About Tom Evslin

Video Profile of Tom Evslin

Follow Tom Evslin on Twitter


Add to Technorati Favorites!
Powered by TypePad
Member since 01/2005


« The End of the Local Number | Main | DARK WATERS: Programming in Small Places »

AT&T: Lesson From the Crypt #3: Vertical Integration Doesn’t Work Anymore

Vertical integration is probably the most important reason why AT&T failed.  SBC’s planned acquisition of AT&T is a mistake, I believe, because it is a step toward vertical reintegration at a time when horizontal, not vertical, excellence is necessary for success and even survival.

Once upon a time, the computer industry was a model of vertical integration.  IBM manufactured chips and computers, wrote operating system and application software, taught courses and provided consulting, and even sold made and sold punch cards and time clocks.  Each of the seven dwarves, the companies that tried to compete with IBM in those days, was vertically integrated as well.

The came the microcomputer.  Society changed as computing power migrated from the splendid isolation that mainframes lived in to the desktops of end-users.  The business of computing changed as well.  It became horizontal. There are a proliferation of retail outlets on and offline; computers are manufactured all over the world from components whose sources are equally varied.  Intel dominates the chip layer; Microsoft the operating system and application layers.  IBM almost died as a result of this change.  Just in time, it adapted and is still alive and ticking although not its former dominant self.

AT&T at its height was a marvel of vertical integration.  It made phones, made and strung wire, owned the subs that tend undersea cable, launched satellites, made switching equipment; and, of course, provided local and long distance service.  It had a fleet of cars larger than Hertz.  It employed a million people.  It was so large that it insourced almost everything.  Most phone companies around the world were similarly vertically integrated.

It may have been that in yesterday’s world of poor communication, vertical integration was an advantage.  If you have a monopoly in  any single layer and can sustain that monopoly, then you can have even greater profit by moving into dependent layers above and below and using your monopoly position to assure dominance and high profits in those layers as well.  The carriers of the world were government-sanctioned and government-regulated monopolies if not simply a branch of the government.  They survived and prospered in their vertically-integrated form until the dawn of competition.

Vertically integrated companies can’t compete!  The oxymoron of “internal customers” is poison to a competitive culture.  That is the lesson of the computer industry and it is a lesson the telecommunications industry apparently has not learned yet.

At its launch AT&T WorldNet Service was able to get a competitive advantage as an ISP by being able to use 200 existing AT&T owned and operated dial PoPs (Points of Presence) which had been built for another service.  These provided the local numbers essential to making Internet access affordable and offering a fixed monthly rate.  Score one point for vertical integration.

Meanwhile UUNet, a very focused horizontal company, was busy building local PoPs; but they started out far behind AT&T.  UUNet was not an ISP itself.  Instead, it rented its PoPs out to retail ISPs by the minute to enable them to provide local access.  Early customers included Microsoft’s fledgling MSN as well as small local ISPs.  UUNet had many customers for its PoPs.  Only AT&T services, mainly WorldNet, could use AT&T PoPs.

The UUnet PoPs were new and state of the art;  AT&T’s were not.  The UUNet PoPs were not guarded by a union work force.  AT&T’s were.  So AT&T PoPs had a much higher cost per minute of use than UUNet PoPs.  As the captive “customer” for AT&T PoPs, we at WorldNet weren’t happy with the price we paid even though we appreciated the reliability.  But money to upgrade the PoPs and to build new ones would have to come from the budget for the Network Division (a different layer!), not from WorldNet. The Network Division had other priorities and potential business from WorldNet alone didn’t justify a very aggressive buildout.

UUNet had real external customers and built to meet their needs.  AT&T had only itself for a “customer” and so didn’t have enough incentive or enough business to build the world’s best network of PoPs.  Soon UUNet had more PoPs than AT&T.  That meant than a tiny startup ISP like EarthLink, on the day they went into business, had both more PoPs available and a lower cost per minute than AT&T WorldNet.  EarthLink inherited the breadth and economies of scale of its supplier UUNet.

Eventually WorldNet wanted to buy from UUNet to lower its costs and increase its geographic coverage.  But these purchasing decisions are made by corporate politics and not by the marketplace in vertically integrated companies – it’s called “serving the greater good” or “being a team player”.  WorldNet was not allowed to take its traffic off AT&T’s network and out of its PoPs.  WorldNet could not use its own size to achieve a good deal with UUNet.  WorldNet was now at a competitive disadvantage to EarthLink.  Score two points against vertical integration. Eventually AOL sold its own PoPs and used UUNet, among others, to provide access.  AOL did not let itself be a prisoner of vertical integration.

Meanwhile, AT&T was not selling access to its PoPs to other ISPs because that would have been bad for WorldNet.  AT&T, in theory, could have built a network much bigger, better, and cheaper than the one UUNet cobbled together.  In fact, UUNet did not hesitate to lease network connectivity from AT&T where it made sense to them to do so.  So, not only did AT&T lose the opportunity to be a great retail ISP, it also lost the opportunity to be a dominant supplier of local access.  Score two more points against vertical integration.

A horizontal company has a high surface to volume ratio.  It sells all of its outputs in a competitive market.  It is free to buy all its inputs in a competitive market.  Its managers are not isolated from the markets they compete in.  A vertical company spends most of its time and energy dealing with itself rather than the external market.  Meetings are dominated by esoterica like transfer pricing between divisions because this is what determines internal success.  The real marketplace is distant from most of the managers trapped inside the vertical structure.  Buy vs. build and capital allocation decisions inside a vertical company are made by office politics in a vain attempt to optimize across the whole vertical organization.  Horizontal competitors optimize only for the layer they are competing in and so end up being superior to the vertically integrated company layer by layer.

SBC clearly believes it can make vertical integration work.  I don’t think so but it’ll be interesting to watch and learn.

Lesson from the Crypt #1 is don’t manage for quarterly results.

Lesson from the Crypt #2 is you can’t innovate flawlessly.

Lesson from the Crypt #4 is don’t sent your losers to heaven.

Lesson from the Crypt #5 is navel gazing is a bad culture.


TrackBack URL for this entry:

Listed below are links to weblogs that reference AT&T: Lesson From the Crypt #3: Vertical Integration Doesn’t Work Anymore:

» Homeshoring and other names from Future of HR Services and Outsourcing
This past year's hubbub surrounding [Read More]

» Horizontal vs Vertical Integration from robhyndman.com
Tom Evslin writes about the death of vertical integration, a propos of considering the SBC / AT [Read More]

» Fractals of Change: AT from Aria Systems
Link: Fractals of Change: AT [Read More]

» Search Before the Purchase Report, Link Development, Webmaster Communities, Integration from SEO Book.com
DoubleClick releases a search before the purchase report. A few threads about link & community development. Link to a post about why horizontal integration is often better than verticle integration. [Read More]

» Labels are (Vertically) Dis-Integrating (As Well They Should) from Rags' Soapbox
Record labels continue to vertically disintegrate. [Read More]

» The Perfect Shopping Mall: Inside Google from SEO Book.com
Post about why I think Google has a better business model than eBay, or most other companies for that matter. [Read More]

» Full of ourselves from Jason-Preston.com
Theres a lot of doomsaying on blogs. For the past few days Ive been tossing a little newsblog aggregation idea around in my headsomething not terribly unlike Batelles Federated Media, but I just cant see it working.... [Read More]

» adobe p2p? from NextBlitz
Om Malik wrote this interesting blog post about Adobe's P2P ambitions. Read Om's excellent writing for the full story. The short story: Adobe owns Flash (via Macromedia), and just bought a small P2P company, and recently announced a relationship with [Read More]


Tom Evslin

Hi, David:

EarthLink was before WorldNet which launched (GA) on March 14, 2006 and was announced a little earlier. Our model for the allyoucaneat pricing was PSInet but I do know that there were other ISPs who had adopted that as well. Our role was to use the AT&T brand to popularize the concept but we certainly weren’t the inventors.

I became aware of EarthLink and Skye Dayton somewhat later and was appalled to find (to your credit) that EarthLink, thanks to its wise hookup with UUNET, had more PoPs and a lower cost of connection than we did.

Thanks for posting the comment.

David Beckemeyer

Nice post, Tom.

I wanted to clear up one bit of minutia regarding the history and timing of EarthLink/UUNET.

Your statement "on the day they went into business, [EarthLink] had both more PoPs available and a lower cost per minute than AT&T WorldNet" suggests that AT&T Worldnet was already in business when EarthLink started.

I was at Earthlink when we went nationwide on UUNET and that was in the fall of 2005 (the company began operations in 2004). We were the first to introduce the $19.95 all-you-ca-eat nationwide pricing model at a time when the going rate was $1-$2 per hour. As far as I know, AT&T Worldnet did not exist yet at that time, or at least the service was not yet announced anywhere. I think AT&T introduced the Worldnet service about a year later, also offering the same $19.95 per month pricing as EarthLink.


very good read, you can see corresponding processes in many ILECs across Europe. Telcos being vertically integrated 'silos' offering different services will not be able to be competitive in a (digitally) converged telecoms market. horizontal integration is the way forward


I never knew the world of telcom was so fascinating and instructive til I started reading your posts. Thank you from a naive grad student!

Post a comment

If you have a TypeKey or TypePad account, please Sign In.

Now on Kindle!

hackoff.com: An historic murder mystery set in the Internet bubble and rubble

CEO Tom Evslin's insider account of the Internet bubble and its aftermath. "This novel is a surveillance video of the seeds of the current economic collapse."

The Interpreter's Tale

Hacker Dom Montain is in Barcelona in Evslin's Kindle-edition long short story. Why? and why are the pickpockets stealing mobile phones?

Need A Kindle?

Kindle: Amazon's Wireless Reading Device

Not quite as good as a real book IMHO but a lot lighter than a trip worth of books. Also better than a cell phone for mobile web access - and that's free!

Recent Reads - Click title to order from Amazon


  • adlinks
  • adsense