numerous forms of innovation for policy management and charging.

This new research paper examines the roles of 10 technology-led approached to data traffic management. Some of them have been introduced in this blog before - offload has been discussed for over two years, and mobile policy management and traffic-shaping for over four years.

I started to discuss the huge diversity of approaches earlier this year - this thought-leadership paper expands on the analysis, based on dozens of extra meetings and interviews I have conducted over the past few months.

*** LINK TO MORE INFORMATION AND PURCHASE OPTIONS HERE***

There is no easy answer - the best approach will depend on the existing customer base and its behaviour, the mix of smartphones and laptops on the network, the operator's spectrum and cell site holdings..... and its forecasts and beliefs about the future.

In many operators, there is also an organisational and management problem: there is often no single individual who "owns" the issue of data traffic, who can develop a holistic solution. Instead, there are often diverse individuals who pursue narrow goals, which can have unintended consequences elsewhere in the network - or impacts customer experience.

In almost no operator is there an individual with the job title of "Policy Manager".

Vodafone recently announced that 85% of its mobile data traffic comes from laptops - but it is seeing faster growth from smartphones now. This highlights the complexity of the decision space - which also affects decisions about spectrum purchase and timelines for LTE/HSPA adoption.

More details on the new Disruptive Analysis paper on Top 10 Technologies for Mobile Broadband Traffic Management are here.

As this is a shorter and more tactical document than full strategy reports, the price points are correspondingly lower, starting at just $350 + tax if appropriate.

Alternatively, please inquire about more customised workshops or consulting projects on this field - an area that Disruptive Analysis has tracked for longer than any other analyst firm.

Read More

Thoughts on the LTE conference

So, two days being bombarded by LTE technology and operator trial/deployment presentations.

Some take-aways:

- When it works, LTE apparently works very well in terms of peak speeds etc.

- For a greenfield operator in an uncontested market, with limited current 3G uptake and users without pre-existing expectations, it would likely be a winner in the medium term

- Unfortunately, LTE is going to deployed into a very messy world, with entrenched business models, diverse frequency allocations, existing "good enough" voice and data technologies that have predictable user experience - and capital constraints imposed by the economy and investors.

The frequency thing is a bit of a killer. So far, Verizon is doing 700MHz, one of the Swedish networks is at 900MHz and the other at 2.6GHz. NTT DoCoMo is using 2.1GHz and later 1.7GHz. China is looking at the TD flavour of LTE in 2.6GHz. Some operators were talking about refarming 1.8GHz for LTE, and some of the US operators mentioned their unique AWS band. O2 is trialling 800MHz digital dividend band in the UK. And I think I've seen a reference to 2.3GHz as well somewhere.

In other words, it's a mess - and quite a few early operators were trying to talk up their preferred option, in an attempt to drive scale. Then add in the fact that most of these bands are quite small in terms of outright size - not enough for competing operators with 2x20MHz allocations, for example. And the fact that the largest spare band at 2.6GHz is (a) still be auctioned in many places, and (b) is, according to O2 UK "very disappointing" in terms of propagation in the real world.

One other message that came out quite strongly is that investors do not seem happy about the notion of another big round of radio network capex, especially given the doubts over whether mobile broadband can be monetised beyond connectivity.

(It was notable that my term "happy pipe" seemed to resonate with quite a few people at the event, which was itself, amusingly, held in a part of Amsterdam called Der Pijp).

My conclusion is that in many markets, LTE networks will develop in a patchwork form, either for specific hotspots to use new spectrum to manage high densities of data users, or else perhaps for city-wide deployments. But especially in Europe, the use of 2.6GHz is going to struggle to enable LTE-only networks - they will either need to be dual-band with a sub-1GHz frequency from Day One, or else will have to rely heavily on HSPA as a fallback. (In which case, why bother with the added complexity - why not just use 2.6GHz HSPA instead? or WiFi?)

This means that the concept of an "LTE application" looks pretty weak in the medium term. And I really don't buy the idea of M2M devices (or cars) being ideal for LTE rather than 2G or 3G - coverage is king. Who wants an LTE tablet that doesn't work in at least as many places as their existing dongle, let-alone an LTE healthcare terminal?

One other thing is likely to disappear from the near-term wishlist is roaming - especially because of the likely diversity in LTE voice implementations for the next 10 years. And if voice roaming is ditched in favour of 2G / 3G circuit connections - why is LTE data roaming urgent anyway? Frankly, 21Mbit/s roaming costing $5 per second is quite enough - who really wants 100Mbit/s at $25 per second? Let's wait for the data roaming price to lose a few zeros before worrying too much about cross-border LTE, eh?

The other elephant in the room is that of user expectations. To be honest, the whole concept of "the user" was woefully lacking in the event. I think I only heard the word "battery" mentioned once. Customers have been accustomed to both price and experience for HSPA and will expect that as a baseline for LTE. There is also strong evidence from the fixed broadband world that the feasible premium for (lots of) extra speed is typically only 0-30% on the price, except for a handful of enthusiasts. Once we get to LTE phones, users will also have reasonable expectations that voice quality, reliability, battery life and coverage are at least as good as that achievable with a $20 GSM handset.

Overall, some aspects of LTE technology development sound positive. But I'm still not expecting any sort of miraculous or revolutionary shift in user perception of mobile broadband on a 5-year view, versus what we have today.

It will also be very interesting to see if any of the new Indian 3G licence-holders opt for LTE rather than HSPA. I suspect they will find the business case quite tricky, given market immaturity and an immediate pent-up demand for cheap mobile broadband in a market with little copper or cable.

NEW Mobile Broadband Traffic Management Paper

NEW Broadband Business Models Strategy Report

Read More

Mobile local search - just leave it to Google. Again.

I've written various times before about my skepticism around "Mobile Search", and especially the idea that any form of "local search" is useful on a handset - at least, beyond that done by a decent implementation of Google.

I've heard any number of straw-man arguments that "people want to find things, not search for them", usually made by wannabee directory-services companies that want to charge companies to be listed.

I've heard similar arguments about the evolution of Yellow Pages to something altogether more interactive and cool - or about using SMS to query the network operator's own database, hooked into your cell ID or other details.

The problem I have is that most of these things just don't work that well, because they rely on an incomplete database that is usually compiled by someone else. It's then interpreted by an imperfect filter, and updated infrequently.

I think that a case can be made for mobile search for "commodity" services, where the user doesn't need the absolute closest option, nor need a complete list of all relevant providers. Plumbers, window-cleaners, maybe dry cleaners, conceivably petrol stations.

But for those areas were there is a qualitative, subjective opinion involved.... you really want as broad and accurate a base as possible. Ideally, you wouldn't want to choose a restaurant from a guide with a random selection of 25% of the possible options - or, worse, where a particular chain had cut a deal and accounted for 30% of the options. And in many cases, mobile search is competing against two other options - Google on your PC (good for non time-critical things like dry cleaners), or Google search & maps on your handset.

I tried a new option out today - BT's Exchanges app for the iPhone, apparently developed by a firm called Locayta. So, I tried it out on my local area in central London.

Immediate "FAIL". Under the tab 'pubs', it misses my favourite local - as well as three others in the area. Under "petrol stations" it gets the closest one right, but also includes the location of another that shut four years ago. On the first screen of 10 hospitals, it misses the enormous new University College Hospital which has full casualty facilities - because there are 8 private clinics on Harley Street that are 300 yards closer to me.

Google Maps is a bit better - it gets my local boozer OK, and doesn't list the long-defunct Texaco. But it also doesn't get UCH on the first selection(maybe I'm being unfair because Harley St is a bit of an anomaly).

But best of all is good old Google Search, for which "Pubs Baker St" immediately throws up beerintheevening.com, which lists everything locally - with unbiased reviews, maps etc. A similar strategy yields UCH at #6 on the main search page - and #3 with "hospitals marylebone".

Put simply - the source data that Google Search uses is , it seems, far better for these types of look-up action. If I'm locked out of my house and need a locksmith, then maybe a yellow-pages type thing is useful. But for most searches - you already have all you need.

Read More

A quick thought on operators and APIs

Operators want to sell access to certain capabilities and assets in their network. Many are looking at exposing APIs to developers, enterprises or content providers. The GSMA is pushing its OneAPI programme as a lowest-common denominator set of offerings, while most major operators have major efforts for proprietary API exposure.

In general, I am a believer in this - and it goes to the core of many of the two-sided business models expounded by my partners at Telco 2.0.

But... if the general notion of API exposure and consumption is truly believed by those at the network operators, shouldn't they be "eating their own dogfood"?

As a starting point, wouldn't you expect the amount of API value to be imported from other sources, to be roughly equivalent to the amount exported to these new customer groups?


Now obviously the operators' large databases and infrastructure probably gives them a natural bias here, but that's not to say that similar sources of convenience and value lie elsewhere, which could be bought in.

If APIs are genuinely *that* valuable & available in the short-term, telcos ought to be avid purchasers of each other's capabilities, or those from banks or Google or Facebook or numerous vertical-market specialists. If customer data is valuable, shouldn't a mobile operator want to get access to fixed customer insights, and vice versa?

NEW Mobile Broadband Traffic Management Paper

NEW Broadband Business Models Strategy Report

Read More