This old weblog article from 2008 discusses the cost structure of a typical mobile operator, and I show below the structure graphically.  The key figure is transport costs (backhaul) are roughly 7% of total costs back in 2008.  Some operators have seen that jump to over 15%-20% of total costs in under 2 years.  The reason is the tipping point achieved in mobile broadband adoption, discussed in this article. 
Backhaul is the transport link from the radio base station into the mobile network's core.  Its generally leased line (copper or fiber), microwave or metro Ethernet; the choice is highly influenced by the incumbent fixed operator's role in the mobile market, level of competition and local regulation. For example, BT has been aggressive in its metro Ethernet pricing to keep its backhaul business in a competitive market, while in the less competitive French market France Telecom is able to keep backhaul costs significantly higher (outside Paris.)

At the moment we're seeing lots of excitement around LTE (Long Term Evolution), which requires operators to deploy a third radio access network (after 2G and 3G) and a separate core network that is claimed to somehow magically lower operational costs.  Back in 2008 this weblog article discussed that fallacy.  Unfortunately the NEP's marketing engines have kept on rolling in the face of common sense and we're now seeing some operators realize that its not an air interface issue, HSPA and HSPA+ are perfectly adequate for their customers' needs.  Its a backhaul issue, those 3 sectors running at 20Mbit/s require 50+ Mbps of backhaul, while there had only been 8 Mbps in place.  Quadrupling the capacity has doubled / tripled the transport cost, hence the jump in backhaul costs from 7% to 15-20% depending on the market.

To alleviate this problem Sycamore Networks have come up a very smart technology, IQStream - props to Scott Hilton, Talbot Hack, Yang Cao and the rest of the team for their excellent innovation.  As shown below it can provide up to 90% capacity saving in the backhaul, of course the design rule will be more conservative. At least Sycamore Networks are focusing on the key part of the problem, rather than encouraging operators to make decisions not in their best interest.  For the sake of our industry the NEPs must stop conning operators, and act in their customer's best long term interest not their own short term interests.

North America is an exception in the mobile broadband (MBB) market amongst developed nations.  At the end of Q1 2009 there were 17.3 million data-only HSPA subscriptions in Europe, 14% of all broadband connections are now mobile and mobile is growing at 5 times the rate of fixed broadband.  Depending on how you calculate it, the MBB connections in the US are somewhere around 4% of all broadband connections - and the overall broadband penetration in the US is only 60% with the US at number 20 in the broadband league.  At the end of 2008 mobile operator 3UK was the third largest UK ISP (Internet Service Provider) after the incumbent operator (BT) and cable provider (Virgin) with nearly 1M MBB subscriptions; the merger of ISPs Carphone Warehouse and Tiscali knocked 3UK off that spot.

Just comparing MBB contract plans:

• At the low end in the US 250MB costs $40, while in the UK 1GB costs $7.50; and
  
• At the high end in the US 6MB costs $60, while in the UK 15GB costs $25 - and even netbooks are bundled with the plans in the UK.

The US is between 6 to 20 times more expensive than the UK!

Roughly 95%-97% of MBB traffic is from laptops/netbooks (source EU operator).  Mobile broadband is now a mass market service in many developed markets.  Data-only subscriptions involve either a USB dongle (>90%) or embedded broadband (<10%), data cards are no longer relevant in most markets.  Customers prefer the USB dongle as it can be shared between devices and customers.

The USB modem has created a separation of network and device, minimizing many scaling issues faced by smaller operators.  This enabled small operator 3UK (roughly 4M customers) to disrupted the UK market with its aggressive MBB tariffs, bundled netbooks and prepay mobile broadband.  In the UK, customers now expect a netbook when signing up for a MBB contract - the US gap is not just the 600 to 2000% higher price!  The European netbook market in Q4 '08 accounted for 20% of laptop sales, in Q1 '09 this accelerated to 29% - that's >5M units per quarter principally driven through mobile operators.  Below I show a comparison of MBB plans that include bundled netbooks from Broadband Genie. 

The US has a long way to go before it reaches the value being offered in other markets.  The reasons behind this are ineffective regulation (e.g. local loop unbundling regulation in the US in embarassing compared to other countries) and we have essentially a duopoly in fixed and mobile broadband.  Something needs to change.

In the article Reinventing the On Device Portal - The App Store, I reviewed how Apple's implementation of its App Store on the iPhone provides a template for operators and how the struggling ODP (On Device Portal) can come into its own. 

I've reviewed the ODP landscape, its struggles and evolution in previous weblog articles.  Originally created to improve the operator's walled garden portal experience, it was over-sold as providing a solution across all devices.  In practice, the user experience of an ODP on a significant minority of devices was a good way of deterring customers from data services.

The Operator's App Store is not a new concept; there are early adopters, for example: Verizon AppZone is built using mPortal's ODP.  The critical issue is not technology; operators must simply commit, pre-load their ODP App Store, and have an integrated storefront strategy.  Fortunately, given the processing power in phones today, most devices are now addressable by ODPs.

Cricket's MyHomeScreen is an excellent example of such as implementation, supplied by mPortal.  Firstly, its preloaded; its front-and-center of the customer's experience, see figure below - its the overlay bar/carousel with cute icons; its much more than a widget engine it has a back-end to provide a unified storefront and integrated into the operator's back-end systems.  Services included in MyHomeScreen:

• Website widget, and of course any website can be presented as a widget
• Storefront widget for graphics, tones, themes, games or ringbacks.  Here Cricket can aggregate a number of catalogs to present a unified storefront - see the The Emerging App Store Ecosystem article on why this is important to operators;
• Account status widget to see the prepaid balance, call detail records, status of orders, etc;
• And of course the the usual weather, news, gossip, entertainment widgets;

Cricket has shown the industry how to tackle the consumer electronic and OS app stores head on.  Rather than complain about Nokia Ovi, operators now have a template on how to deliver an integrated storefront experience across the web, content, games, customer relationship management and all the other services they provide, all front-and-center of the customer's phone experience.

In my journeys around APAC (Asia Pacific region) over the past few months, a common theme is the explosion of mobile broadband (MBB), as reported in previous articles in this weblog such as Mobile Broadband Explosion, The Internet's Gone Video, and Mobile and DSL Broadband Compared.  However, the impact of that explosion on the operator is quite different from country to country.  In this article I'm just picking on four countries: Hong Kong, Singapore, Australia and the Philippines; to compare and contrast the impact MBB pricing and MBB terms and conditions (T&Cs) have on customer behavior.

Hong Kong: Mobile broadband as a complement

In this market the two main mobile broadband plans termed 'Unlimited Local Usage Plans' are:
$188/month ($24 USD) up to 3.6Mbps down/1.5Mbps up; and
$348/month ($45 USD) up to 7.2Mbps down/2Mbps up.

Local Usage refers to within HK (Hong Kong) region, as roaming is common given the relatively small size of the region.  Usage is limited to browsing, email, intranet access, streaming, uploads and downloads, VoIP and instant messaging.  The fair usage policy blocks peer-to-peer file sharing, machine-to-machine applications, and use that unfairly exploits SmarTone-Vodafone.  

In talking with HK residents it's apparent that the T&Cs have impacted behaviour.  The limitations have curtailed the savvy early adopters that look to exploit arbitrage or 'open door' opportunities.  Mobile broadband is generally perceived as a complement to fixed broadband in the home, rather a replacement.  Hence the availability of a daily use tariff plan of $68/month ($9USD) + $18 ($2USD) per day usage for up to 7.2Mbps down/2Mbps up (with a 1GB per month limit, after which there is an additional usage charged at $2 (26c USD) per MB.

Singapore: Mobile broadband war

In this market the focus has been on true unlimited data plans, without the constraints on traffic types, with pricing is similar to HK:
Monthly Subscription /  Download Speed / Upload Speed
$22.42 ($15USD) / 1Mbps / 512kbps
$27.10 ($18USD) / 2Mbps / 512kbps
$35.95 ($24USD) / 4Mbps / 1Mbps
$48.51 ($32USD) / 7.2Mbps / 2Mbps

In talking with the residents the 7.2 Mbps MBB package is perceived favorably to the 8 Mbps DSL package at $45, because of the higher upstream rate of 2Mbps compared to 512 Kbps for the DSL service which significantly improves peer to peer performance.  As MBB usage on an individual basis increases, and the numbers of customers taking the service also increases, it will be interesting to see whether the 'profitable majority' can continue to subsidize the 'unprofitable high-usage minority.'

Philippines: Almost giving it away, but not...

Smart in the Philippines has pricing simplicity nailed, 10 Pesos (20c) for 30 minutes.  Let's assume a mean throughput of 100 kbps, that's 22 MB for 20c, which is the equivalent of $7USD for 1GB, with no apparent limitations on traffic type.  With the faster upstream rates compared to DSL broadband (which has the same pricing structure in the Philippines) MBB presents an attractive proposition for the individual user.  However, for those who download 100GBs, or remain connected 24/7 then the bill can rack up to a cap of $288 USD.  So Smart's usage based pricing provides a simple to understand model which imposes a natural cap for the 'big downloaders.'

Australia: Training customers with Usage caps

Australia is in the middle of a bragging rights competition with respect to who's got the biggest mobile broadband pipe.  Telstra recently demonstrated HSPA+ (21 Mbps) on their network, and are on course for launch early next year, as promised at MWC (Mobile World Congress) in February of this year.  Australia's usage caps avoid the hassle of defining traffic types, below I show a sample of MBB pricing.  With DSL broadband available at similar pricing and caps to wireless, the critical decision for the customer is whether the broadband service needs to be shared, i.e. broadband for the family at home.

In talking with residents, firstly, they're aware of the premium they're paying compared to other markets, especially when I described my FiOS package from Verizon :)  But what was interesting was the variety of packages people choose.  For example, some would choose a wireless package for the family as they have a weekend apartment, so it was cheaper to take the wireless option, than two fixed packages.  Some had both, fixed for the family and mobile for business; and a few MBB only.  The overriding feature of the market was regardless of their package they were all highly aware of how their online behavior impacts monthly usage, to the point in a few cases of stifling their usage.  I didn't see this awareness in the other markets reviewed in this article.

Telstra Pricing (at low usage wireless/DSL packages are comparable, but at the top end DSL has the edge)
$29.95     ($19 USD)    200MB
$59.95     ($38 USD)    1GB
$89.95     ($57 USD)    5GB
$129.95   ($83 USD)    10GB

Optus Pricing (comparable between DSL and wireless)
yes' Wireless 2GB     2GB     $39.99 ($25 USD)
'yes' Wireless 5GB     5GB     $49.99 ($32 USD)
'yes' Wireless 6GB     6GB     $59.99 ($38 USD)

Three (Note there's a coverage issue outside the main cities, which means customers roam onto Telstra and pay a roaming fee)
1GB    $15 ($9 USD)
3GB    $29 ($18 USD)
6GB    $39 ($25 USD)

Conclusion

The exact figure at which a customer becomes unprofitable for mobile broadband varies greatly between operators; i.e. on the service pricing, the size and distribution of the network, and the cost of backhaul from the incumbent fixed line provider, etc.  However, as a very rough rule of thumb, if a customer is going north of 30GB per month, then there's a fair chance they're unprofitable.

In summary:

• Hong Kong and Singapore provide a good contrast on the power of T&Cs in controlling certain traffic types that limit high use early adopters, but also how that impacts how the market views MBB.  With people in Singapore viewing it as a competitive DSL replacement. 
  
• In the Philippines the power of very simple usage based billing provides a natural brake to high usage.  
• Australia demonstrates the power of getting people to think about their usage.

As a customer, Singapore provides an ideal market, a competitive price, no need to worry about usage, and the freedom to do whatever I choose.  Philippines is an interesting model, especially for the majority of users who are not intensive broadband users.  The Australia model has the greatest spread in pricing across the operators, it will be up to the customer to vote with their wallets on where the caps and pricing go.

This article follows on from a previous article that reviewed LTE (Long Term Evolution, also know as 4G).  Several people asked why I had not included a fuller discussion on the 4G core in that article.  It is because LTE is an access technology; my focus of the article was on the access where there is a lot of hype at the moment.  However, to support LTE a new core is required, as specified by the SAE (System Architecture Evolution). 

In the standards community the names have now become E-UTRAN (Enhanced UMTS Terrestrial Radio Access Network) for LTE, and EPS (Evolved Packet System) for SAE.  Think of the SAE as a pure IP core without RNCs (Radio Network Controller) or SGSNs (Serving GPRS Support Node).  So in principle an operator would need to run two core networks to support the existing 3G network and the new 4G network.  Given the slight benefits of LTE compared to HSPA+, as discussed in a previous article, this would appear to be a significant barrier for LTE adoption.

However, the NEPs (Network Equipment Providers) are providing a range of solutions that enable the existing 3G core to evolve towards the SAE vision such as:

• One Tunnel, moving the SGSN into the control plane;
• Direct Tunnel, moving the SGSN and RNC into the control plane; and
  
• Internet HSPA: control-plane SGSN and RNC is integrated into the NodeB.

The 3GPP core market is roughly $1B in size, and the top three suppliers in this market are: Ericsson (34%), NSN (33%), Huawei (15%).  Who unsurprisingly are the three players behind the three different solutions that evolve the core towards the SAE vision.  Given we're likely to see the number of mobile broadband customers worldwide explode to over 1B over the next 4 years, mobile core costs could potential to explode as well.  Hence the immediate focus on moving the RNC and SGSN out of the data path and into the control plane, to better manage broadband growth.

Specifically the motivations for this early transition are:

• Lower cost/bit - depending on the architecture it has the potential to reduce equipment costs by 80%;
• Preparation for 4G - in other words avoid a dual core situation which would delay operators buying LTE BSRs (Base Station Routers).  Note the base stations are where NEPs make their money, so they're keen to remove any barriers; and
• Ease of integration of non-3GPP access (i.e. network convergence, avoiding multiple cores and service platforms).

Hopefully this sets out why we'll likely see the '4G core' happen before the '4G access.'

LTE (Long Term Evolution, also known as 4G) is entering the hype-cycle, and being rapidly pushed up to the 'peak of inflated expectations', just like IMS (IP Multimedia Subsystem), PTT (Push To Talk), FMC (Fixed Mobile Convergence), and far too many other technologies to name here.  So to cut through the hype and hopefully present a pragmatic view on the technology I thought it timely to make a few points, as I find I'm often repeating myself on this topic (I hope its not old age.)

A few pointers on what the technology can do:

• LTE will not greatly improve spectral efficiency compared with HSPA+ (High Speed Packet Access).  Within a 5MHz slot you could achieve perhaps 80 Mbit/s with HSPA+ (using MIMO (Multiple In Multiple Out) technology), compared to 100 Mbit/s with LTE. A 20% increase is unlikely to have customers demanding LTE.  The only tangible difference will be a slightly lower round-trip delay for LTE perhaps down to 10-20ms compared to perhaps 30-35ms with HSPA+, though with a flat IP core (moving RNC (Radio Network Controller) and SGSN (Serving GPRS Support Node) out of the data path) that could drop to about 20ms for HSPA+.
• LTE enables a higher peak data rates by using more bandwidth, scaling from 1.25 MHz (useful for the CDMA operators, hence Verizon's decision to adopt LTE) to 20 MHz and beyond.  LTE uses OFDM (Orthogonal Frequency Division Multiplexing), the same technology used in DSL (Digital Subscriber Loop), that is lots of little carriers, rather that one big one carrier as in HSPA, which gives LTE its flexibility.
• LTE can use FDD (Frequency Division Duplex) and unpaired TDD (Time Division Duplex) bands bought during the heady days of 3G spectrum auctions, where HSPA and GSM cannot operate in those bands
• LTE is a global standard, with a global frequency plan, hence can leverage global volumes, another reason behind Verizon's decision to adopt LTE, rather than pay an average $15 premium for its (CDMA) devices.  This is also the fundamental reason why the total cost of ownership of WiMAX networks will remain more expensive than LTE.
  
• HSPA has some funny technologies hidden away in it such as macro cell diversity, i.e. a terminal may use 2 or 3 cells at the same time. LTE does away with these complexities so does not need RNC nodes. LTE also has built in operational and maintenance capabilities, such as self-tuning to help control operational costs.
• By 2010 most GSM base stations in Western Europe will be over 20 years old. Most were put in during the "cheap-energy days."  Today, finally, green credentials matter to companies so LTE will enable the power consumption of mobile operators RANs (Radio Access Network) in the long run to be reduced by up to 50%.

A few pointers on the market's development:

• DoCoMo and Verizon are in unique positions that force a move to LTE faster than the rest of the market, so by 2010 they will likely be deploying LTE, and will be the first movers.
• Most in the GSM community will be working through HSPA+, which will significantly delay their need to move to LTE.
• The Flat IP core I referred to previously is a way of leapfrogging to the SAE (System Architecture Evolution), which can be thought of as the 4G Core, while LTE is the 4G RAN.  With the core upgraded, and a relatively fast RAN in HSPA+, most operators are only going to deploy LTE based upon operation needs.  Hence you'll likely see LTE being deployed more broadly in the market around 2013.
• And of course the good old chestnut of handset availability.  This is actually likely to be less of an issue compared to say 3G as most of the customers using LTE will be using a laptop with a USB fob.  Just remember to bring a spare laptop battery when you use LTE!
  

A mobile broadband Tipping Point has been reached: HSPA (High Speed Packet Access) is 'good enough' for customers and prices have fallen by a factor of 16 in less than two years.  Three UK (H3G UK) now offers mobile broadband prepay, 10 GBP ($20) buys up to either 1GB or 1 month of access, and its prepay so for regular travelers are no longer subjected to the extortionate 15 GBP ($30) per night internet access charges at UK hotels.  Specifically on the 'factor of 16' data point using Three UK as an example, in '06 the price per MB was 8p/MB, it is now 0.5p/MB (contract package 15 GBP per month for 3GB).  Mobile broadband is growing faster than any previous service, including voice.  Operators are talking about annual subscriber growth rates of over 400%, in one case 100% per month!  It reminds me of the good old 'mobile-gold-rush' days in the '90s, and aren't we supposed to be in a recession?  Mobile Broadband is generally a USB HSPA modem connected to a laptop.

However, the problem is this has happened at the same time as two other trends:

• The internet has gone 'video,' so data traffic is growing exponentially on these broadband ISPs; and
• Computing has gone personal; laptop penetration in some countries is close to 40%.  As data points: the Asus Eee laptop costs between $300-$400, and there will be roughly 40 million laptops sold this year in both the US and Western Europe.

In the UK, the BBC iPlayer is causing fixed broadband ISPs (Internet Service Provider) to complain, customers are now regularly consuming multiple GB per month through their ISP, most of it watching video.  Also new internet-connected HD (High Definition) devices such as the Sony PS3 provide access to games demos and video trailers (all available for free to the customer) where one demo often requires >1GB of data.  The fixed broadband ISPs are now revisiting their network economics.  

For Mobile Operators they the advantage at the moment that customers do not yet expect 'unlimited' mobile broadband access, so the 1, 3 and 7GB limits are not yet a significant deterrent.  However, even with such limits and customers now more readily filling those limits, hence mobile operators must change their network economics from 1c per MB to 0.1 c per MB.  Note, a typical smart phone uses about only 10MB of data per month on average, so laptop access has the potential to increase usage per device by a factor of 100!   The average mobile data bit rate per subscriber was about 10 kbit/s per session, a fixed broadband ISP's average was 25 kbit/s ('07).  Fixed broadand ISPs are estimating their average rate will grow to 70 kbit/s by (2010 or 2011).  The driver as mentioned before is video consumption over HTTP, iPlayer (P2P video), P2P (Peer to Peer protocol) sharing, Sony PS3 downloads, etc.

This tipping point creates a number of opportunities.

On the services side:

• The Mobile web is not just about smart-phones; it's about mobile laptops.  The web is not just about desktop PC access, it's about mobile laptop access.  The use cases are different, hence the opportunities.
  
• With Bluetooth coming as standard in most laptops sold in Western Europe the opportunities for VoIP bypass are significant.  But it will be interesting to see what substitution really takes place.  It can not replace my mobile phone, but it could help save me a few dollars for expensive international calls when I'm settled at a coffee shop.
• Those places where people sit and wait: train stations and trains, coffee shops, airports, in the back of the car, etc. will be far more likely to have sophisticated internet devices connected.  Presenting opportunities in customer service, customer retention, marketing, ways of selling and doing business.
• For enterprises this also changes the mobile office, enabling greater transparency and security of enterprise applications.
• And as discussed at length in other articles on this weblog, provides new opportunities for the Telco API to enhance the web experience of those mobile laptops.

On the network side:

• The core network of mobile operators is going to need to flatten, and do so quite rapidly.
• The backhaul network in mobile operators is already a well covered topic, announcements between O2 and BT show the solutions being adopted.
• HSPA+ is required not for speed, but for the capacity it offers to meet demand in high traffic zones.  HSPA+ was discussed in my Mobile World Congress Summary.
  
• The usual device delays that 3G and HSPA has suffered, may not impact LTE (Long Term Evolution) as greatly as its first application may not be a smart-phone, rather a USB mobile broadband modem powered from the laptop.

Ovum predicts the number of HSDPA connections will reach 16.5 million by the end of 2008 in Western Europe, thanks in part to the simplicity of plugging in a USB HSDPA modem and it just working rather than the complex configurations and reboots of a data card; as well as the simplicity and attractiveness of the pricing plans at just 15 GBP per month.  Mobile operators are now working on providing value added services on top of this mobile broadband access.  One such service is secure remote access (SRA), i.e. enabling secure access to the corporate LAN and protecting remote corporate devices (laptops and smart-phones).  SRA can add $15-$30 per device per month in ARPU (Average Revenue Per User).

What would appear at first blush to be just a matter of using IPsec (IP Security) or SSL (Secure Sockets Layer), has spawned a whole industry comprising the following groups:  

• Dedicated SRA providers such as iPass and Fiberlink providing a package of dial-up, Ethernet, WiFi and HSDPA access options bundled with a security package for the global traveler;
• Mobile operators' offers range from HSDPA and a simple connection manager, e.g. O2's Mobile Connection Manager; to complete packages of managed security and remote access, e.g. Vodafone's Secure Remote Access;
• Fixed and Mobile network operators offers similarly range from access centric to a complete managed solution across the corporate LAN as well as remote access, e.g. Orange Business Services (which is in part a resell of iPass) and BT MobileXpress;
• IT security vendors such as Symantec and Checkpoint with the unified threat management systems (both LAN, remote access and remote offices);
  
• Both local and global IT system integrators that package together solutions from the above providers; and
• The enterprise's IT department.

To expand a little more on what SRA covers: for global access it's about providing connectivity to the corporate LAN and internet through WiFi, hotel Ethernet, dial-up and mobile broadband.  On security it covers: anti-virus, firewall, anti-spyware; encryption of traffic between the PC and corporate LAN; multi-factor authentication; data protection including encryption, device control, information policy, back-up and restore; one stop connection client; policy management and control of end user behavior to meet IT policies; management reporting; central portal providing IT with improved visibility and control; all backed up with Professional Services and Help Desk.  

So which supplier will dominate?  Based on a recent enterprise survey I ran in the UK, Germany and the US it very much depends upon what the enterprise is buying.  If it's a security solution, then it will be from an IT security vendor or a managed solution from an operator or IT system integrator.  If it's remote access, then the operator (fixed and/or mobile) is the likely choice.  For secure remote access, the choice is a little more complex, and will depend more upon the type of enterprise and its situation.  For example, a large financial services enterprise, with a large IT department will have a tendency to build a solution from an IT security vendor.  Though with that said we are seeing departments within large enterprise select an operator's SRA for their team. For a SME, which has a LAN security solution, but not the remote access component, a solution from an operator will likely be their choice.

This situation puts the operator in an interesting position, it can choose to be:

• Just an access provider, which sets it on the path of being a utility;
• Find a mix of security and access that meets a segment of the enterprise, which is difficult as enterprise IT providers already bundle remote access security solutions into their unified security solutions; or
• Deliver a complete managed access and security solution, which given Verizon Business bought CyberTrust and Vodafone bought Aspective shows this is a growing focus for those operators strong in the enterprise segment.

Perhaps secure remote access will be the tipping point upon which an operator breaks-through and delivers managed services to its enterprise customers, or secede the space and focuses upon access and being a utility.

Some DSL (Digital Subscriber Loop) providers have certainly led the industry in communicating rates for their service that are unlikely to ever be experienced by their customers, with statements such as "Speeds upto 50 Mbit/s,"  note that all-important 'upto' term.  Ofcom in the UK has started an investigation into this practice.  Physical loop limitations mean that unless your house is next to the local exchange, the actual throughput will be much lower, typically one half to one quarter of that rate.

Earlier this year I did a quick survey examining the mobile industry and its claims as multiple wireless technologies are offered to customers.  The US is focusing upon average download speeds and defining the conditions upon which those speeds are achieved.  The rest of the industry uses 'upto' with in some cases guidance on what can typically be expected.

For DSL, through-put is dependent on loop-length, noise at the local exchange (generally Gaussian), concentration in the carrier's network, and congestion on the site being accessed.  For example, at home I know I get 2.8 Mbit/s rain or shine.  However, for my mobile data service the throughput is dependent on my distance from the tower, the number of people simultaneously using that tower, whether I'm indoors, the weather (fading is Rayleigh not Gaussian, so sometimes it just does not work), which device I'm using, and what smarts the operator has in their network to squeeze the data; as well as the concentration in the carrier's network and congestion on the site/service being accessed.

My experience with throughput on my mobile service is definitely more variable.  This raises the question does average download speed mean anything?  Upto speeds have been used in playing marketing games about size.  When really what matters is how long it takes to download an email with a 1MB attachment, or how long will it take for the BBC news website frontpage to download, or how often will a webpage or email download timeout?  It's definitely something that an independent agency, such as JD Power and Associates or Consumer Reports, should be evaluating and comparing across operators as mobile access to the Internet becomes as pervasive as mobile voice services. In an ideal world it would be nice to see a number of measurements around a region, for example NYC metro area, comparing operators' performance with mean and percentage of measurements above some 'adequacy' user experience level.