=-- This distribution rendered properly... -=
=---- http://www.prairienet.org/~yamada ----=
=---- mailto:yamada@acm.org (haiku) --------=
=-- when these lines have equal length. ----=
The Incarna Parser (TSX/Web -- Satisfaction Engine, Web edition)
Architectural framework design document, version 0.2.1.
This release: 21 December 1999
7,236 words, max line length = 67 chars
Comments: yamada@acm.org
Copyright 1999 by William Paul Fiefer, all rights reserved.
Copyright specifics: http://www.prairienet.org/~yamada
"...he who believes that a clock has soul
attributes the maker's glory to the work..."
-- Johannes Kepler,
letter to Herwart von Hohenburg, 1605
===========================================
I N C A R N A
T S X / W E B
Satisfaction Engine
- Web edition -
D E S I G N D O C U M E N T
Analytic processes
===========================================
======================
C O N T E N T S
======================
------------------
[01-00] What's New
------------------
[01-01] What's new?
----------------
[02-00] Overview
----------------
[02-01] What is this document?
[02-02] What background do I need to use this document?
[02-03] What do TSX/Web and the Satisfaction Engine do?
[02-04] How does TSX/Web define a site?
[02-05] How do TSX/Web and the Satisfaction Engine work?
[02-06] Who needs TSX/Web?
---------------------------
[03-00] Satisfaction Engine
---------------------------
[03-01] Why do we discuss the Satisfaction Engine first?
[03-00] What is the central idea of the Satisfaction Engine?
[03-00] What is the Satisfaction Engine?
[03-00] What is an Injection Console?
------------------
[00-00] Indicators
------------------
[00-00] What are indicators?
[00-00] What are fundamental indicators?
[00-00] What are some TSX/Web indicators?
------------------
[00-00] Dimensions
------------------
[00-00] What are dimensions?
---------------
[00-00] Sources
---------------
[00-00] What are sources?
[00-00] What are fundamental sources?
[00-00] What are some TSX/Web sources?
---------------
[00-00] Weights
---------------
[00-00] What are weights?
------------------
[00-00] Transforms
------------------
[00-00] What are transforms?
[00-00] What is the proposed model for building transforms?
[00-00] What are fundamental transforms?
[00-00] What are some TSX/Web transforms?
--------------
[00-00] Graphs
--------------
[00-00] What are graphs?
----------------
[00-00] Analyses
----------------
[00-00] What is an analysis?
[00-00] What are some TSX/Web analyses?
[00-00] What are the most important TSX/Web analyses?
---------------
[00-00] Domains
---------------
[00-00] What is a domain?
[00-00] What are some TSX/Web analysis domains?
-------------------
[00-00] Methodology
-------------------
[00-00] What role does methodology play in TSX?
[00-00] What does the outline of the methodology look like?
--------------------
[00-00] SourceVision
--------------------
[00-00] What is SourceVision?
----------------------
[00-00] TSX/Web design
----------------------
[00-00] How is TSX/Web designed?
[00-00] What language is TSX/Web programmed in?
-----------------------------
[00-00] Project, introduction
-----------------------------
[00-00] What needs to be done?
[00-00] What if I want to help?
[00-00] How do I highlight my modifications?
-------------------------
[00-00] Project, notebook
-------------------------
[00-00] Where do I put uncataloged indicators?
[00-00] Where do I put uncataloged sources?
[00-00] Where do I put uncataloged transforms?
[00-00] Where do I put uncataloged graphs?
[00-00] Where do I put uncataloged analysis ideas?
[00-00] Where do I put analysis domains?
[00-00] Where do I put major issues?
=================================
[00-00] W H A T ' S N E W
=================================
TSX/Web is the draft name only of this project and is used
throughout the remainder of this document; the master project
final name is `Incarna.'
=------------------
[00-00] What's new?
=------------------
[06 October 1999]
The operation of TSX/Web (but not its general function or design)
is extended significantly.
==>> Definition of a site recast
TSX/Web can now reside on one or more machines and analyze the
network traffic at that (those) machine(s). TSX/Web can interpret
this traffic as originating from a single site, regardless of the
true source of the traffic.
Put another way, TSX/Web can parse the traffic at a proxy server
or network interface and think that all of it is local.
==>> The redefinition produces more accurate usage analyses
TSX/Web produces a report analyzing the browsing habits and
information use patterns indicated by the traffic stream. All
TSX/Web transforms used in a site analysis remain valid in
parsing this stream. Data gathered through interviews and other
sources is used to validate and extend the stream analysis.
Placing TSX/Web at the machine responsible for connecting a large
entity like a corporation or university to the Internet produces
the information consumption analysis of that entity's staff.
Placing TSX/Web at a single workstation produces a similar
analysis on a single user.
==>> The redefinition is a more powerful abstraction
The TSX/Web view of the places browsed as being one site adds
more information. TSX/Web can analyze the structure of a single
site, as defined traditionally, but if that site's users consume
one percent of their pages there and 99 percent from a remote
site, the analysis is probably moot.
Better then to consider the sites users visit (consume) as a
single site, "the site of User X," rather than persist in
preserving provincial, localized definitions. Each user -- each
entity -- works inside a personal site built of all of the sites
consumed. This is a higher level of abstraction that later can be
factored into the traditional definition of a site.
Further, building the TSX/Web transforms that parse a single
traffic stream is easier, by magnitudes, than writing the same
transforms to traverse complex data structures and merge data
from multiple streams.
[21 September 1999]
Added sections `Project, Introduction' and `Project, Notebook'
Removed section `Project.'
[15 September 1999]
Edited document to focus on TSX/Web.
[14 September 1999]
Added `SourceVision' section.
[10 September 1999]
Added the sections `What's New' and `Satisfaction Engine.'
[09 September 1999]
Added insurance risk rating systems to the candidate models for
TSX/Web analytic processes.
Selected an augmented Web browser as the candidate model for the
injection console.
[08 September 1999]
Selected credit and bond rating systems as candidate models for the
TSX/Web analytic processes.
Selected Perl as a candidate language for programming TSX
transforms.
Added an enclosing category `Injection Consoles' to describe the
class of software TSX belongs to.
[04 September 1999]
Minor clarifications.
[03 September 1999]
Rearranged sections.
Added indicator `dimensions.'
Added `Dimensions' section.
Added `Domains' section.
Changed analysis `category' to `domain.'
Changed the comment and change notation.
Deprecated the comment and change notation.
Added HTML so this document works cleanly in a browser.
[31 August 1999]
Minor clarifications.
Quotations.
[30 August 1999]
Changed TSX to TSX/Web.
Defined an audience.
Added sample reports.
Clarified many points.
Simplified the rules for adding comments and changes.
Moved the `Transforms' section.
Inverted the analysis domains tree in `Project.'
Globally changed `composite' to `compound.'
[29 August 1999]
Changed TGW to TSX.
Changed `report' to `analysis.'
Changed `scales' to `graphs.'
Added `transforms.'
Added `Transforms,' `Weights,' and `Graphs' sections.
Changed `Random' section to `Project' section.
[17 July 1999]
TGW.
===============================
[00-00] O V E R V I E W
===============================
"In the discovery of secret things and in the investigation of
hidden causes, stronger reasons are obtained from sure experiments
and demonstrated arguments than from probable conjectures and the
opinions of philosophical speculators of the common sort."
-- William Gilbert
=-----------------------------
[00-00] What is this document?
=-----------------------------
This is the TSX/Web master design document. It tells you what
TSX/Web is and how it works. It also tells you a little about how
TSX/Web will be marketed.
This document is in text format and any viewing tool can render it.
Read `What's new?' for an overview of major changes in a release.
Minor editorial changes are not marked or noted. If you've missed
reading more than the previous release, you should skim this entire
document. The overall design of TSX/Web is stable, so the important
changes will stand out. Catching up won't take you more than five
minutes.
Otherwise:
* If you want a quick sense of what TSX/Web is, read the
`Overview' and perhaps `Satisfaction Engine.'
* If you want to participate in this project and are reading this
for the first time, go through it from beginning to end.
* If you are adding to this project, everything you need frequent
contact with is at the end of this file in the `Project'
section. Go right to the end now.
=------------------------------------------------------
[00-00] What background do I need to use this document?
=------------------------------------------------------
You have it. But a few points need highlighting.
(1) This document assumes the information stream flowing between
computers (the traffic) can be examined (parsed) and analyzed
semantically just as the conversation stream between a person and
anyone they communicate with can be analyzed.
(2) This document expands the concept of a browser. Any device that
can link to and exchange information over the Internet (more
broadly a TCP/IP network) is a browser.
(3) This document expands the concept of a site.
* It includes single unified sites (traditional sites):
www.mycorp.com
www.big-university.edu
www.national-defense.mil
* It includes tightly linked sites (virtual sites):
www.big-firm.com
www.supplier-to-big-firm.com
* It includes sites grouped for specific purposes:
www.university-one.edu
www.university-two.edu
www.university-three.edu
* It includes the sites visited by a user, group, department,
firm, state, nation -- any defined entity -- over a specific
time period or having specific attributes.
=------------------------------------------------------
[00-00] What do TSX/Web and the Satisfaction Engine do?
=------------------------------------------------------
TSX/Web is an edition of the Satisfaction Engine.
The Satisfaction Engine, from here on TSX, examines data about a
process (domain) and produces analyses about what it finds. This is
the Web edition (Web module) of TSX, from here on TSX/Web. It
examines Web sites and produces analyses about what it finds. Put
another way, TSX/Web takes as input one or more sites plus a set of
queries concerning those sites. From this it produces an analysis
of that (those) site(s).
For example, TSX/Web can
* Determine, describe, and assess the site value as an Internet
property or production zone
* Identify, describe, and assess which global
business-to-business strategies the site implements
* Determine, describe, and assess how well the site implements
global business-to-business strategies
* Determine, describe, and assess how well site content retains
customers
* Identify usage (consumption) patterns for a one or more users
* Describe the cognitive site (defined in the segment immediately
following) constructed by one or more users
TSX/Web does this by snooping network traffic, by traversing the
site to collect data, and by taking as input data collected from
documents and other non-Web sources and from interviews. TSX/Web
takes this collected data and performs a series of calculations on
it to generate one or more reports on specific areas of interest to
a particular client.
=--------------------------------------
[00-00] How does TSX/Web define a site?
=--------------------------------------
In two primary ways:
(1) TSX/Web can be configured to think all network traffic in and
out of one or more machines is a site. TSX/Web parses this
traffic stream as if it were one site.
When this definition is applied to one user or a defined
cluster of users, it is sometimes called a `cognitive site.'
(2) TSX/Web can be configured to analyze a site as traditionally
defined, which is a physically distinct collection of
machines offering information.
=-------------------------------------------------------
[00-00] How do TSX/Web and the Satisfaction Engine work?
=-------------------------------------------------------
TSX/Web (and any edition of the Satisfaction Engine) is formed of
nine component ideas:
(1) Indicators -- the attributes of a Web site
(2) Dimensions -- the values assigned to indicators
(3) Sources -- providers of information about indicators
(4) Weights -- relative rankings of indicator values and sources
(5) Transforms -- converters of indicators into information
(6) Graphs -- pictures that compare and contrast
(7) Analyses -- TSX/Web conclusions
(8) Domains -- groupings of similar analyses
(9) Methodology (SourceVision brand name) -- making the analyses
The following are not now TSX/Web components:
* Queries, which determine the analyses TSX/Web can produce;
* Sites, which are what TSX/Web examines;
* The Injection Console (defined in the `Satisfaction Engine'
section).
Given one or more sites and one or more queries, this is how the
nine pieces combine to produce an analysis:
INDICATORS and SOURCES are assigned WEIGHTS that signify their
relative importance in a particular analysis. A specific
DIMENSION is assigned to an indicator based on information
provided by a source. Indicator and source information is fed
to a set of TRANSFORM routines that produce GRAPHS of their
characteristics. The graphs are used to make ANALYSES (reports,
white papers). Analyses are grouped into DOMAINS to make them
easier to market (sell). All of this is controlled by a
METHODOLOGY, which has the brand name `SourceVision.'
TSX/Web turns sites into numbers and numbers into analyses. It's
essentially a data conversion utility.
You could also say TSX/Web indicators and sources are the raw
materials that make TSX/Web analyses (products), and the TSX/Web
methodology is how this factory runs.
=-------------------------
[00-00] Who needs TSX/Web?
=-------------------------
Anyone wanting answers to questions like this:
* How does our site fit into our strategic business plans?
* Do our Web development services deliver what they claim?
* Are we optimizing our use of our site?
* How coherent is the message we deliver to our clients?
* How coherent is the message we deliver to our employees?
* How coherent is the message we deliver to our investors?
* What do our competitors want to accomplish with their site?
* Has our site delivered a return-on-investment?
* What is the market value of our site? Our competitor's sites?
* Does our site promote the employee interaction we need?
* Is our site designed to evolve as our business evolves?
For example:
* Vice presidents of marketing
* Digital publishers
* Brand managers
* Executive producers
* Interactive marketing managers
* Public relations managers
* New media managers
* E-commerce managers
* Web site content managers
* Channel marketers
* Editors
If you're responsible for the success of a strategic information
complex, if you plan to launch or invest in a major Web venture, or
if you are considering the business valuation of an Internet
property, you need TSX/Web.
================================================================
[00-00] S A T I S F A C T I O N E N G I N E
================================================================
=-------------------------------------------------------
[00-00] Why do we discuss the Satisfaction Engine first?
=-------------------------------------------------------
We start with the Satisfaction Engine because TSX/Web is an edition
of the Satisfaction Engine.
=-----------------------------------------------------------
[00-00] What is the central idea of the Satisfaction Engine?
=-----------------------------------------------------------
The central idea of the Satisfaction Engine (and TSX/Web) is that
any aspect of a process (for TSX/Web, this is a Web site) can be
represented through a set of variables called `indicators.'
Indicators describe the characteristics of a process.
=---------------------------------------
[00-00] What is the Satisfaction Engine?
***
Indicators, sources, and analyses.
=---------------------------------------
The Satisfaction Engine (TSX) is the software representation of a
process analysis methodology. This edition (module) of TSX focuses
on the World Wide Web and is called TSX/Web. The methodology steps
are
(1) Auditing
(2) Benchmarking
(3) Competitive analysis (comparison and contrast)
(4) Differentiation (prescriptive analysis)
All domain-specific TSX editions work similarly because the same
TSX architecture is at the core.
TSX/Web tells you the effectiveness of a site as measured from many
perspectives. You can use this information as input to other
analyses.
TSX/Web looks at the technical and social side of a Web site,
analyzing processes inherent to the site as well as processes that
interact with or are emergent from the site. For example, TSX/Web
can analyze the network response time of a site (an inherent
process), user perceptions of that response time (an interactive
process), and whether improving that response time will pay off in
improved productivity (an emergent process). This last example
means TSX/Web can measure how the site fits into the community it
serves.
If there is no design specification, TSX/Web can derive one, in
effect telling you what a Web site is really doing. This feature
will tell you what a competitor is doing with their site and what
your site's designers are really up to.
TSX/Web does this by looking at Web site indicators that it
collects from sources. TSX/Web can, for example
* Determine the degree of deviance from a design specification
* Compute a set of social indicators about a site
* Generate ad hoc performance data about a site
The indicators that provide information about the social aspects of
a site are a superset of the indicators covering the technical side
of the site. They add human- and social-factor indicators.
Indicators are not all equal. Some are more important in any
particular analysis and are weighted to reflect their relevance
compared to other indicators used in that analysis.
To perform the calculations behind its analyses, TSX/Web relies on
a set of transforms; these transforms are modules added to the
Satisfaction Engine. The Satisfaction Engine is the heart of the
system. The transform modules make the Satisfaction Engine
extremely flexible. For each particular aspect of a site TSX/Web
analyzes there is a corresponding transform in the Satisfaction
Engine.
TSX/Web uses the analytic observation tools of business, finance,
advertising, communications, journalism, social science, behavior
science, and computer performance testing among others. Each of
these fields measures soft, hard-to-isolate phenomena.
TSX/Web's calculation routines function as do the quantitative
routines used to produce credit ratings, track financial markets or
predict weather patterns. For example, if network response time
effects productivity, then it might be represented similar to a
coefficient of friction or drag. Again, each of these fields
measures soft, hard-to-isolate phenomena.
TSX/Web examines its conclusions and evaluates their quality based
on the nature of the site indicators and sources. The output is a
reliability value; an analysis with a reliability value is called a
validated analysis.
=------------------------------------
[00-00] What is an Injection Console?
=------------------------------------
The Satisfaction Engine belongs to a class of software I call
`Injection Consoles' because they insert and control the action of
other programs (here, the transforms) in a new environment. An
Injection Console is a marketing term for an `agent.'
By adding this class of software the TSX design can extend
horizontally while the moment at which the code and design break is
delayed.
The candidate model for the injection console is an augmented Web
browser.
=====================================
[00-00] I N D I C A T O R S
=====================================
The attributes of a Web site.
=---------------------------
[00-00] What are indicators?
***
What TSX/Web wants to know.
=---------------------------
TSX/Web collects specific pieces of data from sources. It looks for
Entities (objects) -- pages, for example
and records the
Behaviors (processes)
and
Characteristics (attributes)
of those objects. The characteristics (attributes) are
`indicators.'
This deliberately mimics how databases work so TSX/Web will have a
more natural interface to relational database management systems.
Indicators are assigned weights -- multipliers -- that reflect
their importance relative to other indicators. See the section
`Weights' in this document for a discussion of weights.
The rules of the TSX/Web methodology control the permissible
TSX/Web vocabulary. This simplifies TSX/Web's internal design
because the methodology requires that each indicator object be
expressed in terms that are also used in TSX/Web's programming
modules. This lets us program TSX/Web in terms of the subject
vocabulary (Web sites) rather than in terms of the machine
vocabulary (computer jargon).
Indicators fit into a tree hierarchy. The value of a indicator
reduces completely to the indicators composing it. For example, the
value of the indicator `site' is a compound of all of its
components. (You'll see this shortly.)
[Note: When you make additions to the Indicators tree, preserve its
structure or submit a tree redesign. (The tree is displayed soon.)]
The value of the topmost Indicator element is the same as the site
satisfaction value, which measures how well the site meets the
criteria on which TSX/Web judges it.
***
The presence or absence of any given indicator is noted by TSX/Web,
and that plus a weight (multiplier value) assigned to the indicator
is handed to the validation routine.
Normally within a hierarchy we can omit (factor out) levels that
have only one level below them, as do Indicators and Site. We keep
the redundancy so we can add new arms to the hierarchy later while
delaying the time when our system breaks.
Indicators that still need to be positioned in the tree are called
uncataloged indicators.
=---------------------------------------
[00-00] What are fundamental indicators?
=---------------------------------------
Fundamental indicators are indicators whose value are not
determined using other indicators. Every indicator at the leaves of
TSX/Web's Indicators hierarchy is fundamental.
Indicators with values computed using the values of other
indicators are compound indicators.
For example, if a indicator called
Daily-Outbound-Throughput
measures the number of bytes sent out from a site each day and is
computed by dividing 30 into the indicator
30-Day-Outbound-Throughput
then it is compound.
Most interesting indicators are compound.
=----------------------------------------
[00-00] What are some TSX/Web indicators?
=----------------------------------------
Some sample indicators follow. When you make additions to the
Indicators tree, preserve this structure or submit a tree redesign.
This sample is for illustration only and is not a final skeletal
form.
...................................................................
+-Indicators
+-Site ((URLs, people, and platforms?))
+-URLs
+-Quantity
+-Total
+-Frames
+-Type
+-MIME
+-Purpose
+-Form
+-Applet
+-Information
+-URL page indicators
+-Byte size
+-Number of images
+-Number of internal links
+-Number of external links
+-Tags
+-BASE
+-IMG ALT
+-IMG HEIGHT
+-IMG WIDTH
+-Passwords
+-Tools
+-Quantity
+-Total
+-Type
+-Browsers
+-Type
+-Plugins
+-Quantity
+-Total
+-Languages
+-Quantity
+-Total
+-Type
+-HTML
+-Version
+-Java
+-Version
+-JavaScript
+-Version
+-Perl
+-Version
+-Platforms
+-Quantity
+-Total
+-Servers
+-Quantity
+-Total
+-Type
+-Web
+-Application
+-Purpose
+-Main
+-Mirror
+-People
+-Quantity
+-Total
+-Purpose
+-Developers
+-Quantity
(( position these indicators ))
+-Throughput
+-Inbound
+-Outbound
...................................................................
=====================================
[00-00] D I M E N S I O N S
=====================================
The values assigned to indicators.
=---------------------------
[00-00] What are dimensions?
=---------------------------
Dimensions are the values assigned to indicators. If an indicator
has a weight other than `1,' its dimension is this weight
multiplied by its value as determined from its source. A `final
dimension' always refers to the weight-value multiple.
============================
[00-00] S O U R C E S
============================
Providers of information about indicators.
=-----------------------------------
[00-00] What are sources?
***
Where TSX/Web looks for information.
=-----------------------------------
TSX/Web determines the characteristics of a Web site using input
from interviews, measurements, descriptions, documentation, and
other repositories. These repositories are `sources.'
TSX/Web can automate certain parts of this information gathering
while others parts must be completed by humans filling out forms.
The TSX/Web methodology governs the layout of all such forms so
they have the uniformity needed to later implement them in such
electronic tools as PDAs.
Sources fit into a tree hierarchy. The information provided by a
source reduces completely to the sources composing it. For example,
the information from the element
`Site'
is a compound of all of its components.
[Note: When you make additions to the Sources tree, preserve its
structure or submit a tree redesign.]
Sources are assigned weights -- multipliers -- that reflect their
importance relative to other sources. See the section `Weights' in
this document for a discussion of weights.
The presence or absence of any given source is noted by TSX/Web,
and that plus a weight assigned to the source is handed to the
validation routine.
Normally within a hierarchy we can omit (factor out) levels that
have only one level below them, as do Sources and Site. We keep the
redundancy so we can add new arms to the hierarchy later while
delaying the time when our system breaks.
Sources that still need to be positioned in the tree are called
uncataloged sources.
=------------------------------------
[00-00] What are fundamental sources?
=------------------------------------
Fundamental sources are sources whose value is not determined using
other sources. Every source at the leaves of TSX/Web's Sources
hierarchy is fundamental.
A source having a value computed using the values of other sources
is a compound indicator.
=-------------------------------------
[00-00] What are some TSX/Web sources?
=-------------------------------------
Some sample sources follow. When you make additions to the Sources
tree, preserve this structure or submit a tree redesign. This
sample is for illustration only and is not a final skeletal form.
...................................................................
+-Sources
+-Site
+-Objects
+-Network topology
+-Production tools
+-Process
+-Type
+-Copywriting
+-People
+-Staff
+-Consultants
+-Documents
+-Design specifications
+-Work orders
+-Mission statement
+-User surveys
+-Design
+-People
+-Staff
+-Consultants
+-Documents
+-Design specifications
+-Work orders
+-Mission statement
+-User surveys
+-Development
+-People
+-Staff
+-Consultants
+-Documents
+-Design specifications
+-Work orders
+-Mission statement
+-User surveys
...................................................................
============================
[00-00] W E I G H T S
============================
The relative importance of indicator values and sources.
=------------------------
[00-00] What are weights?
=------------------------
The value of a particular source is not necessarily equal in
importance to other sources. Nor is the value produced by analyzing
a particular indicator of equal importance to other indicators.
More important sources and indicators are assigned multiplier
values or `weights.' These weights are not constant and may differ
when that source or indicator is used in a different analysis.
When a source or indicator has been evaluated, it is either final
or is used in subsequent calculations. Every source or indicator
is multiplied by its weight prior to being used in a transform.
A weight can have a negative value. An infinite weight is undefined
and must trigger an error recovery activity.
Depending on context, the weights `0', `1', and `-1' mean
0 1 -1
--- ----
absent present
false true
no yes
invalid valid
unreliable reliable
=====================================
[00-00] T R A N S F O R M S
=====================================
Converters of indicators into information.
=---------------------------
[00-00] What are transforms?
=---------------------------
TSX/Web produces analyses by converting a set of indicators into
useful information. The algorithms that manipulate indicators are
called transforms. They define the TSX edition.
A transform may be fundamental or compound. compound transforms
must build entirely upon new transforms, reuse old transforms, or
combine old and new transforms.
For each particular analysis there are defining transforms in the
Satisfaction Engine. A given analysis must reuse transforms,
require new transforms, or combine old and new transforms.
Transforms that provide valuable information but still need to be
assigned to one or more analyses are called uncataloged transforms.
=----------------------------------------------------------
[00-00] What is the proposed model for building transforms?
=----------------------------------------------------------
The analytic systems using by credit bureaus and rating houses to
produce credit and bond ratings.
=---------------------------------------
[00-00] What are fundamental transforms?
=---------------------------------------
Fundamental transforms are transforms that are not built using
pre-existing transforms.
A transform built using other transforms is a compound transform.
=----------------------------------------
[00-00] What are some TSX/Web transforms?
=----------------------------------------
You decide. For starters, we suggest, by domain and data structure:
Content
-------
* Fog (readability) index
(INTEGER)
Usage
-----
* Popular links
(TABLE)
* Frequency of search engine use
(INTEGER-RATIO:SEARCHES/USERS)
* Common search strings
(TABLE:ALPHANUMERIC)
* Time spent per page
(INTEGER:SECONDS)
* Pages used en route to a target
(HASH TABLE)
* Common en route pathways
(TABLE)
Structure
---------
* Average page structural complexity
(INTEGER)
* Prevalent links
(TABLE)
=========================
[00-00] G R A P H S
=========================
Pictures that compare and contrast.
=-----------------------
[00-00] What are graphs?
=-----------------------
Analysis results are plotted along two or more indicators. These
plots are called `graphs.' When two or more sites are analyzed and
plotted on a single graph to compare and contrast their
characteristics, we have the information required for benchmarking,
competitive analysis, and differentiation analyses.
Graphs are descriptive and predictive.
* Audit analyses graphs are predominantly descriptive.
* Benchmarking analyses graphs are predominantly descriptive.
* Competitive analyses graphs are predominantly prescriptive.
* Differentiation analyses graphs are predominantly prescriptive
and predictive.
===============================
[00-00] A N A L Y S I S
===============================
"If useful data is surrounded by useless data, whether by accident
or design, you can get the useful data anyway (if you know what
the useful data looks like).
"If you have useful data it can tell you what you need to know.
"If the useful data doesn't tell you what you need to know, it can
still tell you more than what anyone else knows."
-- Luke Hawley
=---------------------------
[00-00] What is an analysis?
***
TSX/Web conclusions.
=---------------------------
An analysis is an answer to a line of inquiry.
TSX/Web describes a Web site by producing one or more analyses. The
analyses describe specific aspects of a Web site. Analyses are
descriptive, prescriptive, and predictive.
* Audit analyses are predominantly descriptive.
* Benchmarking analyses are predominantly descriptive.
* Competitive analyses are predominantly prescriptive.
* Differentiation analyses are predominantly prescriptive and
predictive.
TSX/Web analyses are grouped into domains so that similar analyses
can be
* Identified and performed as a group
* Market and sold as a group
* Classified for machine understanding
* Programmed modularly
* Developed independent of anything Web-related
=--------------------------------------
[00-00] What are some TSX/Web analyses?
=--------------------------------------
Here are some sample analyses. These samples are for illustration
only and should not govern the scope of what TSX/Web can report on.
(The first sample analysis includes a sample indicator and its
sample source.)
* How well a site implements the five core technologies
(1) Automation
Sample indicator: Using many software tools
Sample source: A tools inventory
(2) Collaboration
Sample indicator: Using an interactive network
Sample source: The presence of groupware
(3) Customization
Sample indicator: Using task-specific interfaces
Sample source: The software documentation
(4) Knowledge access
Sample indicator: Using a database
Sample source: The presence of a database
(5) Speech and other perceptual capabilities
Sample indicator: Using machine vision
Sample source: Image input systems for computers
* The business valuation of a site
* The quality of a Web development environment
* The appropriateness of the Web development tools
* The maturity of the Web development staff skills
* The balance and interplay between open and proprietary systems
* The ideas communicated by the site
* The uses made of a site by particular departments
-=---------------------------------------------------
[00-00] What are the most important TSX/Web analyses?
=----------------------------------------------------
Here are some suggestions. These are for illustration only and
should not govern the final set of key TSX/Web reports. Look at the
TSX uncataloged analysis ideas at the end of this document.
* The purchase value of a site
* The degree to which a site promotes cross-disciplinary
interaction
* How well the site supports an online community
* The consistency of the site's cognitive model
* The clarity of a site's message
* The cognitive load created by a site
* The accessibility of the site to users of differing abilities
* How well a site uses the five core technologies
* The validity of a site's specification
* How well the site specification is met
============================
[00-00] D O M A I N S
============================
Groupings of similar analyses.
=------------------------
[00-00] What is a domain?
=------------------------
A group of similar analyses. TSX/Web analyses are grouped into
domains so that similar analyses can be
* Identified and performed as a group
* Market and sold as a group
* Classified for machine understanding
* Programmed modularly
* Developed independent of anything Web-related
=----------------------------------------------
[00-00] What are some TSX/Web analysis domains?
=----------------------------------------------
Here are some sample analysis domains. These samples are for
illustration only and should not govern the scope of TSX/Web
analysis domains.
* Business and accounting assessments
* Network performance assessments
* Software engineering assessments
* Physical structure assessments
* Conceptual structure assessments
* Editorial performance assessments
* Group identity and community building assessments
* Impact/vision assessments
* Community-Web interaction assessments
========================================
[00-00] M E T H O D O L O G Y
========================================
Guidelines for making analyses from indicators.
=----------------------------------------------
[00-00] What role does methodology play in TSX?
***
The methodology controls TSX's behavior.
=----------------------------------------------
The methodology has four parts:
(1) Audit
(2) Benchmarking
(3) Competitive analysis (comparison and contrast)
(4) Differentiation (prescriptive analysis)
TSX/Web is the software rendition of this site analysis
methodology. The methodology has the brand name `SourceVision.'
A large piece of TSX/Web centers on analysing the indicators that
render judgments on specific aspects of a Web site.
=----------------------------------------------------------
[00-00] What does the outline of the methodology look like?
***
The methodology looks like a project plan.
=----------------------------------------------------------
A sample outline follows. When you make additions to this area,
preserve the sequential structure of the methodology.
The methodology *must* be operational regardless of whether the
Satisfaction Engine is used. The system must run off paper.
...................................................................
--Define project scope
--Run audit
--Collect indicators
--Identify target resources
--Construct target resources identity database
--Locate target resources
--Construct target resources location database
--Interrogate target resources
--Construct target resources response database
--Validate indicators
--Unnecessary
--Missing
--Unreliable
--Degree
--To come
--When
--Reliable
--Degree
--Transform indicator data
--Organize indicator data
--Analyze indicator data
--Synthesize indicator data
--Produce Satisfaction score
--Produce Validity score
...................................................................
============================================
[00-00] S O U R C E V I S I O N
============================================
=----------------------------
[00-00] What is SourceVision?
=----------------------------
SourceVision is the brand name of the TSX methodology set. The
TSX/Web methodology is a component of SourceVision.
==============================================
[00-00] T S X / W E B D E S I G N
==============================================
=-------------------------------
[00-00] How is TSX/Web designed?
***
Objects and components.
=-------------------------------
Briefly...
TSX/Web analyses are object-oriented and the objects are culled
from the Web domain. The TSX/Web analysis procedure
(1) Breaks a site into entities (objects)
(2) Examines the behaviors (processes) and
characteristics (attributes) of the entities and analyzes
these.
Where entities interact, TSX/Web describes the interaction using a
standardized vocabulary (API). This means
(1) The language TSX/Web uses to do its work is the same as the
language it is programmed in;
(2) TSX/Web easy to design and modify.
Internally, TSX/Web is formed of three main pieces:
(1) Input: Collects information and converts it to
machine-readable formats
(2) Primary logic: Performs a validated analysis on a set of
indicators (the transforms inside the Satisfaction Engine);
the Satisfaction Engine belongs to a class of software I call
`Injection Consoles' because they insert and control the
action of other programs (here, the transforms) in a new
environment
(3) Output: Takes the analysis and formats it for human use
By keeping the primary logic of the Satisfaction Engine
mechanically distinct from the TSX/Web input and output, and by
defining transforms as components of the Satisfaction Engine, we
can use the Satisfaction Engine as a component in other software
that renders judgments on data collected in other, unrelated
domains.
We minimize the number of states the TSX/Web and its transforms can
enter and we simplify each valid state. State means `mode,' and the
minimal modes for any program are STARTED, RUNNING, and STOPPED.
But these distinctions are not enough. RUNNING, for example, does
not tell us what evaluation is running. So we have more RUNNING
states with each built as simply as possible.
=----------------------------------------------
[00-00] What language is TSX/Web programmed in?
=----------------------------------------------
One proposed language is Perl for the transforms.
Perl is powerful enough to express all the ideas found in TSX/Web
and Perl routines can be written and modified quickly. I think TSX
should be written to accept Perl transforms in addition to any
other transform language.
TSX itself now has no candidate language. But TSX must be written
in an object-oriented language that uses its forming components
(indicators, dimensions, sources, weights, transforms, graphs,
analyses, domains, methodology, and possibly injection consoles)
directly in its syntax.
===================================================================
[00-00] P R O J E C T , I N T R O D U C T I O N
===================================================================
"Day was departing, and the darkening air
Called all earth's creatures to their evening quiet
While I alone was preparing as though for war..."
-- The Inferno of Dante, Canto II
=-----------------------------
[00-00] What needs to be done?
=-----------------------------
Most important, we need to know who might be interested in using
TSX/Web.
Next, we need a list of the most important TSX/Web analysis
domains and analyses. A simple set is best, focusing on what we
want to know about a site, its processes, and environment.
Then we need a collection of methodologies. While the general TSX
methodology will be unchanged among analyses, specific aspects of
the methodology must be tailored to each domain. I believe research
methodologies drawn from each domain provide the best tools for
tailoring the general methodology.
Later, we will need to know where to find information to help us do
this and how we interpret that information. We need the indicators
for the analyses, the sources for the indicators, and the
calculation algorithms (transforms).
That means asking yourself:
(1) Who are TSX/Web's potential clients?
(2) What do clients want analyses on?
(3) What domain does that analysis fall under -- if any?
(4) What research exists about similar aspects of the domain?
(5) What indicators underlie the analyses?
(6) Where do we find those indicators?
(7) How do we weigh and combine the indicators?
Alternatively:
(1) Who needs TSX/Web analyses?
(2) What domain can analyses fall under?
(3) What analyses belong in those domains?
(4) Who has looked at similar aspects of the domain and how?
(5) What indicators underlie the analyses?
(6) Where do we find those indicators?
(7) How do we weigh and combine the indicators?
=------------------------------
[00-00] What if I want to help?
=------------------------------
Everything in this document is tentative. Supplement, extend, or
change the ideas here and send me your changes.
You can work across the entire range of TSX/Web's abstraction
model, from the top suggesting analyses we can produce with
TSX/Web, or from the bottom adding indicators. We are not
programming, we are shaping the system: TSX/Web must run off paper,
as a functioning methodology, before it can run properly as code.
=-------------------------------------------
[00-00] How do I highlight my modifications?
=-------------------------------------------
Don't.
There is a notation for indicating comments and changes. It is ugly
and no fun to use so consider making a second document recording
these items and telling me how they fit with the master document.
Why is the notation like this? The notation must
(1) Identify where comments and changes begin and end;
(2) Allow searches for the beginning or end of such blocks;
(3) Not change the format of inserted or surrounding text;
(4) Be easy to machine edit;
(5) Not interfere with HTML.
When you insert comments or changes into the document, sandwich
them inside
((
))
eye-catchers. The eye-catchers either go flush left in the margin
or enclose (inline) the new material. Make it obvious whether the
modification is a comment or change.
For example: ......................................................
Flush left style...
-------------------
((
Hey, Bozo, make an HTML version of this doc.
))
Documents
Design specifications
Work orders
((
Mission statement
))
Inline style...
---------------
This is the TSX/Web master document. (( Sez who? ))
If you are adding to this project, everything you need frequent
contact with is at the end of this file in the `Random' section.
User surveys (( User polls ))
...................................................................
=======================================================
[00-00] P R O J E C T , N O T E B O O K
=======================================================
=---------------------------------------------
[00-00] Where do I put uncataloged indicators?
=---------------------------------------------
Right here.
* Search engine
* Index
* Table of contents
* META tags
* Keywords
* Frames
* Privacy notice
* Email address
* Mouseover
* Click here statement
* Cookies
* Fog index
* PDF
* Fonts
* HTML line length
* Scrolling volume per page
* Design consistency
* Load speeds
* Number of colors used
* Number of external links
* Number of internal links
* Rank-ordered linked-to sites
[14 September 1999]
* Revenue
* Traffic
* Company influence
* Business category
* Clickthroughs
[16 September 1999]
* Stickiness
* Ability to personalize the site
* Page turnover
* Degree or granularity of indexing and searchability
[00 mmmmmmmm yyyy]
=------------------------------------------
[00-00] Where do I put uncataloged sources?
=------------------------------------------
Right here.
* Webmaster
* Stylesheets
[00 mmmmmmmm yyyy]
=---------------------------------------------
[00-00] Where do I put uncataloged transforms?
=---------------------------------------------
Right here.
[00 mmmmmmmm yyyy]
=-----------------------------------------
[00-00] Where do I put uncataloged graphs?
=-----------------------------------------
Right here.
[00 mmmmmmmm yyyy]
=-------------------------------------------------
[00-00] Where do I put uncataloged analysis ideas?
=-------------------------------------------------
Right here unless you have a domain for them. If so, then put the
idea in one of the analysis domains of the preceding section.
* Identify, describe, and assess the online business model
* Identify, describe, and assess organizational changes resulting
from site use
* Determine and describe how well employee training takes into
account the networked environment
* Identify, describe, and assess the processes needing rework or
deprecation
* Determine, describe, and assess how well senior management
supports site content management
* Determine, describe, and assess how well site content is
managed and under what framework
* Determine how much profit e-commerce has produced
* Determine, describe, and assess the site value as an Internet
property or production zone
* Identify, describe, and assess which global
business-to-business strategies the site implements
* Determine, describe, and assess how well the site implements
global business-to-business strategies
* Determine, describe, and assess how well site content retains
customers
* Identify, describe, and assess the methodology for tracking
visitors at the site
* Identify and describe who visits the site and why
* Identify, describe, and assess whether visitors seek
information, employment, or solutions to problems and determine
how easily they find answers at the site
* Determine, describe, and assess how well Web traffic analysis
is used
* Determine, describe, and assess how well Web traffic analysis
is used by customer service, marketing, and sales
* Determine, describe, and assess how well Web-generated customer
data is collected and used (consider efficiency and
effectiveness)
* Determine, describe, and assess how well advertisers help to
ensure quality site content
* Determine, describe, and assess how well the site increases
sales and builds strong electronic partnerships
* Identify and assess whether and to what degree the site is
static or dynamic
* Determine, describe, and assess how well the site will work in
a language other than English.
[22 September 1999]
* How searchable is the site?
* How does the site look to various search engines?
* How well does the site use meta tags?
* How does the site look to various devices?
[00 mmmmmmmm yyyy]
=---------------------------------------
[00-00] Where do I put analysis domains?
=---------------------------------------
Right here. Analyses are of two broad types: social and technical.
The social is a superset of the technical. An analysis may fall
under more than one domain.
SOCIAL
TECHNICAL
* Impact/vision
- Communication of organizational identity
(One of the ultimate goals for the Web)
- Clarity of purpose
* Group identity and community building
- Degree of shared corporate reality:
Is it one firm; do you share one thing
* Editorial performance
- Proofreading
- Content volume
- Eye candy volume
- Content/software integration
- Presentation
* Conceptual structure
- Accessibility
* Physical structure
- Navigability
- Accessibility
- Searchability
- Appearance to various devices
* Software engineering
- Buying into development myths
Lengthy analysis and design is a POOR investment
Cheap hardware/software
Volatile business climate
- Tightness/formality/rigidity of development model
* Network performance
- Data throughput
[00 mmmmmmmm yyyy]
=-----------------------------------
[00-00] Where do I put major issues?
=-----------------------------------
Right here.
[22 September 1999]
* Proposed that the `Project' sections of this document be split
off to a separate document.
[00 mmmmmmmm yyyy]
|
