Table of Contents
Understanding the dynamic complexity of the editorial process
for an employee portal - Lessons learned at Lufthansa German
Airlines
Katrin Schillinger*, Alexander Zock**, Andreas Gro8ler*
* Tndustrieseminar >** Lufthansa Geman Airlines
Manheim University CA/K
D-68131 Mannheim D-60564 Frankfurt/ Main Airport
Gemany Gemany
Phone: +49 621 181-1583 Phone: +49 69 696-46983
Fax: +49 621 181-1579 Fax: +49 69 696-90011
Email: katins@uni mamheimde Email: Aleeander-Zock@dlh.de
agroe@is.bwi.uni mamheimde
Abstract
Corporate intranets and portals have attracted increased attention among information
managers (Detlor 2000). Although intranet and portal developments aim to diminish the
costs for internal information publishing and to increase corporate information dispersal
(Rice 1996, Thyfault and Marx 1996), real-world implementation projects show that for a
successful content management, there are still a lot of open questions. Particularly the
management of a standardized editorial process suffers from clear concepts. To address this
issue, this paper explores the structure of an intranet editorial process, the dynamic
behavior resulting from this underlying structure and presents a first conceptualization for a
successful process management.
1, Introduction
Corporate intranets and portals have attracted increased atfention among infonmation
manages (Dellor 2000). Although intranet and portal devdopments aim to diminish the
costs for intemal infonmetion publishing and to increase corporeie infomation dispasal
(Rice 1996, Thyfault and Marx 1996), real word implementation projects show that for a
successful implementation of an editorial process management environment accompanying a
portal or intranet project no clear concepts have been established in the literawe This
means that there ae no decision nies and infonmetion or key process indicators for
managerial decisions available that could for example help a manager to decide how many
people are needed to manage a given number of intranet pages that are maintained through a
defined editorial process and are subject to clearly defined quality standards (eg. design,
layout, style and age).
The tem process management comprises in this context two major components shown in
figure 1, namely the physical and the infomation network. The physical component depicts
the process structure as illustrated by step 1 to 4 (links between steps are indicated by axows
between the different graphical derents). The infonration network highlights the
accompanying decision rules or policies applied to a process. In figure 1, for reasons of
simplicity, only one policy is shown, which influences step 2 and 3 on the bese of
infonmation derived directly from the process (Infonvetion I; and Infometion I2) or the
process environment. (Information 1; and Information I,). Based on this modd, the graph
indicates that a policy can be thought of as an information processing procedure (Stexman,
2000). Thus, it can be written as:
Policy =f (Infonretion I,, Information b, Information Is, Infonvation L).
f f \
os —
Information I,
Infomation I,
ifgiewesiiewns Information|
network|
Figure 1: Physical and infonmetion structure of a process management (including single
linear and also feedback relations).
To enable such a process management for a portal or intranet editorial process environment
especially the inherent dynamical complexity of the considered physical process needs to be
understood in a more thorough way to support a sustainable success of an intranet project
This fact is reflected in the observation that most intranet process environments are managed
now a days quantitatively in a deficient way if one compares the expected process results
with the used resources and management concepts. The main reason for this situation is
believed to be rooted in a missing quantitative understanding of the dynamical complexity of
a typical intanet editorial environment This finding is not supising as it is in aeareatt
with observations made in other process dynamics related studies (eg. Sterman 2000,
Wann 2002). It is a remarkable finding in these studies that even for structurally extemedy
simple process environment most organizations lack a dear quantitative and/or qualitative
understanding of the related process dynamics (eg. Waren 2002). This situation can mainly
be attributed to the insight that even simple process structures, especially if they include
delays, can produce nonlinear relationships between process input and output parameters
(eg. Sterman 2000, Warren 2002).
To ovewome this limitation, this paper will explore the dynamics produced by a simplified
physical structure of the intanet editorial process as it is implemented at Deutsche
Lufthansa (DLH). The main goal of the study is to understand the dynamical complexity
created by the process strucue in quantitative teams in the frame of the given process
parameters.
To address this issue the presented study starts with a defailed description of the editorial
intanet environment established at DLH Based on this infonvetion a simplified process
model for the further analysis is proposed. This model contains only the physical network of
the editorial process system The infonmetion network including the policy design for the
editorial process environment are intentionally not considered here, but will be analyzed in
defail in a separate paper. Such a two-step process was chosen to reduce the structural and
dynamical complexity of the considered process modd to a mininum It is hoped that by
this restiction a Cearer undestanding of the physical process characteristics can be
achieved, which is required before a successful policy design for a process management
approach can be set up. The simplified process model is then analyzed in two ways: (a) a
System Dynamics (SD) based simulation approach to investigate the system behavior over
time (far away from an equilibium sale) and (6) an equilibrium amalysis appoach to
investigate the equilibrium state of the system in tems of the given system parameters. The
authors believe that such a combined approach is of high interest as both methods provide
complementary insights. SD based simulation mms enable the exploration of an unknown
dynamics hidden in the given structure of a physical process modd far from equilibium
conditions. In contrast to that an analytical equilibrium analysis puis more emphasis on
analytical relationships between system parameters and the system equilibium stale, thus
enabling a clearer undesstanding of the influence of the contol parameters for the system
dynamics.
At the end of the paper different editorial process scenarios based on different parameter sets
are discussed and evaluated for their practical relevance. The outlook gives a summary of
the main results and stresses the need for a policy design (information network) for the
editorial process. In that respect the presented work should only be understood as a first step
towards a complete editorial process management, as the extension of the presented work by
the process information network remains an open issue for further work.
The results presented in this paper are first results of an ongoing research project conducted
at Deutsche Lufthansa in the frame of ongoing efforts at DLH to redesign its intemal
eBusiness activities. Due to the close link between this study and the intemal eBusiness
framework at DLH the authors are convinced that the results of this study can be of practical
interest for other companies too that are in the process of activdy redesigning their intranet
2. The promise of corporate portals and the situation at Deutsche Lufthansa
The motivation behind corporate intranets is the idea that the intemal use of Intemet
technology can act as a business process transformer (Lindstém and Frank 2000). It is
hoped that through a calalytic effect of the used technology a whole company and its
respective value chains will potentially be optimized Furthennore, direct effects of the use
of portals like diminishing ooss for intemal infomation publishing and an increased
corporate information dispersal (Rice 1996, Thyfault and Marx 1996, Lindstim and Frank
2000) add to this optimization. The devdopment of corporate intranets can be divided into
three different evolutionary stages: (a) In the beginning, the main goal is only to “be online”
by having an information on the Intranet without paying too much atfention to content and
functionality. (b) In the next stage (intemal) customer-oriented savices ae offaed on
portals, which triggers first process redesigning efforts (c) The last stage represents the
fundamental change of the company with respect to its intemal stucue by becoming nm
e@ransformed company (Lindsttim and Frank 2000).
Although corporate portals are described to “become the new melaphor for desktop
computing in business’ (Konicki 2000) and the Meta Group predicts that nearly 85 percent
of businesses will have portals by 2003 (Cruz 2000), most companies are nowadays between
stage one and two. Only few companies have fully reached the last stage of this three step
business transformation (Lindso6m and Frank 2000). The three stages of business
transfonmation described above reflect the typical path hig corporations have to go
nowadays. Despite the darity of this statement especially for hig companies thee is no
detailed description of how they have to change their way of doing business if they ae
considering implementing a corporate portal.
Generally speaking, the use of an enterprise portal is diverse and depends on the fidds of
activity: thee exist Business to-Business (B2B), Businessto-Consumer (B2C) and
Business- to- Employee (B2E) portals (Trowbridge 2000) dependent on the main target group
of the porta. Fiom a fimctional point of view common derents of ocomorde portals
function as integration pldfomms and contain knowledge management, oollaboration and
communication modules (Dellor 2000). Furthermore, they typically provide a search engine,
an index for structured and unstructured data, links for both intemal and extemal Wéb Sites
and infomation sources (Borck 2000, Quellette 1999). But to evolve their full capability,
corporate portals must also comprise concepts for security, personalization, and application
integration. The latter ranges from simple web applications to ERP (entemprise resource
planning) or legacy mainframe systems (Borck 2000). Last but not least every corporate
portal also requires a standardized, role based editorial process environment, to supply the
portal with fresh content and an infonvetion structure that is able to organize and host all
integrated systems in a user friendly and meaningful way. Without the latter every portal is
in danger of providing only infonretion fog to its usas. With all these feahes, Dellor sees
the opportunity for portal to offer “a rich and complex shared information work space” for
employees (Dellor 2000, p. 93).
Deutsche Lufthansa already uses Business-to-Business (B2B) and Business to- Customer
Portals (B2C) (“Infoflyway” and “Corporaie Hyway”). In the fidd of Business-to-
Employees (B2E), Lufthansa was in 2001 in the transitional phase from stage one to two. To
advance further to stage two and three in a structured fashion, DLH launched an employee
portal project called “eBase’! in January 2001. The main goals of this project are: (a) to
provide a new infrastucure for DLHs intemal eBusiness plafform (eg. Public Key
platform), (b) to design and implerent a new oorporate wide infonmetion stucuure, (c) to
' The title “eBase” is a combination of the two words ,, eBusiness” and ,, home base".
increase the connectivity of the enployees and (d) to ensue “skill and will” on the
employees side. Through these activities it was hoped to tigger the intemal ¢Transfonmation
at DLH, 90 that oost reduction, productivity and innovation potentials could be realized for
DLHs intemal processes.
For a company with a decentalized omganization as DLH especidly the aspect of
infonmation structure and quality is of crucial importanoe, as it must ensure the employees
ability to locate all infommtion he/she needs. As mentioned above the so-called business
transformation of an enterprise through a corporate portal or intranet is driven by a diverse
set of portal functions. From a process point of view, a portal or intranet environment is
supported by two main processes: the editorial process and the process of application
integration. Both processes have been considered and analyzed at DLH in the frame of
eBase. Nevattheless, in the following we will focus on the editorial process environment, as
it establishes the natural framework for the process of application integration’.
3. The editorial process at Deutsche Lufthansa
The editorial process developed by Deutsche Lufthansa aims to show a standardized way to
produce content The meaning of the word content in this context is very diverse as the
definition varies from the simple conection of spdling faults of already existing content to
the production of new and complex content with pictues etc.. Hence the different content
types should be differentiated according to the following aspects: conplexity, piotity,
scope, cyde time for updating and onnfidentidity. Hom this divesity logically rest
different quality standards for each type of content. But although there ae different types of
content at DLH, the editorial process stays the same for all types of content Before we will
jpresent these steps, the different roles involved in the editorial jprocess will be described.
In the editorial process at DLH, four different roles are defined, which ae shown with their
list of activities in table 1. The roles ae author, subject specialist, editor and chief editor
The author is an expat, who produces the specific or technical oontent after having
discussed a request for publication with he executive of a department The subject specialist
is responsible for stylistic comections and helps with the definition of citeria for meta data
The editor is responsible for the look and feel, the comectness of meta data and the strategic
fit of the content. Furthermore, he is usually responsible for the structure of a home page of
a department Chief editors exist in evey DLH business unit and for each global topic in the
DLH portal information structure’. The chief editor manages the complde editorial process
environment of a business unit or a global topic. Furthermore, he is responsible for the final
release of content (delegation is permitted). The chief editor also advises the editors and
solves editorial conflicts. With respect to the above described role modd for the editorid
yprocess at DLH, it should be mentioned that in general one person can fill out several roles.
? Information about the standardized process for application integration into the DLH employee portal eBase
galled , Track“ can be obtained from the authors on request.
* The 1st level navigation of the DLH employee portal comprises besides links to the business units of DLH
also links to so called global topics (e.g. News, Work and Life, Knowledge or Marketplace).
Role
Activities
Author’
Coordinates request for publication with]
executive from competent
Defines taryet goup and maintenance of|
content (together with editor)
Produces and updates content
Proposes a link and navigation structure
Creates a teaser and headline (together
with editor)
Determines date for publishing and period
of validity (together with editor)
Saat pea
Checks the oontent for spelling, style]
expression, logical order, consistency and]
velidness
Commumicates all necessary changes to)
author
Checks the link proposition and puts it in}
practice afterwards
Defines meta data and provides oontert|
with meta data
Veiifies meta deta and content on the
basis of a check list (together with editor)
Evaluates look and fed of content
(together with editor)
Checks navigation structure (together with]
editor)
Identifies a request for publication and an]
Evaluates request for publication and
assigns content to a business unit or global
topic area
Checks the link proposition and puts it in|
practice afterwards
Veiifies meta deta and content on the
basis of a check list (together with subject
specialist)
Evaluates look and fed of content
(together with subject specialist)
Checks navigation structure (together with]
subject specialist)
Creates a teaser and headline (together|
with author)
Determines date for publishing and period
of validity (together with author)
Takes care for archiving of content
Chief editor ¢ Verifies the completeness of content on]
the basis of a check list
¢ Releases content and informs author and]
all parsons involved of released content
Table 1: Roles in the editorial process
The editorial process itsdf consists of five different steps: concept phase, production,
completion, publishing and achiving. These steps will be descibed in more detail in the
following and are shownas a graph in figure 2.
In the concept phase, an editor identifies a request for publication and an author, who
coordinates the request for publication with an executive from the ive department.
Author and editor define a target group and the maintenance procedure for the content In
the next step the editor evaluates the request for publication and assigns the content to a
business unit or a global topic. Finally, the editor informs the author of the assignment, so
that at the end of the concept phase, there is an evaluated request for publication with an
assignment for a business unit ora global topic.
The production phase starts with the production of new and the updating of old content by
the author. Separately, the author proposes a link and a navigation suche. Next, the
subject specialist checks the content for spelling, style, expression, logical order, consistency
and validness. After having checked the content, the subject specialist commumicates all
necessary changes to the author The result of this phase is produced content, which has
been quality checked for editorial aspects.
After having checked the proposition for a link struchre, the editor or subject specialist
starts the completion phase by putting the link structure into place Furthermore, the subject
jalist defines meta data and provides the content with meta data In cooperation with the
editor, the subject specialist veifies meta data and content on the besis of a check list,
evaluates the look and fed of the content and checks the navigation structure Finally, the
editor in cooperation with the author creates a teaser and a headline and determines a date
for publication and a period of validity for the content Filly quality conbolled content,
which is provided with meta data is the goal of the compldion phase At this stage, the
content is ready to be released with all components.
In the publishing phase, the chief editor verifies the completeness of the content on the basis
of a checklist Afterwards, the chief editor releases the content and informs the author and
all persons involved about this fact The final result of the editorial process is content
published on the eBase live site.
> aw aw aw aa
Author, Author, Subject
Editor Subject specialist, ae
specialist Editor Editor
Figure 2: The standardized editorial process at Deutsche Lufthansa
After the period of validity of the content has passed, the decision has to be taken if the
content has to be updated or archived. In case of archiving the editor takes care that the
content will be removed from the live site without q@using link enors. Finally he stores the
content in a structured way in the eBase archive.
For the development of a standardized editorial process at DLH different motivations can be
stated:
¢ the need to define dear responsibilities and rights in the orgmizdiond units working
with the editorial process,
¢ the wishto establish deer organizational guidelines for content publishing,
¢ the goal to achieve a standardization of the existing historically emerged process modes
by using a newly developed standard editorial process, so that for example the use of a
centralized Content Management System (CMS) is eased (number of process modes
that need to be implemented),
¢ more transparency about interfaces and
¢ anincreesein efficiency and effectiveness in content production.
Besides these structural aspects of the standardized editorial process, for a successful
process Management, as stated above, a dear understanding of the hidden dynamics in this
process structure is important. Therefore, in a next step the structure of the editorial process
at DLH is modeled and analyzed using a system dynamics model that will be presented in
the next section Such a model is of high interest as it enables an analysis of both, the
qualitative and the quantitative dynamics of the process.
4. The editorial process in System Dynamics notation
The physical structure of the DLH editorial process can be mapped by a stock and flow
diagram consisting of four different levds: concept pages, HTML pages, (published) content
and old content. The first phase of the editorial process - the concept phase - is illustrated by
the rate production of concept pages, which leads to the leva concept pages. The production
hase is represented by the rate production of HTML pages going fiom the level concept
pages to the levd HTML pages The completion and publishing phase are combined to one
tate called quality management to undedine the importance of the quality management
yprocess in these two steps of the editorial process. As a result of these two phases, the
content exists on a live site in eBase After a certain petiod of time, the content gets old,
which is illustrated by the rate content aging, which leads to the lev old content Content
aging is modded in this framework as a first order delay, which means that on average all
content that is older than a threshold age T, is transferred to the stock old content. This
threshold is in general content type dependent In this study the assumption is made that T[
is constant for all content types. T, must be interpreted as an average threshold age.
Old content can either be updated or flowing out of the system and by that gets archived.
Hence, the achiving phase is portrayed by the rate content flowing out. For reasons of
simplicity this rate is also modded as a first order dday*. The rates quality management,
content aging and content flowing out are assumed to be time-dependent The variable
“The archiving rate could be modeled as a work related rate (e.g. production of concept pages) too. This
approach is not used here, as it is assumed that archiving will mainly be achieved through an automatic process
only requiring small working efforts.
quality management is influenced by the aniliay time for quality management which
reflects the time needed to go through the quality management procedures in the editorial
process. The rate content aging depends on the lifetime of the content, which is represented
by the aniliay time to age Finally, the axiliay time to flow out influences the rafe content
flowing out and therefore the amount of old content flowing out of the system.
The decision to produce concept pages and HTML pages are dependent on the following
variables: productivity, workforce and hours worked. As the production of concept pages
needs a different amount of time, people and productivity as the production of HTML pages,
figure 3 shows such a differentiation.
HOURS WORKED
AveRAGE —_-FORUPDATING
PRODUCTIVITY )vonesonce
UPDATING
Updating” UPDATING
>| Sage ge fHTML Pages| ge) Content OM
production of production of, quality Sk, Contest content
ong es a page a ae content aging a
AVERAGE AVERAGE TIME FOR QUALITY \
PRODUCTIVITY PRODUCTIVITY MANAGEMENT TWMETOAGE IME IO.
PRODUCTION OF PRODUCTION
CONCEPT OF HTML
PAGES WORKFORCE
pMlORKFORCE PRODUCTION OF
CONCEPT PAGES HIME PAGES:
HOURS WORKED FOR
HOURS WORKED FOR PRODUCTION OF HTML
PRODUCTION OF
CONCEPT PAGES
Figure 3: The system dynamics model mapping the editorial process at DLH.
For a more deailed dynamical arelysis of the existing process structure, it was decided to
reduce the given process mode to the genetic suche of an aging chain, a we ae
convinced that this simplified modd can capume the main dynamics of the more complex
one. With this attitude we follow Sterman who states: “a broad mode boundary that
captures important feedbacks is more important than a lot of detail in the specifications of
individual components” (Sterman, 2000). However the discussion about the model size still
takes place in the system dynamics commumity. Hence, Keating names the assessment of
mode soope including time horizon and system boundary, one key aspect in the modd
design phase (Keating, 1998).
By aiming for a rdaivey simple stuchre, the different steps to produce content were
condensed to one rate, which is simply called production The latter depends on the
variables average productivity production, hours worked production and workforce
production. Despite this condensed view on the process, figure 4 shows that the basic
strucue stays the same with respect to the aging process from content to old content. The
decision on how many pages are updated is influenced by three oonstants: average
productivity updating, hours worked updating and workforce updating. In order to represent
a vey simplistic (“policy-free”) decision making process, updating only depends on these
exogenous factors, not on old content (non- negativity of old content is assured).
HOURS
AVERAGE WORKED
PRODUCTIVITY UPDATING
UPDATING WORKFORCE
UPDATING
=F content Old
production r Content content
content aging XQ _ towing out
AVERAGE Bi
PRODUCTIVITY TIME TO FLOW
PRODUCTION THE TO.AGE Gur
HOURS WORKED
PRODUCTION
WORKFORCE
PRODUCTION
Figure 4: Basic modd for the simplified editorial process.
AVERAGE
PRODUCTIVITY HOURS |
UPDATING
UPDATING
\ (a et
-_——_ UPDATING
Content + updating
So
) +
Loa content agin +QY
a ae 2 Old Content
AVERAGE
PRODUCTIVITY = Re
PRODUCTION © i
content
flowing out
HOURS WORKED TIME TO AGE 5
PRODUCTION \
WORKFORCE
PRODUCTION TIME TO-FLOW
Figure 5: Causal loop diagram of the simplified editorial process.
The resulting condensed structure of the simplified editorial process is the stucure of a
second order material dday.
Considering the dynamic stucure of the process, figwe 5 illustates that the snmple
strucume of the editorial process consists of two negative feedback loops. The first one
comprises the following dynamics: an increese of content leads to a higher rae of content
aging, which reduces thereupon the leva of content again The same feedback strucure
exists for old content, which is flowing out of the system The more old content exists in the
system, the higher the rate of content flowing out, which decreases the level of old content.
A bese mun of the simple editorial process model shows the typical goal seeking behavior of
negative feedback loops. Put into other words, the amount for content and old content
teaches an equilibrium state. Hence, an analytical analysis of the equilibrium slale of the
system appears to be of high interest and will be presented in the following section.
Content
Time (Day)
Content : base run no updating ————— pages
Old Content
10
Time (Day)
Old Content: base nm no updating ——— pages
Figure 6; Dynamic behavior of content and old content in the basic model.
5. Analytical analysis of the system equilibrium state
Before we can start with an equilibrium analysis of the editorial system, the equations for the
system variables need to be identified and presented. First the equations for the five different
tates and second the equations for the two levels will be shown.
The five system rates can be expressed as follows:
¢ The production rate (P), which measures how many pages, can be produced per day. It is
assumed to be the product of the average productivity for production, hours worked for
production per day and workforce for production. The equation can be written as
follows:
P —-=AP, xh, xN, [pages day] @
AP, | =average productivity production [pages/ (hx person)]
h =hours worked for production per day [h/day]
N, =workforve for production [person]
¢ Content aging rate (CA), which equals the amount of pages that age per day. The rate
equals the ratio between content and the time to age and can be described as a first-order
metenial delay:
CA — =C/TA [pages/ day] (2)
with
Cc =Content ornumber of fresh pages [pages]
TA =time to age [day]
Fora short explanation on how the variable content aging can be modified for practical
needs see the appendix.
¢ The updating rate (U), which describes how many pages are updated each day: U is the
product of the average productivity for updating, hours worked for updating per day and
workforce for updating. This gives:
U =AP,x h, x N,[pages/ day] (6))
with
AP, — =average productivity updating [pages/ (hx pason)]
h, =hours worked for updating per day [h/day]
Ni — =workforve for updating [persons]
¢ Content flowing out rate (CO) indicates the amount of old content, which is flowing out
of the system per day. As stated above, the rate reflects simultaneously the degree of
archiving and can be consequently thought of an achiving rate. The activity of archiving
would claim workforce too. For reasons of simplification, the rate content flowing out is
defined in the same way as content aging. Henoe the equation is similar
CO =OC/TO [pages/ day] 4
with
OC =Old content or number of outdated pages [pages]
TO =timeto flow out [days
Fora short explanation how the variable content flowing out can be modified for
practical needs see the appendix.
As displayed in figure 4 the structure of the editorial process contains two levels that can be
described by the following equations:
Fresh oontent: the amount of content increases through the inflows of production and
updating rate and decreases through the outflow of content aging, which gives:
dC/ dt =P +U- CA i)
Using equations (1), (2) and (3), the differential equation for C looks as follows:
dC/ dt =AP,xh,xN, +AP, xh, xN,- C/TA ©
Old content: the amount of old content increases through the inflow of content aging and
decreeses through the outflow of updating and content flowing out:
dOC/ dt=CA - U- CO 1)
Inserting equations (2), (3) and (4), the differential equation for OC can be written as:
dOC/ dt =C/TA - AP, xh, x N,- OC/TO (8)
Based on these equations for the levels and rales of the system, an equilibrium analysis can
be performed. The aim of such an amalysis is to deduce analytical expressions for the
raationship of the stocks and flows under equilibium conditions. The equilibrium analysis
can be peformed by using equations (6) and (8). In dynamical equilibrium conditions the
system state is characterized by dC/ dt = 0 and dOC/ dt =0. Under this assumption the
following relationship for C results:
AP, xh, x Np +AP, xh, xN,- CTA =0
C=TA (AP, xh, xN, +AP, xh,xN) Q)
C=TAx(P+U) (10)
In equilibium, the amount of fresh pages depends on the sum of the production and
updating rate multiplied with the average aging time for content, which is time to age.
The equation for old content can be derived in the similar way resulting im
C/TA-AP,xhxN,-OC/TO =0
OC =TO (AP, xh, xN,) (1)
OC =TOxP (12)
The last equation shows the interesting insight that the old oonent in eqilibim ally
depends on the production rate multiplied with the average delay time for old content, which
is time to flow out. This is remarkable, because the production rate does not represent a
direct inflow to the levd variable old content. Furthamore, the direct in- and outflows to the
variable - content aging, updating and content flowing out- have no influence at all.
For a further amelysis in a next step, thiee ratios rdating the two system levds in diffeent
‘ways were developed:
Ratio;:
Ration:
Ratio3:
Ratio of content to old content, which relates the amount of fiesh pages to
already outdated pages. This ratio is called Ratio, in the following By using
equations (9) to (12), Ratio, can bewitten as follows:
Rao, =C/OC
= TA (AP, xh, xN, +AP, xh.x Ny)
TO (AP, xh, x N))
TAx(P+U)
= TOxP
Raiio; =TA/TOx (1 +U/P) (13)
Ratio of old content to content, which equals the reciprocal of Ratio1:
Ratio. =OC/C
= TO(AP, xh, xN)
TA (AP, xh, x N, +AP, xh, xN)
TOxP
= TAx(P+U)
1
Ralio, =TOTA x (14 UP) (14)
Ratio of old content to total content, which equals the amount of old content in
the system and is called Ratio; in the following. By using equations (9) to (12),
Ratio; can be written as follows:
Ratio; =OC/(C+0C)
TO(AP, xh, xN))
=TA (AP, x hy XN +AP, x hx N) +TO (APp x hy X Nb)
TOxP
TA x(P+U) 410xP
1
Ratios = TA/TOx (1 +U/P) +1 (15)
Interpretation of these results of the equilibrium analysis leads to three main insights that can
be formulated as follows:
¢ In cese of no updating which means U =9, the equation for total onnient is the
C+0C
=TA (AP, xh XN,) +TO (AP, xh) xN,)
=TAxP+TOxP
=Px (TA4TO)
This means that the stock of content in transit is the product of the inflows (roduction
rate Pand Updating Rate U50) and the two different average delay times time to age
and time to flow out). This rdationship is know as Little's Law which states that in
equilibrium the stock in transit is equal to the product of the total dday time in the aging
chain and the inflow rate into the chain (Sterman, 2000).
Tn case of no updating, the three ratios are only dependent on the two time variables time
to age (TA) and time to flow out (TO).
Ratio, =TO/TA, Ratio. =TA/TO, Ratio, =1/(TA/TO +1)
These time variables characterize the different types of content or the different quality
qiteia for aging of content eg. content pages indicating news typically have a lower
time to age than content with a longtem character Hence, it can be concluded that
without updating a change in the ratios can only be cased by a change in the time
variables, which means a change in the quality standard or the type of content
Tn cese of updating, the different ratios change their values dependent on U. For example
Tatioz, which indicates the ratio of old content to content, changes its value in tems of
TO/TA from 1 to 0, dependent on the degree of updating. This can be illustated in
figure 7.
c/0c oc/c
(Ta/to) (rojra)
1,2
3,8
4 We 1
3,4 75
0,8
0 05 1 15 2 25 3. UIP)
Figure7: Ratio; and Ratio in dependence of the updating effort.
On the x-axis, the updating effort is shown in tems of the production effort: On the first y-
axis on the let side one can see rafio; scaled by TA/TO. The auve shows that in
equilibrium an identical updating and production effort (U(P)=1) leads to a mumber of fresh
pages scaled by TA/TO which is twice as high as the number of outdated pages.
The y-axis on the right side shows ratio, scaled by TO/ TA. This means that for an updating
effort of 2 (updating effort is twice as high as the production effort) the ratio of old content
to content scaled by TO/TA is 1/3. Consequently, the graph shows two critical lines showing
vaio, and ratio, in tams of TO ad TA in equilibrium These anves may be helpful as
indicator for managers, who have to decide about their distribution of resources between
content production and content updating. Figure 7 may give them for the equilibrium state of
the system a first idea about the effects of a certain workforce distribution on the number of
outdated pages compared to the number of fresh pages and vice versa.
To further daborate the presented results the next paragraph will illstae difffeet
scenarios for the editorial process environment. Firtheamore, at the end the practical
televanoe of the lessons leamed from the introduced framework will be presented.
6. Different Scenarios for the editorial process
(1) Cortent production without updating
In this scenario, it is assumed that no updating is done which means that the updating rate
U=0. For the variables influencing the production rate P, workforce for production was
assumed to be 10 persons with 8 hours working per day and a productivity of 0.25. This
means that each person needs four hours to produce one page of new content, a value, which
has been determined after an intemal discussion with expats from Lufthansa Geman
Airlines. One should keep in mind that in these four hours not only the production of the
page but also the conceptualization takes place. The variable time to age was set to 7 days
and the auxiliary time to flow out to 30 days respectively. The latter means that old content
will be archived once a month.
In equilibrium this parameer setting leads to the following results for oontent and did
content:
AP, =0.25 TA =7
h =8 TO =30
N, =10
Cc =TA (AP, xh) xN, +APaxhxN) =140 pages
OC =AP,xhxN,xTO =600 pages
For the three different ratios one obtains the following results:
Ratio, =C/OC
=0,23
Ratio. =OC/C
=4,29
Ratios =OC/(C+0C)
=0,81
With a team of 10 people working 8 hous a cay with a productivity of 0.25, in the
equilibrium 140 fresh pages can be produced. On the other hand, 600 pages will be already
outdated, which leads to a percentage of old content of 81% in the system.
To lower the percentage of old content, the parameter of hours worked production was
enhanced to 12 hours. This leads to a content value in equilibrium of 210 pages, which
means that with a working day of 12 hours for 10 people working with a productivity of
0.25, a rise of 70 pages could be achieved. On the other hand, old content in equlibium
increased as well to avalue of 900 pages. The interesting consequence of this behavior is the
fact that the values of the three ratios remain constant.
Another strategy comprised the enlargement of the workforce: it was increased to 20 people,
which represents an increase of 100% of the workforce Although content in equilibium
Climbed up to 280 pages, the amount of old content rises in paralld to a value of 1200 pages.
Looking at the three ratios in this situation shows again the same results, which means the
ratio values emain constant. Figure 8 illustrates the ascent of content and old content in the
basic model for the above considered parameter changes.
Content
400
mo | a
aan
BS
A
100
Time (Day)
‘Content : base run no | pages
‘hours worked 12.no upceting poe
(Content : workfare 20 no undeting 2-9 -n anna n enna nnn nn nn paces
Old Content
2,000
1,000 os ae
0 bet
100
Time (Day)
(Old Contert:: bese nun no updeting §—§ pages.
(Old Cortert:: hours worked 12.n0 upceting rege:
(ld Content: workfarce 20 0 upitting» --------------——--————— =
Figure 8: Parameter change for hours worked and workforce compared to base rn.
Themain insights for Deutsche Lufthansa were fourfold:
¢ Although a change of single parameters like hours worked or workforce lead to a higher
leva of content, the growth is dways limited Therefore, any inceese in paamdas
cannot avoid that there are limits of gowth considering the production of content.
« An increese of single parametas did not only lead to a increese of content, but
automatically also to an increase in old content. As a consequence, the three ratios
rdding old content and content keep constant values.
* To change the value of the ratios and thereby the ratio between content and old content
in the system, a single parameter adjustment for P or U is not sufficient
¢ Changes in the ratios can be only achieved by changes in TA and TO which represart
structural changes in the system eg. changes of content types or different definition for
the aging threshold (which might equal a change in the quality policy).
The above results ae vey interesting as missing resouvss (i. e workforce) ae often
considered to be key suooess factors for content production, which is obviously not the case
when one is looking for a system with a low ratio3. People are obviously convinced that an
increase of resources would lead to a continuously growing amount of fresh content. The
focus on content production represents a mental model centered on the input of fresh pages
into the system This does not imply that updating is not an issue at all, but it shows dealy
the biased nature of the existing mental modds when it comes to production and updating
efforts. Henoe, it can be stated that in the mental models of the editorial pasonnd at DLH
the two following aspects wee not too dominant yet the possible gowth of content
production is limited and always counterbalanced by content aging, so that content needs to
be updated regulally. In this context, it is very probable that the editorial pasonnd at DLH
is not able to see Clearly the undelying dynamics of the editorial process producing the
observed goalseeking behavior for the content. Henoe, the simple system dynamics moda
may already help to change the awareness for the editorial process dynamics. Furthamore,
the importance of TA and TO and the accompanied system parameters appear to really new
insights, which means that an important extension of the mental modds of the involved
personnel seems desirable.
Therefore, DLH decided to enter into the process of discussing the implications of the
simple moda and its structural and dynamical insights with its editorial personnd.
Finally, it is interesting to note that this scenario maps vay nicdy the history of the first
intranets, which often ended in a situation of “senescence” and as a consequence of that of
limited usage.
(2) Content production with updating
For this scenario apart from content production, updating of content was considered. It was
assumed that the workforoe is strictly divided to one of the two operations. This means that
one person is unable to fulfill both activities at the same time Hence, as the workforve stays
at 10 persons, it was assumed that 5 persons are responsible for content production and 5
persons are doing the updating. Furthemore, the working hours for production and updating
are 8 hours per day in the base run. Besides, it was assumed that the average productivity for
updating is twice as high as the average productivity for production. Consequently, the
value for AP, is 0.5 whereas the value for AP, stays at 0.25. The variable time to age was set
to 7 days and time to flowout to 30 days respectively.
In equilibrium this parameter setting created the following results for content, old content
and the three ratios:
AP, =0.25 AP, =0.5 TA =7
h = h =8 To =30
Np = Na =5
Ratio, =OC/(C40C)
=0,59
Hence, with a team consisting of 10 people working 8 hours a day, half of them on content
production and the other half on content updating, 210 fresh pages can be produced in the
equilibrium by assuming an AP, of 0.25 and an AP, of 0.5. As the number of outdated pages
lies at 300 pages, the amount of old content in the system can be reduced to 59%.
Figure 9 illustrates a change by showing the dynamic behavior of content and old content in
a comparison to the no updating scenario. It can easily be seen that because of the updating
activities, more new content can be produced while the amount of outdated pages decreases
compared to the no updating scenario.
Content
”
200 U L444
A
100
Timea)
(Content : hase run updating: pees
Content : hase run no updating pages
Old Content
icon) | | I }
300 |
100
Time (Day)
Old Content : base rn updating pages
Old Content : base run no updating: paRS
Figure 9: Dynamic behavior of content and old content in scenario no updating and updating
In this context it is also interesting to note that the shape of the graph of content in the
updating scenario is different from the scemario without updating. This is due to the
modding of the updating rate, which is done by the use of a MIN Function in this scenatio.
Thereby the minimum of the values for old content and the linear updating rate - depending
on workforce for updating, hours worked for updating and average productivity for updating
- are considered. Consequently, in the first period with a very low levd of old content (much
smaller than the linear updating rate), all old content gets updated and the available updating
force might not be used 100%. Hence the updating rate is equivalent to the amount of
outdated pages. As soon as the amount of old content summounts the value of the defined
linear rate of updating, the linear rate of updating characterizes the intensity of updating.
Intapretation of this soenario leads in addition to the before mentioned results to the
following additional insights:
¢ With the inplerentation of an updating adivity, the rdaiive amount of old oontent in
the system can be lowered from 81% to 59%.
¢ Although the value of the system ratios changed, the overall system behavior i. e the
goal seeking behavior stays the same Any increese in production or updating activities
cannot avoid that the growth is finite.
In this soenario, a second activity wes introduced. Hence, the workforce must be divided
into the two activities for content production and updating. In this context, the question
arises how to distribute the scarce resource workforce on the two activities to obtain the best
results. Figure 10 shows the influence of workforce for production on the amount of fresh
pages. The graph shows that an increase in the production workforce leads to a decrease of
fresh content pages in the equilibrium This is a result which is suprising at the first moment
as one would have guessed that one of the first measures to increase the amount of fresh
pages is to augment the workforce responsible for content production. At a second glance, it
soon gets obvious that due to the importance of updating, a single focus on oonient
production does not lead to satisfying results. Though a concentration on content production
leads to a temporary effect of more content But as the system structure causes that an
increase of content leads to more content aging and therefore more old content, this effect is
quickly counterbalanced. Without enough people for updating in equilibium, the number of
outdated pages cannot be reduced which means that the amount of fresh oontent will
decrease more and more.
i
o 123 4 5 6 7 8 9 10 Np
Figure 10: Impact of workforce production [N,] on fresh content [C]
Figure 11 undedines this insight by showing the effect of a change in the production
workforce on ratio; that indicates the percentage of old content in the system. It can be seen
that the more people are involved into content production, the higher the percentage of old
content gets in the system, as stated already above.
OC/C+0C)
in%
100%
80% >
60% —* 0C/(C+0C)
4%
20%
0% >
012 3 4 5 6 7 8 9 10 Np
Figure 11: Impact of workforce production [N,] on ratio3 [(OC/ C+OC)]
Tt can be concluded that both graphs are of great help to illustrate the effects of a shift of the
workforce distribution between production and updating. Therefore, a manager may take
these two graphs as an orientation for his/ her decisions about the workforce distribution.
Apart from the discussion about content production and content updating, the aspect of
content archiving seams to be important, too. Therefore, the next soenarios will discuss the
impact of archiving on the system behavior.
(3) Content production with updating and without achiving
For this scenario, the parameter setting of the updating soenario wes kept except for the
variable time to flow out. As stated above in case of no archiving the value for time to flow
out will be rather high to illustrate that old content stays a long time in the system before
flowing out of the system Therefore the overall parameter setting in this scenario can be
givenas follows:
AP, =0.25 AP, =05 TA =7
h = h = TO =100
Np = Na =
Cc =210 pages
OC =1.000 pages
Ratio, =C/OC
=0,21
Ratio, =OC/C
=4,76
Ratios =OC/(C+0C)
=0,83
The value of fresh content in equilibrium stays the same as in the updating scenario because
it is not influenced by the variable time to flow out, which is the decisive factor in this
scenario. Figure 12 illustrates that the graphs for content are the same for both scenarios. On
the other hand, the amount of old content rises up to 1.000 pages, resulting in a ratio of old
content to content of 4,76 which equals on the amount of old content in the system of 83%.
Based on these results for this scenario two main insights can be formulated:
¢ Compared with the second scenario (content production with updating) the “archiving”
scenario leads to an increase of 24% of old content in the system Furthemore, the
percentage of 83% is nearly the same as in scoenatio 1 (content production without
updating) with 81%.
¢ The negetive effect of the missing achiving activity counterbalances the positive effect
of the updating activity.
Content
400
200 +7
of
10
Time Day)
Content : base run no archiving pages
(Content : base run updating pages:
(Content: bese nino upteling | ~~~~-~~~~~~-~-~-=-=-=-=-===~ pages
Old Content
cet} [Lt te
[> [4
Lt
300 L-17|
La" | “e
eg ee
ay
0 ee
100
‘Time (Day)
(Old Content : base run no archiving § pages
Old Content : bese run updating ‘pages
(Old Content: basenmnp upthling | -~==—----======--===-——--- pages
Figure 12: Dynamic behavior of content and old content in scenario no updating, updating
and no archiving
To gain further insights the next somatio will foas on a smulaion oonaining a high
achiving activity.
(4) Content production with updating and archiving
As this soenatio focuses on the aspect of archiving, the variable that modds the achiving
activity, time to flow out, is set to a low leva. This means that the time that the old content
stays in the system is vey short which means that it will be achived quiddy. The oveall
‘parameter setting in this scenario is the following:
AP, =0.25 APR, =0.5 TA =7
h = h =8 =2
N = N =5
Cc =210 pages
OC =20 pages
Ratio, =C/OC
=105
Ratio, =0,095
Ratio; =OC/(C+0C)
=0,087
As in the scenario above, the value of content in equilibrium stays the same because it is not
influenced by the variable time to flow out, which is the decisive factor in this scenario.
Figure 13 illustrates that the value for content in equiliium is the same for the thee
scenarios updating, archiving and no archiving. On the other hand, the amount of old content
can be lowered to 20 pages, resulting in a ratio of old content to content of 0,095 which
equals a percentage of old content in the system of less than 10%.
ae
Time(Day)
Old Content
Tine (Day)
Figure 13: Dynamic behavior of content and old content in scenario no updating, updating,
no achiving and achiving
The main insight of the fourth soenario can be summarized as follows:
¢ The decision for or against achiving has a strong impact on the amount of old content in
the system. While updating may lead to a deciease of 22% oconsideting iatio; (old
content compared to total content), a consideration of archiving can decrease the ratio
even more. So the percentage is reduced to less than 10%. This means that in scenariol
without any updating a change in the achiving paameer could lead to temendous
lowaing of ratio 3.
Before giving some last oonduding remarks about the oveall insights gained by the
different scenarios, the values of the main indicators for the different somarios will be
summarized in a table Table 2 compares the results of the analytical approach with the
results obtained from different modd simulations. It is interesting to sve thet the rests
often differ in a significant way. Even with the snellest time sep available in the software
used for the research project (Vensim PLE wes used) the difference oould not be fully
reduced. This is due to the limited capabilities of the software, which means that with a
advanced version the differences could be neatly decreased for example using a different
integration method. Apart from the time step used in the modd, the question of time horizon
Plays an important role, too. As can be seen in figure 12, in the soenatio of no achiving the
system did not reach the equilibrium state in the presumed period of time. After a look on
the parameter setting of this scenario it gets obvious that with an assumed time period of 100
days and a time to flow out of equally 100 days, the moda does not reach an equilibium
Szenario 1: Szenario 2: Szenario 3: Szenaio 3:
Content! Content! Content Content]
Production} productionwith} productionwith} productionwith
without updating updating] = updating without updating and
tivi ivi
A V A Vv A V A V
Content (C)
[pages] 140 | 139,99} 210 | 209,99] 210 | 20999] 210 | 207,61
Old content (OC)
[pages] 600 | 572,13 300 | 277,09] 1.000 |} 543,85 20 19,77
Ratio;
=C/OC 0,23 0,24 0,7 0,76 0,21 0,39 10,5 10,5
[dimensionless]
Ratio,
=OC/C 4,29 4,09 1,43 132: 4,76 2,59 | 0,095 | 0,095
[dimensionless]
Ratio;
=OC/(C +OC) | 81% 80% 59% 57% 83% T2% 8.7% | 87%
[percent]
Table 2: Main indicators of the equilibrium analysis on the base of the four different
scenarios (A - Analytical approach, V - Vensim sinulation).
Given the above results and observations it is interesting to compare the stength add
weaknesses of the two used methods, which will be shortly jaresented in the following.
The analytical solution in this study wes based on an analysis of steady state equilibrium
This approach helps to derive precise values for the model parameters in steady state, on the
bese of analytical expressions relating the system parameters with the equilibrium state of
the system This kind of information would not have been available by help of simulations.
On the other hand, the amelytical results only show the situation in one sngle moma,
which is in our case the steady state equilibium Finally, an analytical approach enables an
eesier interpretation of the importance of certain system parameters for the devdopment of
The use of simulations helps to overcome this weakness of the analytical approach With
simulations, the behavior through time is made transparent and short-tam and longtam
effects can be easily portrayed Firtheamore, a simulation mode is designed to be
transformable and through its examination through time, the dynamic nature of a process is
emphasized. However, as stated above, results obtained from a simulation may only be
understood as approximation to exact values for example for equilibrium conditions.
7, Summary and outlook
The presented paper gives an overview about the challenges of a portal implementation at a
ig company on the basis of a case study conducted at DLH. In this context the importance
of a contolled editorial process environment for a successful portal implementation is
emphasized.
By using a system dynamics model to portray the standardized editorial process devdoped
by DLH, it wes possible to simulate the dynamics of the editorial process. A combined
approach of a process simulation and an aralytical equilibium amlysis alowed pafonning
ascenario based system analysis, which provided several important insights:
¢ In equilibrium, in cese of no updating, total content equals the product of the inflows and
the different average ddlay time. This relationship is known as Litile’s Law.
¢ In cese of no updating, the three considered system ratios are only dependent on the two
time vatiables time to age (TA) and time to flow out (TO) in equilibium These time
variables reflect the different types of content or the different quality criteria Hence a
change in the ratios can only be caused by a change of the time variables, which means
for example a change in the quality standards or the types of content.
¢ Although content production can be intensified by employing more resources the
number of fresh pages in the system is always limited.
¢ Aging of content is a napwal charectetistic of evay qditorial process and therefore
een betsy a sear ae
¢ An achiving system should be implerented in exch atitorial process to avoid the
draining of the system by old content
General insights of this case study can be formulated as follows:
¢ A process is dynamic by definition and a particular vulnerability of process management
lies in theinahility to manage dynamic complexity
« The deveopment of a simple physical process model already helps to show how the
different activities of an editorial process affect each other Hence a decision about the
tight amount of production, updating and achiving may now be taken in a nore
informed way
A fonvel modd is more elicit thn a desciption with words. By building a fomal
modd, the unspoken assumptions are made explicit and compatibility of the existing
mental modds of the involved staff can be tested.
Although the parsons involved at DLH had a vague and more qualitatively oriented idea
about the different insights presented before a quantitative mapping undelined the
‘urgency and necessity fora deeper understanding of these insights.
The smple structue of the modd already tumed out to be valuable to pafom a
soenario-based analysis. Therefore, the mode already saved as a “wind tummd” for
organizational process expaimentation and may be of great vdue for fubre
expaiments, too.
For a policy design, first indicators have been devdoped which show the rdationship
between quality standards and resource allocation. These ratios and the ciitical paths
developed by relating the ratios with other system variables may serve as a basis fora
more detailed process management, which might also include an information network.
Apart froma necessary validation to underline the robustness of the simple modal,
unanswered questions exist with respect to the following aspects:
Which effect has a change of the dday structure on model behavior? In the appendix,
first ideas conceming this area of research will be presented.
Can the results on the basis of the simple model be transferred to the four stock model
jpresented in the beginning of the paper?
In which way is the physical network of the editorial process influenced by the
infonmation network? How should a policy design for the editorial process look like?
Future research will be focused on these topics and results will be presented in following
papers. Thus, a subsequent paper will deal especially with the dements of a successful
policy design for an editorial process environment inside of a major corporation.
Acknowledgement:
The authors would like to thank Silke Lehnhardt (the fonmer eBase project team leader) and
the eBase Project team for their continuous support for this research project.
Appendix
The analysis of content and old content was based on the assumption that there is only one
sort of content with a single rate for content aging and for content flowing out However in
reality there is not one typical content, but various sorts of content with diffeent quality
standards. As a consequence, the type of delay can be different, too. In the paper, for reasons
of simplicity, two separate first order delays have been assumed to illustrate the rate of
content aging and content flowing out Still, it is also possible to use a second, third or even
fourth order delay. To be able to fully map the delay structure of average content, an average
delay function consisting of the different delay times should be devedoped Figure 14 ties to
give a first impression on how in reality a distibution function of the different dday times
for different types of content could look like It shows the pulse response of a predefined
distibution function of different types of content The numba I to VI reflect the different
types of content contained in the considered bulk content.
% of Unit Pulse/
Time Period I st ait Iv Vv VI
>
Time (multiples of
average delay time)
Figure 14: Possible delay distribution curve fora system containing different types of
content with different average aging thresholds.
References
Borck, James R. 2000. Beyond the hype, what will an EIP do for you? InfoWorld. 06/2000:
49-52.
Cnz, Mike. 2000. Plugging into B2E Solutions. Computer Resdler News. 09/2000: 114-118
Dedlor, Brian. 2000. The corporate portal as information infrastructure: towards a framework
for portal design. Intemational Jounal of Management. 20(2000): 91-101.
Keating, Elizabeth, K. 1998. Everything Y ou Ever Wanted to Know about Developing a
System Dynamics Modd, But Were Afraid to Ask. Proceedings of the 1998 Intemational
System Dynamics Conference.
Konicki, Steve. 2000. Powerful portals. Information week. Issue 784: 44-66
Lindstm, Martin and Tim, Frank Andersen. 2000. Brand building on the Intemet. Kogan
Page. London
Morecroft, John D. W. 1982. A critical review of diagramming tools for conceptualizing
feedback system models. Dynamica (8): 20-29.
Quellette, Tim. 1999, Opening your own portal. Computerworld. 33 (1999): 76-79.
Rice Valetie. 1996. Building the case for your intranet. PC Week. 13 (1996): 53-54.
Stenman, John D. 2000. Business Dynamics. MoGraw- Hill. London/ New Y ork.
Thyfault, Mary E. and Wendy Marx. 1996. The intranet rolls in. InformationWeek. Issue
564:15-17.
Trowbridge, Dave. 2000. EIP - more profitable for integrators than users. Computer
Technology Review. 05/2000: 20-21.
Wanen, Kim. 2002. Competitive Strategy Dynamics, John Wiley & Sons Ltd.
Back to the Top
