Internet Strategies for Standard & Poors
March 1995 - Rex Ballard
to Phil Sanderson
From Phil Sanderson to Walt Arvin
From Walt Arvin to Joe Deon
I. Executive Summary
In the past 2 years, the the Internet has become a major focus for the
the community. Although the Internet has been around for over 15 years,
it was primarily a research network used by engineers and scientists at
leading universities, government agencies, and military branches.
Since 1990, when the Bush administration mandated that "The Internet"
be a fully self-supporting infrastructure, it has been a rapidly growing
network which has grown exponentially from less than 1 million users in
1990 to nearly 64 million users by 1995. Secondary markets such as interactive
access known as "Shell Accounts" and access through the World Wide Web
(approximately 1 million independently operated servers) have experienced
similar exponential growth curves. The most recent wave of growth is in
the "Commercial Internet User". These are people who pay for services beyond
basic connectivity. There are currently about 2 million "Commerce Users".
This paper will chart out a strategy for competing effectively in this
rapidly growing market in such a way as to be able to take advantage of
this and subsequent waves of markets.
II. Why is the Internet important to S&P
More than anything else, the Internet represents Communication, Powerful
Relationships, and Opportunity. This is true for users, information providers,
access providers, and carriers. The internet is a carefully balanced set
of agreements or de-facto "standards" established and maintained by a series
of public specifications known as Requests For Comment (RFCs), and Freely
Distributable Reference Source Code (FDRSC). Although most of this code
was written for variants of the UNIX operating system (once FDRSC itself
and avialable in FDRSC variants), several commercial vendors have ported
this code to Windows and OS/2. Currently this "unsupported" software is
supported by over 200,000 engineers and MIS professionals. It is very popular
with Value Added Resellers, Consulting Services, and Large corporations
because they don't have to "build from scratch", yet they can diagnose
and correct problems more quickly than "vendors". Because these corrections
are contributed back to the internet, support is generally quite good.
DEC is now calling it "complimentary" software.
Currently, there are an estimated 60-80 million E-Mail users and an
estimated 10 million users of the World Wide Web (the Web). There are over
4.5 million registered "Web Pages", each averaging about 25,000 hits (uses)
per week. Quality Commercial content can go as high as 25,000 hits per
day. The average web server serves 10-20 pages/second.
The Web provides an easy to use interface that requires a minimal amount
of expertize to use. Most commercial Web Browsers are very easy to set
up. Dial-up users need to provide a phone number and 2 "addresses". Each
address is 4 numbers between 1 and 256, for example 192.131.34.7 this information
is provided by the access provider.
As an advertizing media, the Web provides content providers the opportunity
to provide advertizing to highly qualified customers.
A. What if we ignore the Internet
The internet is growing at a rate of nearly 20 percent/month. There are
a number of metrics being used to estimate the approximate number of users
using the different types of access, but nearly all indicators are that
people are connecting at an an exponential growth rate that exceeds 10%/month.
In some segments (Web users) the growth rate exceeds 20%/month. There is
already a significant presence on the internet. The economics in terms
of band-width efficiency and low switching and link costs (largely due
to FDRSC) has caused the internet to displace traditional on-line "self
contained services". Lexus and Nexus are now on the Internet, as is Dow
Jones. There are several government databases (Edgar) which are provided
in a "raw" format in delayed databases. The on-line services are also now
providing access to the web and other internet services and finding themselves
able to leverage outside data.
Since MCI first started carrying the internet traffic in 1991, the combination
of large amounts of research data, interactive communication (e-mail, bulletin
boards called "news groups", and real-time Internet Relay Chat (computer
conference calling) have lured users into first supplementing their on-line
service (America OnLine, Prodigy, or Compuserve) and eventually subscribing
to a more direct full service direct access provider. On of the more significant
factors is the integration of corporate enterprise networks and the Internet.
Historically, the internet was primarily corporate lines which were "donated"
in exchange for access to research and development information. Today,
fire-walls and routers provide a means of providing secure access using
protected common carriers. The invisible "commercial internet" is also
growing at exponential rates.
The internet is to telecommunications what the automobile and highway
system was to transportation. While we might wish for the relative security
of the Railroads, or attempt get security through Mass Transit (Busses
and Taxicabs), the consumer is willing to take a calculated risk in exchange
for the ability to drive from his home to any destination he chooses and
return, on his schedule. The internet provides the same level of risk and
convenience. As a result, the internet consumer can choose from an extremely
competitive range of products which are more convenient than point-to-point
dial-up services (Compuserve, AMERICA ONLINE, or Prodigy) or dedicated
link services (Rueters, Quotron, A.D.P.).
Trying to ignore internet is like trying to ignore the train that is
coming down the tracks. It is better to be standing on the platform than
between those tracks. Companies who have resisted the Internet have lost
a significant market share.
B. How it fits with an overall online access strategy
The Internet is rapidly becoming commercialized and is now accessible by
nearly anyone in the world. There are several different protocols and services
which facilitate access. Bulletins and clip files can be sent via e-mail,
real-time feeds can be sent using IRC. Finally, we can provide high-quality
services with uniform interfaces whether provided at a national site, or
in the local corporations. For example, customers with high demand could
receive a Comstock feed and have the same Web interface on their local
system.
Providing Internet access would provide access to all Prodigy, Compuserve,
America Online, Delphi, and Direct Access internet customers. Proprietary
strategies could give us only a small segment of the market at best.
By adopting the open standards of the Internet, we can also provide
high-quality low-cost "enterprise servers" which can combine Comstock,
internet (for retransmission of lost sattellite records) and Local Area
Network access for high-performance "Web Servers" within local corporate
fire-walls.
C. What are our real expectations
There are several sources of revenue available. Web pages are now earning
as much as $7000/line when they are relatively busy. The important factor
on the internet, especially WEB advertizing is that the reference can be
placed such that very highly qualified customers are actually intentionally
requesting the advertizing with the intent of making a purchase decision
within moments of starting to browse.
The information consumer market (people willing to pay for content as
well as access) is also growing exponentially. It is not unusual to have
an average of 200-600 simultaneous users on a popular site. For several
types of information (news, financial data, classified searchable advertizing...)
they are willing to pay s much as $1.50/page (Meade), or as much as $100/month
(Dow Jones). Generally there is a wider acceptance of "bulk purchasing"
- 500 pages/month for $5.00/month (the equivalent of a copy of Byte magazine).
Again, this figure was for a local Newspaper offering it's news and some
secondary databases.
Even at a mere penny per page, it is likely that we would be capable
of earning $60,000/day at only 60 pages/second. A 5 segment stock report
might actually be 40-50 different pages.
III. Significant Issues
If we wish to be on the internet, there are some issues we need to manage.
In order to reach the widest range of customers, we need to manage compatibility.
In order to protect the value of our content, we need to manage security.
We also need to provide convenient mechanisms to receive payments for our
content. We also need to provide ways to get the customer hooked on our
content and have them return regularly by providing "free samples" while
not giving away everything.
A. Compatibility.
We must first be aware of the Open Standards. This is the key to wide-spread
user acceptance. Unlike vendor controlled environments where "revolutionary"
releases are shipped once or twice per year, the Internet is a more "evolutionary"
with innovations being available as much as a few years prior to general
acceptance by the entire user community. It is important to be aware of
these innovations, even to provide demonstrations, but it is risky to implement
features that exclude clients. The customer who is greeted with a message
like "The server refuses to accept your connection" will become a regular
customer of the competition.
B. Security and Copyright
There are two sides to the security issue. One is protecting our data from
unauthorized access and modification, the other is overcoming the misinformation
that has been proliferated lately. The WEB uses a communications protocol
called the HyperText Transfer Protocol (HTTP) which provides much of the
session layer management. The Web also uses an application protocol or
"language" called the HyperText Markup Language (HTML). The latest browsers,
including Netscape and Mosaic 2.0+ provide security features such as the
"Secure Sockets Protocol" (SSP) and the "Secure Hypertext Transfer Protocol"
(SHTTP). The SSP is patented and has must be licensed for commercial use.
The SHTTP is available under FDRSC which assures a wider acceptance and
reasonable licensing costs.
The secondary issue is that of copyright. The individual user who posts
a single story once/month probably does more to advertize our product and
service (they need to be guided through an explanation of attribution).
The big concern is the user who routinely clips documents and starts passing
them around in news-groups. Generally, those people should site the URL
instead of quoting the full article. A 2 paragraph summary, or the first
2 paragraphs should be reasonable.
C. Commercial -- how do we get paid
Until recently, commerce on the net was a bit challenging. The risk of
having your credit card and related information intercepted by someone
who would use your card to buy their computer was too high. Recently, relaxation
of restrictions on RSA and DES encryption have reduced the risk of casual
interception (something like looking over your shoulder electronically).
Several companies including First Virtual Holdings, Digicash, NetCash,
and First Boston offer "push-button checks". The set-up fees range from
$10 (First Virtual), to $2500 (Visa), and are likely to go up rapidly as
leaders emerge. They collect commissions ranging from 39 cents to 6 percent.
There are several charging strategies. The simplest, from a technology
standpoint would be the "charge per hit" or the "bulk rate charge per hit".
The concern here is that each user might have to contract with 2000 providers
and each provider has to deal with 2 million customers. There are several
companies which provide a "One fee buys thousands of servers".
These systems let you delegate authorization to them and at the end
of the month, you tally up the number of accesses you submitted and collect
a percentage based on that proportion (1 to 3 cents/page). Users who are
information hungry will be very likely to use this service. Companies that
provide this include the NewsShare corporation, the Author's registry and
the Copyright Clearing house. In addition, many ISPs are including royalty
charges to one or more of these services. Examples of this include NewsShare,
Copyright Clearing House, Folio, and the Author's registry.
D. Content strategy -- how much free
There are several strategies for managing free vs commercial. A simple
strategy is to combine the "puppy dog close" (try it out a few weeks, if
you don't like it, you pay nothing), and the health club plan (get them
on for a year plan so that they feel the must come back). In addition,
we may want to work out bulk rates for access providers (1,000,000 pages/month
at 2 cents/page). The unused or excess pages/month will be factored into
the next month's rate.
We may wish to become a payment recipient who supports several copyright
clearing houses. In effect, we would collect the payment and the clearinghouse
would credit our "pool" and allow the users we put into the pool access
to ours and other publishers information.
IV. Actual Recommendation
What follows are several recommendations, with alternates where desired.
A. The Home Page
In addition to the usual copyright and "more about us", we should also
provide lines which point to search engines (Wais, Folio, Lycos...) Another
line should link to a Marketscope "Menu".
We should also include Payment options, first for payment directly to
us for private and consortium data. We should offer them "apply buttons"
for each of these cards. We should at least support First Virtual, Digicash,
and NetCash. The goal is to make it as easy as possible to turn them into
a paying customer.
All content pages should contain links back to the S&P home page.
Specifically a hierarchy like:
McGraw-Hill, Standard & Poor's, Product (Marketscope...), Topical
(Company profile information), and "previous". The intent here is to guide
customers who receive our pages as a result of search hits back to company
oriented product.
B. Product offering
- Content
- Presentation
- Pricing Model
The "low-budget" customers would be given access to a delayed (batch update)
feed on a very generic server. The "High end" customers would receive real-time
updates and rapid response times. Corporate customers should have the option
of a web-server interface to Comstock/Marketscope data.
We can also point links to "premium" services and invite them to upgrade
their levels after a certain number of "Samples". After they've seen 10
consensus reports within a week, they know what it is and can determine
whether they consider it worth the extra payment.
In either case, advertizing is a definite possibility. The "random spot"
is generally disliked by internet users, on the other hand, relevant information,
such as a link to company home-pages or targeted catalogues can be very
worth-while. Often, corporate users search content as buyers looking for
suppliers. By the time they press the "See advertizing" page, they are
highly qualified and very interested. Some of the most successful internet
commerce servers are little more than "Classified Advertizing".
We will probably want to create "Company Profile pages" which function
as a "table of contents" for a group of products, including prospectus
and advertizing "buttons".
- In house vs. outside development (system operator)
We want to leverage as much of the available technology as possible. It
is important to have source code, to know that that source code is available
to others, and to participate in the supporting news groups. This will
reduce the development effort and at the same time assure a higher probability
of rapid resolution of problems.
The internet is a global network. We need to be aware that just about
the time New York Closes, Hong Kong is opening. When England opens, it
3:00 AM. It can be very useful to be able to get a response from support
group members in England, New Zealand, Australia, and California. As good
internet citizens, we should monitor these groups and contribute where
answers are available "off the cuff". Often, the bug we fixed yesterday
will be discovered by 3 others tomorrow.
Although it is possible to obtain fully bundled turnkey systems, they
are rarely as turnkey as they would appear. In addition the license rates
can run as high as $1 million dollars for software that runs on hardware
that costs less than $50,000. Using the modular software approach including
the Unix Platform, and TCP/IP
C. Other uses
Most of the focus has been on "The Web". The internet also provides e-mail,
news-groups (public forums, many technical and business oriented), and
archives (ability to search nearly 1 billion (mostly text) documents in
a matter of seconds) using "front-ends" which interface to web browsers.
We want to be able to serve these needs as well.
The internet is a natural form of customer feedback, discussion groups,
and free-lance support. As the commerce market increases, there are more
business oriented forums as well.
We should use the "Standard" products. There are about 20 mail interfaces
to Windows, all but the "commercial" packages support internet standards
transparently.
There is also the possibility of providing encrypted feeds through the
internet, or using the internet to provide "retransmissions" for Comstock
Sattellite feeds. Both can provide access to the small "branch banks" and
"Certified Financial Consultant" who can benifit from the combination of
Feed and Internet products.
D. Development timetable
The actual development should be rather rapid. Availabilty of code and
support for Unix on platforms ranging from 386/SX PCs to 1000 Processor
engines can provide the ability to grow at the 10-20% rates. The modular
structure allows us to "tune" the network based on shifting demands.
Prototype servers, and related engines can be configured in a few weeks.
I was able to load and start a web server for Ultrix in about 2 days. A
development infrastructure using BSDI UNIX or Linux can be set up in about
2 weeks. The logistics of coordinating with network administrators and
managing the flow of traffic over the "fire-wall" may add some additional
delay.
We will also need to negotiate licenses with some companies. The "list
price" on a netscape server using a Sun/Solaris engine can run as high
as $300,000 per year. The hardware runs about $200,000. This will serve
about 4 million hits/day (3000 pages/minute).
The smaller, loosely coupled servers provide a very high degree of fault
tolerance as well. This technology has been used on Directory Assistance
Systems for about 25 years with down-times averaging 15 minutes/year.
A BSDI or Linux PC, configured with an NCSA server (Mosaic), runs about
$3000/year and serves about 200 pages/minute. Methods of delegating work
to other low-cost servers can increase this to over 800 pages/minute. This
is sufficient to service a pair of T1 links.
The BSDI and Linux hosts are also used as the building blocks for Routers,
Fire-walls, Database servers, and File servers using freely redistributable
source code provided on CD-Roms with quarterly or monthly updates.
We will also need to negotiate licenses for payment schemes and possibly
encryption.
Since much of our content is either fielded data or SGML tagged data,
it will be fairly easy to adapt it for use on the web. The Edgar content
can be converted to HTML using unix utility languages that can cut development
time by 80% compared to regular C development. Tom Gerhard has already
converted several European Marketscope Pages to HTML using a utility package
from SoftQuad.
In effect we can return a substantial NOI at 200 pages/second ($4 to
$10/second) on increments of $10,000/server. As resource intensive functions
are identified and specialty needs are singled out for optimization, they
can be migrated to performance specific platforms such as Pyramid, VMS,
or RS-6000.
The major development effort will be the development of Menus and artwork.
This can be done in house, or through free-lance designers at $20-200/page.
The interaction is similar to a game. We will need someone who can design
an interesting dialogue.
Consistent with the "Evolutionary" nature of the internet, we should
work toward getting the first "Easy" products (Marketscope, Edgar) onto
the web within 90 days. We can add the various components of stock reports
within 3 to 5 months. Many of the tools for rapid prototyping and development
are provided on the recommended platforms, the remainder can be loaded.
The other major effort will be negotiations. There are ways to structure
systems so that "simultanious users" can be reduced to a very small number
(under 200 oracle licenses for example) while serving several thousand
"browsers". We do want to make sure that we do not find ourselves irrevocably
committed to paying royalties based on each individual user coming in through
the "Pool" or through through subscription. This could easily be a number
in the millions.
E. McGraw-Hill coordination
This cooperation needs to occur at several levels. First, there are the
development logistics. There have been problems with the Heightstown routers,
fire-walls, and address/routing management. We need to have coordination
between the different systems such that we can add new connections to our
side within a few minutes, and can add accounts to the outside within a
few seconds.
The potential value of a McGraw-Hill "Pool" raises the possibility of
providing access to all McGraw-Hill publications on a "Bulk purchase plan".
Users could purchase "Books" of a few thousand pages (remember web pages
are very small) for $30-50 and be able to access those pages repeatedly
without extra charges. Supplementing them with dynamic content such as
S&P Marketscope and Daily News can generate a steady flow of return
business and new "book" purchases.
V. Resource requirements
We should begin with a prototyping system (Linux or Unix, basic service)
and a series of phased enhancements.
Prototype phase:
| A. Pentium CPU as prototype server, Linux or BSDI OS. |
$5000 |
| B. Configure servers, routers, fire-walls. |
5 days. |
| C. Httpd web server porting and configuration. |
5 days. |
| D. Billing system configuration - |
10 days. |
| E. Oracle and WAIS front-ends. |
5 days. |
| F. Convert Marketscope menus to Web Pages |
10 days. |
| G. Convert Marketscope forms to Web Pages - |
10 days. |
| H. Royalty Negotiations |
10 days. |
Dependencies A:B B:CD C:FG D:H E:G
Conversion of marketscope forms may be automatable.
Phase 1 beta:
| MarketScope Alert to HTML driver. |
10 Days. |
| Royalty negotiations with 3rd parties. |
5 Days. |
| Enable Secure HTML. |
3 Days. |
| Blocking Strategy. |
2 Days. |
| Enable "Feed Rate" Customers. |
3 Days. |
| Start Billing. |
2 Days. |
| Total |
25 Days. |
Phase 1 production:
| Promotion. |
10 Days |
| Inclusion on other databases (lycos, WAIS,...). |
3 Days |
| Inclusion links on strategic web pages (Compuserve, America
Online...) |
|
| Inclusion links to strategic web pages (Edger, MIT stats...) |
| Total |
20 Days |
Phase 2:
| Stock Reports text to HTML. |
10 Days |
| Stock Reports Graphics to JPEG and GIF. |
10 Days |
| Arrange advertizing (Initial). |
5 Days+ |
| Arrange other content(Initial). |
5 Days+ |
| Secure Sockets Distribution. |
5 Days |
Phase 3:
| Edgar SGML to HTML parser. |
10 Days |
| Edgar EDI to HTML parser. |
10 Days |
| Access other content. |
5 Days |
Phase 4:
| Based on traffic, Migrate critical functions to alternate
platforms. |
20 Days |
| Based on traffic & Royalty arrangements, optimise resource
usage. |
20 Days |
| Performance enhancements. |
20 Days |
You are visitor #
to site.
Return to Author's
Home Page