`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`THE NETWORK PROVIDERS BUSINESS CASE
`FOR
`INTERNET CONTENT DELIVERY
`
`Peter Christy
`Vice President
`Internet Research Group
`399 Main St., Los Altos, CA 94022
`(650) 559-2103
`pchristy@irgintl.com
`www.irgintl.com
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page i
`
`About the Internet Research Group
`The Internet Research Group is the leading consulting, market research, and information
`services firm focused exclusively on Internet infrastructure technologies for e-business and e-
`commerce. IRG’s mission is to help major Internet technology vendors and key startups make
`faster and better decisions about their product, market entry, and market development strategies.
`IRG provides technical product and market analyses, develops authoritative reports on specific
`markets (such as Internet traffic management and Web caching products), produces focused
`informational Web sites around specific technologies, and creates white papers and other
`information products. IRG’s consulting division helps companies develop successful market entry
`and development strategies for new products in early adopter markets. For more information, visit
`IRG’s Web site at www.irgintl.com.
`
`Caching and Internet content distribution are two key focus areas for which IRG offers Web
`resource centers. For more information about caching, please visit www.caching.com. For more
`information about Internet content distribution, please visit www.cddcenter.com.
`
`For more information about this paper, please contact:
`Peter Christy
`Vice President
`Internet Research Group
`(650) 559-2103
`pchristy@irgintl.com
`
`
`
`
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page ii
`
`Contents
`EXECUTIVE SUMMARY................................................................................................ 1
`Internet Performance Issues ..................................................................................... 1
`Content Delivery Systems........................................................................................ 1
`The Akamai Network System .................................................................................. 1
`Benefits of the Akamai Network.............................................................................. 2
`OVERVIEW................................................................................................................... 3
`INTERNET PERFORMANCE ISSUES .............................................................................. 3
`How Performance Affects Internet Players.............................................................. 4
`INTERNET CONTENT DELIVERY: CHANGING THE TRANSMISSION MODEL.............. 4
`How an Internet Content Delivery service improves performance.......................... 5
`THE AKAMAI FREEFLOW SYSTEM............................................................................. 7
`Servers...................................................................................................................... 7
`Cache compatibility ................................................................................................. 8
`Request routing ........................................................................................................ 9
`Data Collection................................................................................................................ 9
`Data Processing ............................................................................................................... 9
`Domain Name Services............................................................................................ 9
`DISTRIBUTED SERVERS MAGNIFY BANDWIDTH .......................................................... 9
`HOW SUBSCRIBER NETWORK PROVIDERS BENEFIT FROM CONTENT DELIVERY.. 11
`Infrastructure savings through reduced bandwidth demand .................................. 11
`Improved customer satisfaction through better QoS.............................................. 11
`Lower subscriber churn rates ................................................................................. 13
`The Economics of “Churn” ........................................................................................... 13
`USING CONTENT DELIVERY SYSTEMS WITH CACHING........................................... 15
`Can a conventional cache be an Akamai server?................................................... 17
`CONCLUSION ............................................................................................................. 18
`Predicting the Future.............................................................................................. 18
`APPENDIX 1: GLOSSARY OF TERMS ......................................................................... 20
`APPENDIX 2: MEASURING ACCESS PERFORMANCE................................................. 22
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 1
`
`Executive Summary
`The rise of the Internet as a primary communications channel in our society has revealed
`significant performance problems that affect Network Provider revenues and customer
`satisfaction. Content delivery systems are a solution to these performance problems. Using
`Akamai Technologies’ network as an example, this paper explains how content delivery systems
`reduce Network Provider and subscriber satisfaction problems. This paper will address the
`business case for all Network Providers, including ISP operations centers, universities, corporate
`campuses, broadband providers, and all other locations with large numbers of Web visitors.
`
`Internet Performance Issues
`A Web user’s experience is directly related to the speed at which his browser retrieves and
`displays pages from Web servers. This performance can be affected by Web server processing
`delays (due to excess demand on the server providing the content); Internet delays (overloads at
`network peering points or routers between the user and the original content server); or “last mile”
`delays (low-speed, dial-up connections between the user and the Network Provider).
`
`Content Delivery Systems
`Content delivery systems resolve performance problems related to Web server processing
`delays and Internet delays. Fundamentally, content delivery systems create and maintain up-to-
`date copies of popular or high-bandwidth Web content in cache servers at multiple locations at
`the edges of the Internet. As a result, users requesting popular Web content may well have those
`requests served from a location much closer to them (a local Network Provider’s data center),
`rather than from much farther away at the original Web server. By serving content requests from
`much closer to the user, content delivery systems reduce the potential for Web server overloads
`and Internet delays.
`
`The Akamai Network System
`The Akamai Network System is a content delivery service that consists of thousands of
`servers located at Network Provider data centers around the world. Akamai contracts with high-
`volume Web content providers such as Yahoo to distribute their frequently accessed content.
`Akamai uses the Internet to distribute copies of this content to its FreeFlow servers, and then,
`using proprietary logic from its network operations center, Akamai uses the Domain Name
`Services (DNS) system to reroute standard browser contents for its customers’ content to the
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 2
`
`optimal FreeFlow server at the location nearest the user. These servers (nominally configured
`with 1 GB of DRAM and two 18 GB disk drives) are provided at no cost to Network Providers
`willing to host them. They integrate seamlessly with a provider’s network, storing popular, high-
`bandwidth content and serving it to the Network Provider’s subscriber base. The servers are
`typically configured in groups of five with two Ethernet switches, although smaller and larger
`configurations are also available.
`
`Benefits of the Akamai Network
`The primary benefits of the Akamai Network for Network Providers are lower infrastructure
`costs, higher subscriber satisfaction and quality of service, and reduced subscriber churn. By
`serving popular content from local servers at the Network Provider’s data center, the FreeFlow
`system saves Network Providers the cost of retrieving this content from origin servers that may
`be thousands of miles away across expensive communications links. And, by serving subscribers
`from a local server, Network Providers are able to offer a significantly higher level of service,
`presenting popular Web content quickly and reliably, and thereby also reducing churn rate and its
`associated costs.
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 3
`
`Overview
`The Web and the Internet have revealed the power of universal information access, and the world
`hasn’t been the same since. What started 30 years ago as an academic and research tool has
`become the foundation for modern industrial and commercial activity. Successful Web sites
`literally experience a doubling of load every 90-100 days. Internet access has driven PCs into
`more than 50 percent of U.S. homes faster than anyone expected. Billions of dollars are being
`invested in additional data transmission capacity, which for now seems at best likely to just keep
`up with demand. And as the Internet evolves, Web access performance – the absence of delays or
`transmission problems – becomes more important, often relating directly to a Network Provider’s
`brand value and subscriber satisfaction.
`
`Using Akamai’s FreeFlow Internet content delivery system as an example, this paper
`explains emerging Internet content delivery services and how they can make a big difference in
`achievable Web performance, thereby reducing Network Provider bandwidth costs and boosting
`subscriber satisfaction for providers that participate in content delivery systems.
`
`
`
`Internet Performance Issues
`The Internet miracle is based on a growing set of standardized, interconnected networks, and
`a standard for information publishing and viewing (the Web and browsers). We access
`information on the Web by giving our browser the name of a Web site, after which that content is
`fetched and displayed using the Internet to access and transport the data. If everything goes well,
`and all the links are high-speed, fetching Web data seems instantaneous. But as we all know, the
`Web is rarely instantaneous. Internet performance problems often arise in any of three general
`areas:
`
`• Web server processing delays: Traffic loads that are ten or more times greater than
`the typical load on the site as news events or fads drive high access. The site’s
`servers are unable to handle high loads and thrash, halting delivery of data.
`
`•
`
`Internet delays: Delays somewhere in the Internet between the information
`requester and the Web server (often at the peering points between networks).
`
`• Last mile delays: Delays in the connection between the subscriber and the Internet
`(e.g. due to a 14.4 Kbps dial-up modem connection).
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 4
`
`Performance problems are important to content providers and consumers, but they are more
`important to Network Providers, who rely on performance as a key differentiator of their service.
`Performance problems drive subscribers to sample competitive services. Typical Network
`Provider churn rates are over 4 percent per month, which compounds to an astounding 50 percent
`per year loss rate, and with new subscriber acquisition costs equal to 2-10 months of revenue, a
`lower churn rate can directly and significantly impact business profitability.
`
`How Performance Affects Internet Players
`In an ideal world, clicking on a page link would be instantaneous – the new page would be in
`place faster than the mind could absorb its content. At best, today’s world is a far cry from that
`ideal. Human response time is less than a second, but the typical page response time for leading
`sites is more like 10 seconds, according to data measured by Keynote Systems. And 10-second
`page responses are hardly the worst case: a site can be down or inaccessible, or there can be
`instability somewhere in the Net between user and server (packet loss at a peering point or “route
`flap”). A flash crowd could also hit the site, driving the computer systems into severe, thrashing
`overload.
`
`Performance problems affect everyone in the Internet value chain:
`
`• Network providers have lower customer satisfaction, higher subscriber turnover, and
`higher connectivity costs as they attempt to improve content transmissions or
`subscriber access by increasing Internet capacity and peering with additional
`networks.
`
`• Web users seeking information must wait for it to be delivered, which wastes time
`and decreases their satisfaction.
`
`• Content providers seeking to deliver their information see their content’s impact
`decreased through access delays or slow delivery. As a result, site designs are
`“dumbed down,” and lack rich content.
`
`
`Internet Content Delivery: Changing the Transmission Model
`The traditional Internet model of browsers connecting to content servers over the Internet is
`the root of many of these problems. Using a distant browser to request files from an origin
`content server was the elegant idea that created the Web and accelerated the Internet revolution
`(see Figure 1). But as the Internet increases in commercial value, as content becomes richer, and
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 5
`
`where performance improvements translate directly into business value, we start to see the
`limitations in this simple model:
`
`• Servers can’t keep up with the peak loads presented, unless the site is built with
`gross overcapacity.
`
`• Beyond the borders of the U.S., network capacity diminishes rapidly. What works
`well here is intolerably slow for overseas users.
`
`• The data exchange points (peering points) between the various networks that
`constitute the “Internet” become overloaded, so data is lost and must be
`retransmitted.
`
`• The bulk of a Network Provider’s traffic is “backhaul” data to retrieve rich content
`from distant servers.
`
`As content becomes richer and more complex, the chances increase that these problems will
`impinge on performance, bandwidth costs, and the satisfaction of valuable subscribers.
`
`Figure 1. Traditional Internet content delivery.
`
`
`
`How an Internet Content Delivery service improves performance
`A content delivery service alleviates these problems by moving demanding content (popular
`and/or bandwidth intensive) closer to the people requesting it. Rather than serving content from
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 6
`
`the origin Web site, the content distribution model makes copies of key content on a multiple
`content delivery servers sites distributed through the Internet, close to the users requesting that
`content (see Figure 2). This approach addresses all of the problems previously described:
`
`• With the hottest content off-loaded to multiple distributed content servers, the load
`on the origin server diminishes.
`
`• The connection from a local content delivery server to the user is shorter than the
`connection to the origin server, is less subject to delays, and has higher bandwidth,
`producing increased subscriber satisfaction.
`
`• The path from the user to the content server transits fewer peering points (often none
`if the content server is within the same network as the user), greatly reducing the
`sources of packet loss and data retransmission.
`
`• As a system shared by many content providers, a content delivery service
`dramatically increases the “hit ratio” associated with caching, thereby reducing
`Network Provider bandwidth costs.
`
`2. Content Provider’s
`Web Server returns
`HTML with embedded
`URLs pointing to Akamai
`Network
`
`2
`
`1. User enters
`standard URL
`1
`
`3
`
`4
`
`3. User’s browser
`requests embedded
`objects
`
`4. Rich content
`is served locally
`by Akamai
`
`= Akamai Server
`
`
`
`Figure 2. Content delivery with the Akamai Network system.
`
`As the success of systems like Akamai’s clearly demonstrates, a shared content delivery
`service is a much more effective solution to Web access performance issues than is even possible
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 7
`
`by building bigger servers and adding additional interconnection capacity. By solving a common
`problem, content delivery services offer common benefits to everyone in the Internet value chain:
`
`• Subscriber Network Providers save on infrastructure costs and improve their
`customer satisfaction.
`
`• Content consumers get better quality as their access to popular Web site improves in
`performance and availability.
`
`• Network providers improve the performance and reliability of their service.
`
`ICD systems can directly impact performance because they are practical remedies for these
`problems, and because they improve performance in the absence of any severe problem. In other
`words, adding content delivery directly and significantly improves customer satisfaction,
`decreases subscriber churn, and contributes to the Network Provider’s bottom line.
`
`
`The Akamai FreeFlow System
`The Akamai FreeFlow system has three functional components: Akamai servers, a request
`routing system, and Domain Name Services.
`
`Servers
`Akamai’s FreeFlow system runs on thousands of servers distributed throughout the Internet
`at Network Provider operations centers, universities, corporate campuses, and other locations
`with a large number of Web visitors. Each server in this generation of the service has a minimum
`of 1GB of RAM and two 18 GB disk drives and performs the functions of a cache, storing a copy
`of the content that Akamai delivers.
`
`Akamai typically deploys servers in configurations of five with two Ethernet switches (see
`Figure 3), although both smaller and larger configurations of three, nine or eighteen servers are
`also available. The Ethernet switch is used to provide inter-server communications as well as a
`100 Mbps connection to the Internet. The rack-mounted servers are extremely easy to install and
`are typically functioning within hours with absolutely no changes to a provider’s network
`topology or configuration.
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 8
`
`Ethernet Switch
`
`To Internet
`
`Ethernet Switch
`
`Support Server
`
`Akamai Server 1
`
`Akamai Server 2
`
`Akamai Server 3
`
`Akamai Server 4
`
`Figure 3. A typical deployment diagram for Akamai servers.
`
`
`
`Through its Akamai Accelerated Network Program,
`Akamai offers these servers to qualified Network
`Providers at no cost – the provider’s actual investment is
`rack space, power, and local network connection. In
`exchange for these modest costs, Network Providers
`gain the higher bandwidth and improved quality of
`service for their customers while diminishing their need
`for upstream Internet capacity. What's more, the Akamai
`servers are configured to serve only a Network Provider’s downstream customers, preventing the
`possible consumption of upstream bandwidth.
`
`“It’s free and we save
`on traffic to the U.S. As
`more content providers
`sign up with Akamai it will
`get even better.”
`--Kenji Hirota
`KDD Japan
`Akamai Accelerated Network
`
`Cache compatibility
`In the near future, Network Providers that have invested in third-party caches will be able to
`join the Akamai network through a simple software upgrade to their cache software. Akamai and
`Cisco have authored the Open Cache Interface, which will enable third-party caches to
`communicate with the Akamai network and store content for major Akamai customers such as
`Apple, Yahoo, and CNN Interactive. To date, CacheFlow, Cisco, InfoLibria, Network Appliance,
`and Novell have joined this initiative.
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 9
`
`Request routing
`Akamai optimizes access to the content it serves by continuously optimizing how requests
`for that content are distributed to the thousands of servers that serve content. The first step in that
`optimization is data collection.
`
`Data Collection
`To maximize the performance from its set of servers, Akamai first uses methods of
`measuring Internet performance, Akamai content request traffic, and server loads.
`
`Data Processing
`With all this data collected, FreeFlow uses sophisticated algorithms to compute the optimal
`mapping of requests to servers, given request and server loads, the available access to the various
`server centers, and the general condition of various Internet links. Those routings are put in place
`through the use of the Internet Domain Name Services.
`
`Domain Name Services
`Akamai uses the Domain Name Services (DNS) functions within the Internet to steer
`requests for Akamai-served content to specific Akamai servers. All Akamai content is part of the
`Akamai name domain. As a DNS server contacts Akamai to resolve an Akamai URL to the IP
`number equivalent, Akamai considers the location (IP number) of the requesting DNS server and
`feeds it the optimized routing for that location. Then, the DNS system does the actual work of
`steering the request to the right Akamai server. As load and Internet conditions change, Akamai
`re-optimizes the solution in an ongoing fashion in near real time.
`
`
`Distributed servers magnify bandwidth
`It is important to note that Akamai’s Network is not a facilities-based network. Akamai
`content is served from individual servers located as close to the consumer as possible, rather than
`over a dedicated or shared network. Each Akamai server is capable of serving over
`100Mbits/second, so the set of 1,700 servers can serve 170 Gbits/second – a huge capacity
`compared to the requirements of even the largest Web sites – but there is no high-capacity data
`transmission network needed between those servers or back to the NOC. The only network
`capacity used is that right at the edges, where bandwidth is the least expensive.
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 10
`
`ISP
`
`Users
`
`Origin Web
`Server
`
`Users
`
`ISP
`
`ISP
`
`Users
`
`ISP
`
`ISP
`
`Users
`
`Users
`
`Figure 4. Traditional Internet content access – browsers get content from origin server.
`
`Users
`
`Users
`
`ISP with
`Akamai
`Server
`
`Origin Web
`Server
`
`Users
`
`ISP with
`Akamai Server
`
`ISP with
`Akamai
`Server
`
`ISP with
`Akamai Server
`
`ISP with
`Akamai
`Server
`
`Users
`
`Users
`
`
`
`
`
`Figure 5. Internet content distribution with Akamai servers – Akamai servers at Network
`Providers cache hot content and serve it locally.
`
`Akamai contracts with content providers to deliver their content through the Akamai
`FreeFlow service as if it were a physical network. Content is “Akamaized” by changing the URL
`references to the Akamaized version. Requests are subsequently served from Akamai servers
`situated on the edge of the network, near the user. This simple change improves performance
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 11
`
`under normal conditions and avoids the problems outlined earlier. Akamai logs the rate at which
`content is served for each customer, aggregating the instantaneous contribution of thousands of
`servers, and charges the customer as if that content had been served from a centralized server
`through a single big pipe.
`
`
`
`How Subscriber Network Providers Benefit from Content Delivery
`The principal benefits of content delivery services from the Network Provider’s point of
`view are savings on infrastructure costs, increased customer satisfaction, and reduced subscriber
`churn (and the consequent reduction in customer acquisition costs).
`
`Infrastructure savings through reduced bandwidth demand
`By adding content delivery servers to its infrastructure, a service provider saves significantly
`on capital and operating expenses because these servers serve its users exclusively: no requests
`from outside the provider’s network are mapped to the provider’s region of servers. The content
`server delivers content locally that otherwise
`“Akamai has very clever
`would have to be fetched from the Internet. The
`technology that is saving us
`money and improving our service.
`bandwidth needed to connect the server locally
`It has actually improved our
`is essentially free (abundant) whereas the same
`profile in the marketplace.”
`bandwidth over the Internet is costly. For most
`--John Lindsay
`sites, the average bandwidth savings with an
`Operations Mgr., iiNet Ltd.
`Akamai Accelerated Network
`offering like Akamai’s will be 10-15 percent.
`
`The actual saving is probably considerably
`larger than the average, because most Network Providers provision for peak load, not for average
`load. Because the content provided by a content delivery service tends to be exactly the content
`with the highest probability of flash crowds and surge loads, the content delivery service has a
`greater impact under these peak conditions. Whereas the impact on average load might be 10
`percent, the impact on peak loads could be much greater.
`
`Improved customer satisfaction through better QoS
`One of the biggest impacts of Internet content delivery on a Network Provider’s business is
`in the area of customer satisfaction. Viewed as a cost savings, a content delivery system saves
`some percentage of costs; but from the user’s point of view, content delivery can make a black
`and white difference in performance.
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 12
`
`A subscriber Network Provider runs a difficult business. Good access is transparent – fast
`and reliable. Access performance becomes an
`issue when it is less than transparent. Even
`though many performance problems are not
`related to a provider’s network, new Web users
`attribute all of their problems to their Network
`Provider. There can be many causes for
`subscriber access problems:
`
`“It makes the Akamaized
`parts of the Web run like a
`scalded cat. In Australia, where
`people are accustomed to Web
`pages from the USA arriving in a
`sluggish or inconsistent manner,
`the consistent, blazing speed at
`which Akamai served content
`arrives is jaw dropping stuff.”
`--Simon Hackett
`Technical Director
`Internode, Australia
`Akamai Accelerated Network
`
`• Problems in the subscriber access
`line (dial-up connection, xDSL
`line, cable)
`
`•
`
`Inadequate provisioning of
`upstream bandwidth – not enough capacity for the number of on-line subscribers
`either under normal loads or flash crowd conditions
`
`• Problems with the origin server (inadequate capacity, for example)
`
`• Problems in the Internet between the Network Provider and the origin server (packet
`loss and data retransmission from an overloaded peering point, for example).
`
`It’s important to realize that using a content delivery system can make a big difference in all
`aspects of performance except those due specifically to problems within the Network Provider’s
`network. A content delivery system moves highly popular, highly accessed content from the
`origin servers to a content delivery system within the provider’s network, so it alleviates the other
`problems as follows:
`
`• Problems in the access line – Because content delivery servers are placed near
`modem racks, DSLAMs, and cable head-ends, local loops are fully utilized.
`
`•
`
`Inadequate provisioning of upstream bandwidth – The local content delivery server
`reduces upstream bandwidth demand in all conditions and especially in flash crowd
`circumstances.
`
`• Problems with the origin server – The local content delivery server reduces traffic on
`the origin server by as much as 90 percent (for popular, served content) and largely
`eliminates dependence on the server. Again, these improvements are greatest
`precisely under the most demanding flash crowd circumstances.
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 13
`
`• Problems in the Internet between the Network Provider and the origin server – The
`content retrieved from the
`local server isn’t subject to
`Internet delays.
`
`The subscriber satisfaction impact of
`a content delivery system is amplified
`because the content delivery system is
`likely to accelerate access to the most
`popular, most frequently accessed material.
`Delays to relatively obscure content can
`easily be attributed to problems with the
`server responsible for its delivery, but
`access to popular sites such as Yahoo! are
`a natural way to compare the performance
`of different Network Providers.
`
`Akamai has a really good shot at
`solving the global Web content
`distribution problem, and I think they
`have it nailed. Because of our
`partnership with Akamai, I have Web
`sites I can highlight to my customers,
`like CNN and Apple, whose enormous
`download speed becomes a great
`demonstration of the value offered to
`them by choosing Internode. It is a
`strong competitive differentiator in an
`increasingly challenging market.
`--Simon Hackett
`Technical Director
`Internode, Australia
`Akamai Accelerated Network
`
`Lower subscriber churn rates
`High churn rates are a general Network Provider problem. Tables 3 and 4 below give some
`concrete examples of the impact of churn rate on profitability and shows how seemingly small
`changes in churn rate end up making a big difference in bottom line profitability. Happier
`customers mean lower churn rates, lower customer acquisition costs (because less is spent to
`replace departing subscribers), and lower associated administrative costs involved in deleting and
`adding subscribers.
`
`The Economics of “Churn”
`The analysis on the following page shows how reducing the subscriber churn rate (the rate at
`which subscribers discontinue service) can have a profound the impact on a Network Provider’s
`gross revenue.
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 14
`
`
`
`Basic Assumptions
` Churn rate
`Average subscriber acquisition cost
`(in revenue months)
`Non-critical Assumptions
` Average subscriber monthly bill
` Number of subscribers
`Calculations
`Annual revenues
`Annualized loss rate
`Monthly subscriber loss
`Acquisition cost / subscriber
`Cost to refresh lost subscribers / mo.
`Impact of Reduced Churn Rate
` Savings / month
` Savings / year
`Bottom Line Impact
` Savings as a percent of revenue
`
`
`
`
`
`
`4.4% (Average)
`
`5
`
`$20
`100,000
`
`$24,000,000
`42%
`4,400
`$100
`$440,000
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`4.0%
`
`5
`
`$20
`100,000
`
`$24,000,000
`39%
`4,000
`$100
`$400,000
`
`$40,000
`$480,000
`
`2%
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`3.5%
`
`5
`
`$20
`100,000
`
`$24,000,000
`35%
`3,500
`$100
`$350,000
`
`$90,000
`$1,080,000
`
`5%
`
`Table 3. Bottom line impact of specific churn rate reductions
`Source: Internet Research Group
`
`In Table 3, we assume that a Network Provider’s goal is to maintain or grow the customer
`base. We assumed an average churn rate of 4.4 percent (the figure given as an average Network
`Provider churn rate a recent article). We also assume that the average monthly revenue per
`subscriber is $20. We then calculate the cost to replace lost subscribers at the average rate, and at
`two reduced churn rates of 4.0 percent and 3.5 percent. The explanation details the calculations in
`Column B (the 4.4 percent churn rate).
`
`In the Basic Assumptions section, we assume the average churn rate, and a subscriber
`acquisition cost of 5 months (at the $20 monthly average subscriber revenue shown directly below).
`
`Under Non-critical Assumptions, we assume 100,000 subscribers and compute annual
`revenues of $24 million ($20/mo. x 100,000 subscribers).
`
`In the Calculations section, we determine the annualized version of the churn rate just to
`emphasize the severity of the problem: If you lose 4.4 percent of your subscribers a month, that’s
`a 42 percent loss rate on an annualized basis! We multiply the monthly loss rate times the annual
`
`© Internet Research Group, 1999
`
`

`

`The ISP Business Case for Internet Content Delivery
`
`Page 15
`
`revenue to get the monthly subscriber loss, which is also the number of new sub