Thanks to those of you who commented on my essay last week on applying The Goal to telecom. I think some misunderstanding (my fault) occurs around what is meant by “telco”. I’m very much of the same sort of opinion as David Gurle, head of collaboration services for Reuters, who in a September interview with C|Net News.com says:
Then where is IM going? Will it remain a service with millions of loyal users and no way to make money off them? The biggest observation I’ve had in the last two or three months is that IM is the perfect disruptive technology to telecom providers. […] In a sense they are the future local exchange carriers. That will create the connectivity requirements. The more value-added services on these respective networks, the more there will be a need for connectivity because it will become essential for people’s lives.
You’re saying that AOL, MSN and Yahoo will become phone companies?
Yes. I’m saying that Microsoft, AOL and Yahoo will become the equivalent of phone companies of the future, and with a reach that will go beyond their reach today. That reach will be global and without boundaries. With phone carriers, they can only go where the wire goes.
Can they make money like the phone companies? I think the next thing that will happen is they will bundle their value-added communications services with the Internet access businesses. […]
So the communications hubs of the future are the uber-ISPs. I’m already taking it for granted that end-to-end IP is an extinction-level event for hard-wired vertically integrated telcos. The assets of the telcos will go into the bankruptcy laundromat. (Unless of course, like the airlines, the government steps in and a taxpayer expense endlessly bails out incumbents and keeps out prospective foreign competition.)
Funnily enough, what comes out of the wash at first glance looks remarkably like what went in. Some of the brand names may be familiar, because the bankruptcy machine keeps brand colors bright and fresh-smelling. But deeper investigation reveals huge differences. People will buy transport from their ISP because they want to, not because they have to. Not all transport is created equal, and the ISP will tailor the performance, availability, ubiquity and price of their transport offering to the service bundle being provided. Convenience of purchase matters.
For instance, will my TV-over-IP service work with my DSL connection? Does my remote data archive system fit in my bandwidth cap? Does my wireless transport have the coverage I need for the places I go? Vertical integration may occur, just at the retail rather than the application level. (Layer 8 of the ISO stack!)
Going back to our airline analogy from before, we’re entering the era of the package holiday. We’re at about the equvalent of 1970 in terms of the airline industry. Buying your own transport and service separately is possible, but pricey and time-consuming. Travel agents arose to take some friction out of the market. Look at the mess of regulation and incompatible standards. Apparently simple issues like roaming between WiFi networks will probably remain thorns for years to come. Eventually (after a decade or two) the purchasers will mature, and the equivalent of the low-cost airline will emerge for transport. This could, for example, be a result of mesh networking. By then consumers will understand what they need and what they give up by not adopting a full-service product. Eventually more people demand private, secure pipes (the private jet) and QoS (the business-class seat), although these will not be mass-market products.
So Yahoo!, AOL and MSN are the communications travel agents of our age. They’re the ones who take our money, even if they don’t provide all the elements of the service. Eventually they too will be disintermediated. Every centralized function slowly gets pushed to the edge. But not for a long time, and ultimately anything that relies on a contract needs centralized hooks into the analog world of governments, courts and lawyers. The average user doesn’t have the sophistication to ask how many redundant transatlantic circuits the connectivity service offers, yet.
The “telco” I’m referring to is closer to what today we regard as an ISP. Bear in mind that an ISP is not the same as an access provider. Access will probably cease to be a retail business, and will become a wholesale business. The goal of a wholesaler is to fill the capacity of the fiber or spectrum. That’s not the business I’m referring to by “telco” in the new world. By telco, I meant the retail face of the communications business. I guess we need some new vocabulary. Wholesale transport is “data haulage”. We don’t pay separately for the transport when we buy goods from the supermarket — it’s bundled into the cost of service. Tomorrow’s telco is probably better referred to as a “personal communications retailer”. Not very snappy, better suggestions invited. The words “service” and “access” are verboten, mind you.
Today’s telcos are more complex beasts that it appears at first glance. There is a value chain in telecom that comprises more than just transport and service. Discrete elements include provisioning, billing, identity issuance, abuse management, customer service, equipment retail, content retail, and routing/directories (e.g. ENUM). Each of these is up for grabs. There is no reason to vertically integrate these well-defined elements. The integration and transaction costs have fallen too far.
The idea that the connectivity wholesale provider will engage in none of these is as unlikely as today’s telco remaining in all of them. Some incumbents will be particularly good at small slices of the value chain above raw transport, and will keep hold of those pieces. But most functions will move into the ISP business. So it is perfectly reasonable for a next-generation “telco” to have a communications-centric Goal, because that company is in the service retail business, not underlying network connectivity business. The majority of usage will come from a modest pool of dominant applications. The power law will apply. A huge tail of niche interest, reachable from Internet access point, will fill up the rest, and for that you’ll rely on your raw transport to get to it on a best-effort basis.
Deals like SBC Yahoo!, BT Yahoo! and Verizon/MSN are very much the way of the future. Each of these may have slightly different allocations of the value chain between the connectivity wholesaler and the service retailer. The smart incumbent will re-allocate the value chain voluntarity, before circumstances force a move. Earthlink has the right model, just they arrived in the market too early and had to pay too high a price for wholesale connectivity. Their leading the anti-spam and non-pop-ups approach shows they align with the goal I outlined.
The wireless side of wholesale transport will be slower to occur. Today’s wireless networks vary between hopeless to useless at delivering IP-based real-time applications. Sometimes this is for technical reasons (e.g. high latency, lack of multicast support). Often it is for hard-to-resolve business issues (e.g. chicken-and-egg between handsets and networks for introducing new technology — this needs a large intermediary integrating the value chain and committing to simultaneous orders of network and handset equipment; or roaming on WiFi).
So, to respond to some specific comments:
“We do a terrible job of protecting the user from unwanted communications”
Depending on who “We” are, and assuming you mean the operator of the network, that’s because it is not the network operator’s job to.
As I was saying, this is the job of the communications services retailer, not connectivity wholesaler. A super-smart traditional telco might morph to be such an entity.
Cliff notes that telcos are today service providers:
I don’t know of a single Telco that acts as if being a Network Operator is their only job. Telcos sell loads of services, and make most of their money doing it.
The critical thing to see here is that it is not a binary dumb pipe vs. service provider. There are many parts to delivering a service. The question is not whether the access provider will retain any of them, but how many are naturally aligned with the job of access wholesale and retail.
As David Anderson points out:
Martin’s point here is that when carriers start to focus on their true goals and make investments congruent with those goals, we’ll start towards a much more user friendly telecommunications service. A service aligned with subscribers goals will surely survive attack from insurgents - and hence the “telepocalypse” might be avoided.
I just wouldn’t put my money on it, based on my experience of working in the industry. But it’s fun to imagine what today’s telcos could do to rescue themselves. It makes for an interesting day job too, if somewhat of an unhill battle. (Bear in mind the British tendency for understatement, here.)
In both your phone call and spam examples, (and if you know what spam is in some absolute “agreed by all sense,” you’re doing much better than most), it is only the end user who can decide.
Not at all. Whilst Bayesian filters at the end points are great for personalized spam control, the ISP and access provider are uniquely positioned to examine patterns of abuse usage that cover multiple users. Furthemore, they are greatly advantaged in being able to punish outbound abusive/interruptive activity, by disconnecting access service. Some activities are naturally “end-to-end-to-end”. End-to-end doesn’t mean the extinction of all intermediary roles. For instance, Visa are still needed to clear credit card purchases on the Internet.
Thus it is the job of the network to deliver as cheaply and easily as possible, leaving money in the recipients hands to purchase services/products to do what protection they decide they need.
Yes, but much of that will still be purchased from the access provder. The cost will be much lower than today, because the possibility of buying transport and service independently exists. Just like number portability sucks (undeserved) profit from telcos today, because the possibility of moving exists. For another example, a socially scalable version of Skype that physically rings something in your house needs a hook into the analog world to punish misuse. That will cost money. Skype will not always be free, because that part of the value chain inherently links into the analog world and costs money. Just like the freeness of e-mail is an illusion; the infrastructure cost is bundled with access and the time and expense of spam management is not zero.
… let’s be really careful when we talk about messing with the middle of the network. The Internet became the success that it is today because it is an end-to-end network.
Retaining intermediary roles is not the same as violating end-to-end. Assembling a bundle of communications services and access means (including user equipment) that meets customer needs is a valid goal. Increasing the number and value of user communications is a way of expressing that goal. The occasional smart service provider may even get to own their own network(s) and not have to pay middleman wholesale fees. But that will be the exception, not the rule.
In an any-service-over-any-network world, where the network is general purpose and carries everything, the provider of connectivity has no advantage whatsoever in the sale of services, and no mandate to sell ‘em.
This is true is (i) the utopian world of cheap ubiquitous IP networks suitable for any reasonable purpose exists (it doesn’t, and won’t for a long time), and (ii) the value chain and profit pool comprises only access and service (it doesn’t). Access providers and service retailers retain some natural advantages. Sometimes, indeed (horror!) vertical integration with the network is needed. For instance, Nextel’s Direct Connect service required vertical integration with the network. The technology of the day simply didn’t support an IP version. Even today, VoIP equivalents from Sprint and Verizon offer superior functionality (a function of IP) with lower performance (a side-effect of lack of integration with the lower layers of the network). It isn’t invalid to seek to fulfil user needs that are at the edge of what technology can provide.
David Anderson writes: “Imagine you lived in an area with a lot of toll roads like Chicago or Washington DC and 9 out of every 10 cars on your sluggish commute was a “spam” car using bandwidth and not paying the toll.”
Let’s consider why [the road congestion analogy] is inappropriate…
To me, the bandwidth that matters is that of the user. And that is, for practical purposes, fixed. My Goal is to maximize the user’s communications bandwidth. The size of the network pipe isn’t the issue. Plus, the Korean experience of throwing bandwidth at everyone just shows that abuse scales at the same speed. Korea is a top DDoS source because so many people have 100 megabit broadband at home and insecure Windows machines. You still need intermediaries to police the system in an end-to-end world.
What if I want to buy the ability to send/get messages at any amount, without being charged per message or bundle ? Is that impossible ? Will no one sell me that service ?
That’s what an all-you-can-eat bundle offers. The bundle size is really the statistical average use, not infinity.
I summary, I see the telco industry (after some bankruptcy recycling) fragmenting into the value chain components. Access is a wholesale component. There are more than two components in the value chain. The application service is only a fragment of the chain. Some parts of the value chain, but not all, will remain with access providers. There will be good economic reason for this. It does not violate the end-to-end principle. Most of the rest will be picked up by resurgent ISPs, who will become the user-facing retail side of the industry. (So you’ll have someone new to hate for poor service.) The goal for the retailer of communications is (say it one more time…):
To maximize the number and value of user communications events, whilst simultaneously protecting the user?s attention from unwanted interruption.
This will all happen much earlier on the wireline side, because technology constraints on wireless make the end-to-end approach temporarily impractical for many user-valued but demanding applications. On the other hand, that places wireless network operators in a much better position to dig themselves out of the hole by aligning to the new goal before the end-to-end network kills them. Plus Nokia and Sony will probably waste a few billion trying to muscle into the networked service space, slowing things down for a while.
Posted by Martin Geddes at 06:25 PMTrackBack URL for this entry:
http://www.telepocalypse.net/cgi-sys/cgiwrap/mgeddes/MT/mt-tb.cgi/98.
Thanks for the consolidated response Martin, and I won't pick at all of them, but in the matter of the end-point's ultimate knowlege (which you partially acknowlege) of what is and isn't spam to them I must persist...
"Not at all. Whilst Bayesian filters at the end points are great for personalized spam control, the ISP and access provider are uniquely positioned to examine patterns of abuse usage that cover multiple users."
Again the presumption that the core, operator, "telco," "government," whatever, knows what abuse is to multiple users. I'm not sure they even know for certain what one user's preferences are, and the minute you generalise and homogenise, your accuracy drops.
Yes, I see no reason why operators shouldn't offer "protection," but to impose it universally and without opt-out, is reducing the choice of the customer.
As noted somewhere, there are customers who REQUEST information to be emailed to them about viagra... The network, even the wisest, cannot tell.
Posted by: at December 4, 2003 07:44 PMI disagree, but then I would :-)
The information "this email, or one very similar, has been received by 12,000 other subscribers" is something that can be fed into your end-to-end spam filter -- but intrinsically requires a middleman. (OK, I'll relent, and accept that a peer-to-peer comparator and filter is possible, but considerably less likely and open to abuse itself absent of a remotely attested trusted computing solution. Another story.)
I think the key point is that "end-to-end" is not the same as "no middlemen". End-to-end is an architectural design principle. No more. It's perfectly legitimate for middlemen to exist, and also for the access provider to aim be to one of those middlemen. All without violating end-to-end.
For example, the Blackberry runs on multiple networks. In principle I could buy a Blackberry separately from network connectivity in true stupid network style. But if I'm not receiving incoming mail, I only want to phone one customer service number, with no buck-passing between the connectivity and service provider. Hence it is natural for the network operator to also be the retailer and tier-1 customer service provider for this device. But the service operation and follow-up support would come from RIM.
The effect of end-to-end is fragmentation and re-allocation of the value chain; not simple displacement of all network operator activities, even if sometimes that is the case.
Perhaps "Middle-man" is itself an outdated concept for some of this. Imagine the Telco, not as a middle man but as a Super-end, and that the model is of ends-to-ends rather than end-to-end.
When you use a VISA or AMEX card, the retailer checks against an Acquirer (who acts on behalf of an Issuer) for authorisation. But perhaps that's too much like a trusted computing model.
Let me have another go of expressing myself: Although there is no universal definition of spam, clearly some messages are unwanted and can be identified. In the same way I hope that the post office helps filter out radioactive, explosive or biological mail, an electronic service provider may be able to help filter out something like kiddie porn spam. Society does have limits of what is and is not acceptable, even if they are very hard to define or enforce. Users of a service may also have a definition of what constitutes spam, even though hard and fast rules for which maybe difficult to define. Spam filtering would not neccesarily be done by the 'telco'. Why not seperate out this service? As long as you have a sufficient number of subscribers you could form a filtering sevice regardless of who the other providers were. Filtering services could have a broad spectrum of what is acceptable, everything from those who think swear words (like Scunthorpe - think about it) are out to those who think pictures of people having sex with goats on fire are O.K.