s [6]. Recently, workflow management sys-dministrative environment and can also bemponents, that enhance existing applications well as infrastructure components (such ass management functionality [31].

ment systems can increase the process effi-uch as 150% [17], they do not necessarilytion. Since the introduction and deployment

on system architecture is very often a com-vor, it can be observed, that once this kind oflly deployed, many companies resist the urgetem (an effect that can be observed at compa-as well).

workflow projects is only one cause for theards change management. An even more

ency of those relationships, that describe then the organizational, technical, or process can be attributed to the lack of an integrated

and presentation of performance indicators,a workflow-enabled organization and giveadjust in order to increase the organizational

of a workflow application, the workflowbout the different state-changes on the pro-corded in the so-called audit trail, which is

tructure (cf e. g. [19]). Originally designed asng purposes, the audit trail provides informa-ness processes at the operational level.

trail information could enhance traditionalagement information systems by presentingrformance as well as drill-down capabilitiesto the business data connected with these

ting components of many workflow manage-ed evaluation functionality. Especially the trail data with data warehousing technologyflow vendors, but left to extensions by the

Process-driven Management Information Systems -Combining Data Warehouses and Workflow Technology

Michael zur MuehlenDepartment of Information Systems

University of Muensterismizu@wi.uni-muenster.de

AbstractThe use of workflow technology promises efficiency gains through the auto-mation of manual routing, coordination and work distribution tasks. Duringthe execution of workflows, state-changes of the workflow engine arerecorded in a log file or database, the so-called audit trail. Combined withbusiness object data, the audit trail provides exact and timely informationabout the operational behavior of an enterprise. In this paper we discuss thedesign on data warehouse applications that take advantage of workflowtechnology as an information source. We outline evaluation opportunitiesgenerated by the use of audit trail data and point out potential pitfalls withregard to data consistency and integrity.

1 IntroductionThe use of workflow management systems improves the efficiency of busi-ness processes through the automated coordination of the data and resourcesneeded for the execution of the single activities. Workflow management sys-tems rely on a formal representation of the process logic that is designed as aworkflow model during the development phase of a workflow application(also called build-time). During the execution phase of the workflow appli-cation (also called run-time), the workflow engine derives workflowinstances from the generic workflow model and notifies workflow partici-pants about pending activities through their work-lists (see e. g. [14]).

Workflow management systems have found widespread acceptance sincethe advent of this technology in the late 1980s. The Association for Informa-tion and Image Management (AIIM) estimates the worldwide revenue forworkflow technologies to grow from $4.3bn in 2000 to $8.3bn in 2003 at acompound annual growth rate of 31% [6]. Especially in conjunction withdocument management technology, workflow systems are perceived as theenablers of office productivity gains through the elimination of manual rout-

ing and work distribution tasktems have spread beyond the afound as embedded software copackages (e. g. ERP systems) aapplication servers) with proces

Although workflow manageciency of an enterprise by as mlead to a more flexible organizaof a workflow-based informatiplex and time-consuming endeaarchitecture has been successfuto apply changes to the new sysnies introducing ERP packages

However, the complexity ofreluctant attitude observed towsevere cause is the non-transpareffects of workflow changes olevel. This missing transparencyinfrastructure for the gatheringthat describe the behavior of advice on which parameters to efficiency.

During the run-time phaseengine generates information acess and activity level that is reeither file-based or a database sa technical protocol for debuggition about the execution of busi

In theory, the use of auditenterprise controlling and mankey indicators about process pefrom single process instances instances. In practice, the reporment systems offer only limitcombination of workflow auditis not supported by most workuser.

cause-effect relationship between workflows that affect the performance of a workfloworkflow model itself, but also to the invokedhe organizational surroundings of the work-e, the organizational signature power of cer-is increased, these participants couldased number of cases, lowering the numberry and thus reducing the workload of certainven though this measure has no direct effectorkflow management system, its effects are

ss performance (e. g. lower average through-

uction of a workflow management system isar to the development of a complex applica-ing and controlling are cyclic activities thatystem and run in parallel to system adminis-igure 1 shows a procedure model for work-

nd deployment (cf. [33]). The organizationalclic auditing of the workflow application. Itsical loop through incremental changes to the, a continuous improvement process can be

risks associated with a fundamental redesignations.

r Workflow Development Projects

This paper aims at providing a framework for the design of data ware-house applications that are specifically targeted at the integration of work-flow audit trail data. In the following section we develop a taxonomy ofdifferent analysis purposes for workflow audit trail data. We distinguishbetween workflow monitoring and controlling, and show, how both conceptsrelate to the workflow life cycle. In section 3 we focus on the design issuesfor a process information system based on workflow audit trail data. Follow-ing a phases model that spans data extraction, transformation, evaluation andpresentation, we point out benefits and potential problems associated withthe use of audit trail data for data warehouse applications. In section 4 weshow the benefits of workflow-driven process information systems using acase study from an insurance company. A discussion of related work isgiven in section 5. The final section of this paper provides a brief summaryand outlook.

2 A Taxonomy for Workflow Monitoring and Controlling

2.1 Monitoring and Controlling in the Workflow Life CycleCompanies face the need to quickly adapt to changing market conditions andcustomer wishes. Having a transparent overview about ongoing and histori-cal business processes gives companies the flexibility to adjust the treatmentof individual cases as well as the opportunity to make structural changes tobusiness processes. Workflow management systems separate applicationlogic from business process logic, enabling users to modify the business pro-cesses based on intelligence gathered from the use of a workflow applica-tion.

The life cycle of a workflow application has been described by severalauthors as a closed loop (see e. g. [7]), starting with the definition of thebusiness process to be implemented (1), followed by the transformation ofthe process model into a workflow model (2). The enactment of the work-flow model (run-time) makes up the third phase of the life-cycle (3), whereasthe ex-post analysis of executed workflows in the sense of process control-ling (4) generates information that is fed back into the process design phaseOther workflow cycles, such as the one described by HEILMANN (see. e. g.[9]), contain different phases, but they all depict the development, deploy-ment and analysis of a workflow application as a closed loop.

In practice, the closed loop assumption does not reflect the development anddeployment of workflow applications as it actually happens. This is due to

the fact that there is no isolateddesign and enactment. Changecan be applied not only to the wapplication systems as well as tflow application. If, for exampltain workflow participants autonomously approve an increof managerial approvals necessa(potentially limited) resources. Eon the workflow model or the wa noticeable change of the proceput time).

In our experience, the introda sequential process, very similtion system. Workflow monitorfollow the introduction of the stration and maintenance tasks. Fflow application development aloop on the right reflects the cyfeedback is applied in the technworkflow application. This waymaintained, without facing the of existing processes and applic

Figure 1: Procedure Model fo

essing and its impact on workflow technol-

ysis of workflow audit trail data can be current workflow instances, which is per-rocess managers (or process administrators),

workflow instances is mainly conducted byse two analysis purposes promise significantthem in detail in the following sections.

Evaluation Focus

Technical Business-oriented

Single Instance Single Instance

• exception handling (e. g. manual termina-

tion of workflow instances, manual re-assignment of orphan

activities)

• workflow monitoring (e. g. tracking of the

individual process state)

Multiple Instances Multiple Instances

• system monitoring (e. g. number of data-

base connections, response time)

• license monitoring

(e. g. utilized invoked applications)

• workforce manage-ment (e. g. scheduling

of resources based on predicted workload)

• workload manage-

ment (e. g. assign-ment of work items based on individual

capacity restrictions)

Single Instance Single Instance

• workflow debugging• workflow recovery (in

case of a system fail-ure)

• audit purposes (e. g. proof of fulfilment in

case of customer complaint)

Multiple Instances Multiple Instances

• license management

(e. g. utilized invoked applications)

• workflow controlling

(e. g. activity based costing, resource uti-lization etc.)

• workflow planning (e. g. development of new processes and

procedures)

low audit data evaluation

2.2 A Taxonomy for Workflow AnalysisThe analysis of workflow data can be based on either short-term or long-

term observation of the audit trail. This differentiation is typically used todistinguish between workflow monitoring (discussed in section 2.3), whichdescribes the analysis of active workflows, and workflow controlling (dis-cussed in section 2.4), which deals with the ex-post analysis of (potentiallyfinished) workflow instances. While short- and long-term observation pro-vide a suitable differentiation for the data to be analyzed, the purpose of theanalysis can serve as another point of differentiation. With regard to this,technical and business-oriented analysis goals can be distinguished.

From the end user point of view, another separation can be made betweenthe analysis of data for the purposes of an individual user as opposed to thepurposes of an organization (for reasons of clarity, this view has beenommitted from the table). These different perspectives determine both therelevant objects for the analysis (activities, resources, data, applications), aswell as their cumulation (e. g. the evaluation of single workflow instances asopposed to the evaluation of an aggregate of multiple workflow instances).

Table 1 shows a taxonomy for the analysis of workflow protocol data.For reasons of clarity, the table contains different evaluation opportunitiesfor single workflow instances in contrast to aggregate workflow instances.

The analysis of workflow audit trail data for technical purposes is per-formed mainly by system administrators and workflow designers, that aredebugging workflow models that are not executed correctly (since someworkflow management systems use relaxed formalisms for their modelinglanguages in order to accommodate less formal business process modelingtechniques, the resulting processes may contain potential deadlocks or simi-lar problems). Another purpose could be the monitoring of system loads andresponse times in order to determine the proper scale of the underlying hard-and software-components (mostly in terms of databases). In order to identifyworkflow instances that are “stuck” on the worklist of an absent user (andthe respective activity does not have a “timeout” attribute), the systemadministrator may query the audit trail for activities that are assigned toabsent users. But not only human users take advantage of the logged work-flow history. In case of a system failure many workflow systems use theaudit trail file to recover the system state to the last committed transaction,much like a database system uses a transaction log file (for a discussion of

dabase-related transaction procogy see e. g. [2]).

The business-oriented analdivided into the monitoring offormed by workflow users and pwhereas the controlling of pastenterprise controllers. Since thebusiness value, we will discuss

Data Scope Current Data (live)

Historic Data (ex-post)

Table 1: Taxonomy for workf

ses at a specific point in time, the number ofl as the current activities of these instances of staff requirements. Combined with thet systems to prioritize work items accordinge of the case (i. e. the idle time of a pendings companies maintain a consistent level of with high workloads.

transparency is illustrated by an example of. Due to a proposed (and later cancelled)e insurances were subject to additional taxa-n or after January 1st, 2000. This announce-e in life insurance applications during thet the life insurance department worked over-unt of applications, neglecting all other casess. As a result, the structure and age of the and customers complained about the longt back to them regarding their inquiries. Thised easily, if a workflow management systemand prioritized those, that were older than a

s desirable not to expose the detailed processg a certain workflow. An example is the pre-orkflow participants outside of the company

g. customers or suppliers. Figure 2 shows theess at an e-commerce web site. Even though more complex, only four steps are displayed differs from the actual process state in themodel of the underlying process state modelels shadow processes [20]). Figure 3 showslevel between the actual process state and aes in section B and the two activities in sec-ngle status B’ and C’, respectively, whereasr at the same level of granularity in the busi-

the business state model can only contain thean the process state model, since it is derivedsively. If context data such as the values of

2.3 Process MonitoringProcess monitoring deals with the analysis and overview of process

instances at run-time. Using monitoring information, workflow administra-tors and process managers can adjust the behavior of current workflowinstances and react to problems that arise during process enactment. Further-more, process monitoring is used to improve the responsiveness of an orga-nizations to customer inquiries. When the current state of a process instancecan be determined easily, questions such as “Who is handling the customerorder 4711?” can be answered in an efficient manner. For the individualworkflow participant, monitoring provides the ability to identify those col-leagues that worked on a particular case earlier, in case of open issues thatneed to be resolved. Figure 2 shows the monitoring of an online order whileit is being processed by the vendor.

Process monitoring beyond single process instances can be used to pre-dict staffing requirements. If the average processing times of activities allow

for a forecasting of open procesactive process instances as welallow the short-term predictionability of workflow managemento the case attributes and the agcase), process monitoring helpcycle times even during seasons

The importance of workloada German insurance companychange of the tax legislature, liftion, if the contract was signed oment led to a fourfold increassecond half of 1999. The staff atime to handle the unusual amothat were not new applicationremaining cases was unknowntime it took the insurance to gesituation could have been avoidhad kept track of all the cases certain threshold.

Under certain conditions it istructure to the party monitorinsentation of workflow data to wthe workflow is executed in, e. web display of an ordering procthe internal processes are muchto the user. This business stateway that it is a simplified state (NAEF et al call these state modan example for the abstraction business state. The four activitition C are combined into the sithe activities A, D, and E appeaness state model.

In the (simplified example, same number or fewer states thfrom the workflow states exclu

Figure 2: Monitoring of an E-Commerce Order

at have been accepted by a user but that have.

the ex-post analysis of workflow audit trailinstances are aggregated according to differ-mes. Workflow controlling is useful for theents in workflow enactment and the review

mplementations. In order to identify devia- audit trail data is often compared to target business process models. The goal of work-provement of future process enactment, thus than the results of workflow monitoring.

g Facility

certain process relevant variables is taken into account, the coarse states ofthe business state model may be refined into sub-states.

Besides the organizational process monitoring, workflow managementsystems typically provide facilities for technical monitoring, which dealswith parameters such as response times, system load and the like. Withregard to technical monitoring workflow management systems do not differfrom complex application systems that are managed through commercialpackages such as Tivoli [25] or Candle [4]. Figure 4 shows a screenshot ofthe technical monitoring facility of a commercial workflow managementsystem. Besides the current numbers of active users, processes and activitiesthe system also displays the number of pending processes and activities (i. e.

those processes and activities thnot been completely processed)

2.4 Process ControllingProcess controlling deals with data. Here the single workflow ent evaluation dimensions schedetection of long-term developmof already existing workflow itions in process execution, thedata derived from correspondingflow-based controlling is the imits effects are more long-lasting

Figure 3: Process State and Business State

C

B

A

D E

ProcessState

A'

B'

C'

BusinessState

D' E' Figure 4: Technical Monitorin

ese systems rarely store this kind of data in

ctive, workflow audit trail data is anotherfinancial statements from an accounting sys-ion processing systems. In order to enhance data, it needs to be combined with applica- a common repository for this kind of data,rocessing tools exist that support the control-ation of information stored in a data ware-ion of workflow audit trail data into theides the opportunity of enhancing existing the ability to analyze the business processhis task is non-trivial and we discuss the

gn for Process Information Sys-

Data Into A Data Warehouse Schemaation and loading phase of data warehouse

nverted into a format that fits the data ware-e format of the existing audit trail logs, thatt file structures, an appropriate import mech-sing a transformation algorithm, the propri-s converted into the data base schema of the trail data is machine-generated, the removal of little importance during this step.

by the workflow management system has to suited for the analytical purposes of the data different workflow management systems

n schema for all workflow audit trail formatsThe WORKFLOW MANAGEMENT COALITION

trail format, that could be used as a commonflow vendors (WfMC Interface 5 specifica-, most workflow vendors have defined a pro-ata. Out of 20 vendors claiming conformancexpressed interest in supporting the Interface

ually implemented the standard. It should be

Whereas the target audience for workflow monitoring data is administra-tive IT personnel on the one hand (for technical information) and workflowparticipants on the other hand (for organizational information), workflowcontrolling data is mainly used for enterprise controlling purposes. Analyzedautonomously, audit trail data provides information about the temporalaspects of process execution as well as information about resource utiliza-tion on the process and activity level. However, information about the busi-ness context of a particular process in most cases cannot be answered bylooking at audit trail data alone. This is due to the fact that most workflowmanagement systems do not store data that is processed in the applicationsinvoked during workflow enactment. The WORKFLOW MANAGEMENT COA-LITION has published a definition of three classes of data associated withworkflow systems [30].

• Application data is data beyond the control sphere of the workflowmanagement system. It is managed and stored by the applicationsinvoked during the enactment of workflows, e. g. a letter to a customerthat is managed by the word processing system.

• Workflow-relevant data is managed by applications and has an impacton the control flow of the current process. Typically this type of data isqueried at decision nodes during the process, when the workflow man-agement system has to decide which of several alternative paths to fol-low. Workflow-relevant data can also be used to increase the flexibilityof staff assignment rules (e. g. “IF claim.value()<50,000 THEN per-former.role() = accountant ELSE performer.role() = manager”). This typeof data is read (but not updated) by the workflow management system,but only few systems store this information in their audit trail records.

• Workflow-internal data is managed by the workflow management sys-tem itself and contains information about the current process instance,e. g. the ID of the process starter or the name of the performer of the lastactivity. This information is used to realize run-time specific semanticsin the process flow, such as the assignment of an activity to the managerof the process starter. This is the kind of information found in most audittrail formats.

Many commercial workflow management systems adhere to the WfMCseparation of application data, workflow-relevant data and workflow-inter-nal data. Some systems allow the workflow designer to specify complex datastructures and provide facilities for explicit data flow and transformation

specifications. Nevertheless, ththe audit trail file.

From a controlling perspesource of information, just like tem or log files from a transactthe business value of audit trailtion data. Data warehouses areand elaborate on-line analytic pling recipients during the evaluhouse. Therefore, the integratexisting data warehouses provcontrolling infrastructures withperspective of an enterprise. Timplications in section 3.

3 Data Warehouse Desitems

3.1 Transforming Audit TrailDuring the extraction, transformdevelopment, source data is cohouse schema. Depending on thcould be database records or flaanism needs to be deployed. Uetary schema of the audit trail idata warehouse. Since the auditof missing or incorrect values is

The audit trail schema usedbe converted to a schema that iswarehouse. If audit trails fromhave to be integrated, a commoinvolved has to be developed. has published a common audit denominator by different worktion [29]). In practice, howeverprietary format for their audit dwith WfMC standards, 9 have e5 standard, but only 2 have act

ate changes) during workflow enactment areecording of the audit trail information, mostlyze the data at a higher level of integration.tivity level require the integration of severaltem supports multiple state changes betweenctivity or a process instance, the number of

y is not predefined. Integrating these entriesoperations on large data volumes. For theata warehouse this implies, that during theloading phase, summarized datasets have torocesses, to ease subsequent evaluations.

bject Data.4, most audit trail formats do not contain ahas been processed in the workflow instance.ons in a business context almost always to be linked to some business object infor-

ccount involved or the merchandise handledious research prototype PISA (cf. [32]), wecreated a database entry containing the ID ofID of the business object treated in the pro-the audit trail data with (semantic) business navigate from a particular process instancews for the application of numerous evalua-

ith business object information can create a warehouse designer:

t is a business object. If, during the course ofin different application systems is used, anarticular data set has to be created for everyr to make this data set accessible through at-oriented workflow management systems a java-based application server) provide aept, most stand-alone workflow systems arerent applications without an additional wrap-

noted, that currently no official conformance testing facility for the WfMCstandard exists, so even the two implementation have not been evaluated byan independent organization yet.

Figure 5 shows a section of the WfMC Interface 5 data format. Eachaudit trail entry consists of a CWAD (Common Workflow Audit Data) prefixand suffix. While the prefix is defined by the WfMC, the suffix can be usedfor vendor-specific extensions. Depending on the type of event that occurredin the system, a different set of attributes is recorded in the body of the logentry. The example shown in the figure depicts an audit trail entry for thecreation or start of a process or subprocess. Accordingly, e. g. the termina-tion of an activity would result in a different set of data recorded in the bodyof the log entry. For the transformation of audit trail records into a commondata warehouse schema, this results in the filtering of excess data and thereplacement of missing values, that are not recorded for certain event types.

While the events (such as stthe point of reference for the rusers of the audit trail will anaEvaluations at the process or acaudit records. If a workflow systhe suspend and resume of an aaudit trail entries for an activitcan resulting in complex join design of a process-oriented dextraction, transformation and be generated for activities and p

3.2 Integration of Business OAs pointed out in section 2

link to the application data that However, meaningful evaluatirequire the workflow audit datamation, such as the customer aduring the process. In our prevmanually added an activity that the process instance and the Ocess. This artificial linking of data provides the opportunity toto the business context and allotion methods.

The linking of audit data wnumber of problems for the data

• Not every workflow contexa workflow instance, data artificial wrapper for this pworkflow instance in ordeunified OID. While objec(e. g. those running insidenative support for this concdesigned to work with diffeper.

Figure 5: WfMC Audit Trail Format (excerpt)

rtunity to model and change processes withystem is significantly easier than the adjust-ss logic in application systems. Therefore,g workflow applications will adjust these toonment. The change of the process structure, activity with another results in a change ofrocess enactment. If an analysis is to be per- of process instances, in many cases onlyvant, whose execution path is identical, i. e.ces is identical at the type level. Therefore, ite data warehouse information, that for everyrlying workflow model variant is recorded.

kflow audit trail data, different evaluations evaluation methods can be executed on thealuation requirements of the data warehouse [8], some basic evaluation perspectives can

c format, workflow audit trail data supportstive as well as from a resource perspective.

spective focuses on information that can be temporal and logical information associatedns recorded in the audit trail. Without any

s derived from the activity identifiers as wellorded in the audit trail.

trail contains the following timestamps:

cess

when the activity becomes ready for execu-

e when the activity was activated by a work-y selection from the participant’s work list)

• Business data is subject to side-effects. The notion of application datawas introduced by the WfMC to make sure, that mission critical datawas not locked from use while being used by a (potentially long-run-ning) workflow instance. Reverting this view, application data can bemanipulated by applications outside of the workflow context at anytime. If audit trail data is transferred through a batch procedure once aday, intermediate changes to the business data are invisible to the datawarehouse. To avoid this problem, business data would need to be trans-ferred to the data warehouse as soon as the workflow instance treatingthis business object is finished. For performance reasons, this will onlybe feasible in few cases.

• Synchronicity is not always desired. From a controlling point of view, itis possible that the most recent version of a business object is of morevalue than the version that was current at the time of workflow enact-ment. When, for example, a query is issued “Show all order fulfilmentprocesses from Q1/2000 for customers residing in the U.K.” the resultset may or may not reflect the customers domicile during process enact-ment. The correctness of the result set depends very much on the objec-tive of the person analyzing the data. If the purpose of the query was theidentification of buying patterns for a direct marketing campaign, subse-quent relocations of customers should be included. If, however, the goalof the query is the identification of demographic profiles of specificgeographic regions, the customer addresses as of the time when the pro-cesses were executed is relevant. This example shows, that the purposeof the data recipient determines the synchronicity requirements for audittrail and business data during the import phase.

• Recording audit trails creates potentially large amounts of data. LEY-MANN and ROLLER have pointed out the risk of growing audit trails [14].They estimate the typical number of log entries per activity to be 5 (or 1KB for a fully featured audit trail entry). For 10,000 processes with 10activities per day this creates an audit trail of 500 MB every day, notincluding the associated business object information. Numbers likethese are not unusual. The average number of pages received in the mailroom of the insurance company described in section 4 is on average29,000 per day, resulting in 8,900 different cases that are instantiatedevery day.

• Processes evolve. The oppoa workflow management sment of hard-coded proceover time, enterprises usinmatch their changing envire. g. the replacement of anthe audit trail data during pformed across a collectionthose instances will be relewhose set of activity instanis vital for the validity of thworkflow instance the unde

3.3 Evaluation DimensionsDuring the evaluation of wordimensions are used and varioudata repository. While not all evusers may be known in advancebe predefined. In its most basianalyses from a process perspec

3.3.1 Process perspectiveThe process evaluation per

derived from an analysis of thewith the activities and transitiomodification, this information ias the timestamps and status rec

Typically, a workflow audit

• Instantiation time of the pro

• Starting time of the process

• For each activity

• creation time (the timetion)

• activation time (the timflow participant, e. g. b

e information about the reason for this non-

t systems allow for the recording of the per-our experience, during the introduction of a, this technical option often creates mistruste users of the system. The measurement of an information system is subject to privacyermany, for example, the measurement of information technology requires the agree-If no employee’s representatives exist at theotice has to be posted at the workplace under

ity, segment and process level

ActivityLevel

Analysis

SegmentLevel

Analysis

ProcessLevel

Analysis

• completion time (the time when the activity was marked as com-pleted either by the system or the participant in case of a manualactivity)

• Completion time of the process

Using these timestamps and the differences between them the followinginformation can be derived easily:

• Cycle time of the process (the time elapsed between the instantiation orstarting time and the completion of the process)

• Net processing time (the sum of the time elapsed between the activationand completion of each activity of the process)

• Idle time (the time elapsed between the creation time and activationtime of each activity).

• Frequency of process occurrences (which processes were executed inwhich frequency at a certain hour/day/week/month etc.)

Figure 6 shows different analysis scopes, namely at the activity, segmentand process level. The resulting analyses can be performed at the instancelevel (for the analysis of the behavior of single activities and processes) aswell as at an aggregated level (for the analysis of the behavior of activitiesand processes over time or the computation of average values).

A frequency analysis can provide information about the likelihood of theexecution of certain activities and/or process paths. This information is ofpurely statistical value, if no connection can be made to the business objectsthat were treated in the different process instances (i. e. the reason, why aprocess instance was executed in a certain fashion). This does not mean,however, that these results are not meaningful at all. For example, probabili-ties about the execution of certain process paths could be inserted into theworkflow model for simulation purposes, enabling the process designer toexperiment with different workloads or staff assignments before actualchanges are made to the organization.

In combination with the semantics of the process and the businessobjects involved, the results of a frequency analysis can be used to identifypotential weaknesses or bottlenecks along the process. For example, if thevariance of the cycle time for an order fulfilment process is particularlylarge, the analysis of the business objects associated with extremely fast (or

slow) processes may yield somhomogeneous behavior.

3.3.2 Resource perspectiveMost workflow managemen

former ID in the audit trail. In workflow management systemand resistance among the futurindividual performance throughlaws in some countries. In Gemployee performance throughment of the employee’s union. company in question, a visible n

Figure 6: Analysis at the activ

ss, this information could be used to classifyers, according to the frequency of exceptionsriggered by the different companies. Using a and audit trail data, an analyst can navigatewell as the data structure of the associatedhe process perspective (“what is the averageg process?”) the analyst could single out pro-mporal behavior (“display those processeeded the average cycle time by more than an, the analyst could traverse to either theerspective (“out of this set, display the name

exceeded the cycle time the most”). From anve, these are the most valuable insights, the can generate.

luations available at the business object per-y the semantics of the business objects. Ide-d at the time, the workflow application is

that the data required for the analyses isworkflow management system directly, or

ted Data Warehousedesign of a data warehouse that integratesbusiness object information. The extraction,r imports workflow audit trail data as well ason repository. The data is formatted accord-

ets that describe the format and semantics ofes.

rms various evaluation methods on the pro-d delivers the results as an evaluation result

ouse size reasons it may be feasible to creates or analysis purposes (e. g. a specific data analyses). Finally, the visualization layer presents them to the user in a usable formatart, depending on the type of result data).

right of the data warehouse diagram indicateand use of the process-oriented data ware-

surveillance. Therefore, a sensible measure during the initial setup of aworkflow management system is the creation of a formal charter that con-tains the level of detail for the audit trail as well as the evaluation purposes,this data is collected for.

Combining temporal information about process performance with infor-mation about the resources involved provides opportunities to create mean-ingful financial indicators. If the cost factor for the utilization of a resource isknown, the (personnel) cost for the execution of a process instance an becomputed by multiplying the processing times of the activity instances withthe cost factor of the workflow participants. This simplified approach doesnot take into account work that is performed outside the scope of the work-flow management system, e. g. answering to a customer call while anothercustomer order is being processed inside the workflow system. However, fororganizations that use workflow systems for administrative processes (wherepersonnel cost account for the majority of process cost), the use of workflowaudit trail data can lead to a better accountability of overhead cost.

Another evaluation using the resource perspective is the analysis ofresource utilization at the activity or process level. This analysis answersquestions such as “How many different participants were involved in anordering process?” or “How many work items typically reside in the worklist of an accounting clerk on average?”

In [32] we have discussed the application of SASSONE’S hedonic wagemodel to workflow audit trail data. By classifying workflow participants intogroups such as executives, experts and administrative staff, and by assigningcategories to workflow activities that reflect these groups (for example exec-utive tasks, administrative tasks, expert work), the analysis of audit trail dataprovides information about the accuracy of work assignments. Using thehedonic wage model, the financial effects of organizational restructuring onprocess performance can be visualized.

3.3.3 Adding Perspectives through Business ObjectsThe combination of audit trail data with semantic information about thebusiness objects allows for a multitude of evaluations. Depending on theprocess semantics as well as the information available about the businessobjects, numerous evaluation dimensions can be derived.

An example for an analysis originating at the business object perspectiveis the analysis of processing times grouped by the vendor of the process

object. For a supply chain procesuppliers into A, B and C suppligenerated within the processes tcombination of business objectalong the process structure as business object. Starting from tcycle time of an invoice auditincesses according to their teinstances, whose cycle time excweek”). Using this informatioresource or the business object pof the supplier, whose invoices enterprise controlling perspectiuse of workflow audit trail data

The number and kind of evaspective is determined mainly bally, the evaluations are definebeing implemented, to ensurerecorded (either through the through external programs).

3.4 Designing a Process-OrienFigure 7 shows the schematic workflow audit trail data with transformation and loading layebusiness object data into a comming to the respective metadata sthe underlying raw data structur

The evaluation engine perfocess-oriented data warehouse anset. For performance and warehdata marts for specific processemart for activity-based costingreads the evaluation results and(e. g. as a table or a graphical ch

The arrows on the left and the steps taken during design

lfilment of these requirements can be identi- to know in order to solve the problem atements are the major guideline for the selec-ion method (which evaluation delivers theoice of an evaluation method determines theriented data warehouse (what source data iss evaluation method?).

the design of a process-oriented data ware-agerial controlling requirements need to beorkflow is designed. Subsequent changes tosult in necessary adjustments of the underly-ch can be both time-consuming and expen-

orkflow Audit Trail Data in an

tline the practical relevance of our approach was conducted at a medium-sized German

employees. The company had conducted aneering project two years before the caseis project, the major processes affecting thehad been documented and grouped into clus- employees, the average processing time forth a resolution of minutes). Based on thisosting system (ABC) had been implemented.

een in use for two years, it became apparentocesses had changed. Also, certain informa-ses such as capacity planning was not avail-. A project was set up to investigate the-driven process information system for the

activity-based costing system. It is character- most of which are transferred manually intom. The activity structure of the business pro-

house. During the operation of the process-oriented data warehouse, a con-tinuous feed of audit trail data will be integrated (either continuously or viabatch procedures), evaluated and presented to the user. The design processreverses this flow. Starting with the business requirements of the user (whatkind of business problem or decision needs to be addressed by the user?) the

information necessary for the fufied (what does the user needhand?). The information requirtion of an appropriate evaluatresults needed?). Finally, the chrequired entries in the process-orequired for the execution of thi

This top-down approach at house illustrates, that the manstated already at the time the wthe evaluations required may reing workflow application, whisive.

4 Case Study: Using WInsurance Scenario

4.1 PreconditionsIn the following section we outhrough a case study. The studyinsurance company with 1,200large business process re-engistudy. During the course of thinsurance-specific departments ters. Based on statements fromactivities was documented (wiinformation, an activity-based c

After the new system had bthat the underlying business prtion that was required for purpoable from the existing systempotential benefits of a workflowcompany.

4.2 Information AvailabilityFigure 8 illustrates the existing ized by a variety of data feeds,the activity-based costing syste

Figure 7: Design of a Process-oriented Data Warehouse

f a (traceable) transaction in the transactione ABC-system documented 20,000 cases of

insurance per year. However, the actual num-higher, since many customers could be con-on request, before it was put into effect usingm. These cases, even though their processingces, did not appear in the ABC-system. Fur-ction processing system jeopardized the linkture and the transactions, since the measuredrform a certain activity was tied to the trans-

hen a transaction was changed the process-ctivity changed, but this change was not

rkflow Data Integration shortcomings, the concept for a workflow-gned, that should be used as a data feed for9 shows the structure of the target scenario.ture controlling infrastructure is a data ware- repository for all operational data the needs audit trail is transferred into the data ware-rt filters. On the one hand, the process and for the process perspective evaluations. On

rmation (which resource performed whiche perspective evaluations. In order to provide process structure, to-be information aboutransferred into the data warehouse at build-mple phone calls), however, will still not be

ail information to operational business data,d separately. One table contains the relation-nces and the insurance contracts or the cus-ble as a bridge, a drill-down from a singleriggered the process can be performed. The relationship between the workflow partici-counting structure of the enterprise. Since only record a technical ID of the performers

cesses was transferred manually from Excel sheets into the proprietaryABC-system. Transactions from the operational insurance application sys-tem were assigned to activities according to a transformation schema devel-oped during the BPR project. Since the number of transactions wererecorded in the legacy system, the log files were used to compute the actualnumber of activities performed in a certain period. Using the estimated pro-cessing times per activity, the required personnel capacity per period wascomputed using a custom Excel application. This data was combined withthe activity structure and the internal accounting information, which was col-lected in a data warehouse application and manually transferred into theABC-system.

Besides the fact that the whole evaluation process was time-consumingand error-prone, due to many media-breaks and manual transfers, the resultsof the ABC-system did not reflect the actual operations at the insurancecompany. For example, only those processes appeared in the ABC-system,

that resulted in the execution oprocessing system. This way, thcustomers cancelling their auto ber of cancellations was much vinced to revoke their cancellatithe transaction processing systeconsumed time and kept resourthermore, changes to the transabetween the BPR activity strucamount of time necessary to peaction used within the activity. Wing time for the associated adocumented anywhere.

4.3 Improvement through WoAfter the identification of thesebased data warehouse was desithe existing ABC-tool. Figure The central component of the fuhouse, that serves as the centralto be evaluated. The workflowhouse using two separate impoactivity information is extractedthe other hand, resource infoactivity) is extracted for resourcsemantic information about thethe processes and activities is ttime. Manual activities (for exarecorded by the new system.

In order to link the audit trtwo tables need to be maintaineship between the process instatomers involved. Using this taprocess to the customer who tother custom table contains thepants and the internal cost acworkflow management systems

Figure 8: Controlling Data Flow (As-Is)

Data pool

SASData Warehouse

Activity-basedCosting Tool

InternalAccounting

System PC

Internalaccounting

sheets

BPRProcess

documentation

Personnel capacity usedper activity (manual transfer)

Activity structure(manual transfer)

Internalaccounting sheets(manual transfer)

InsuranceTransactionProcessing

System(legacy)

Data

DataData

OperationalStatistics

Number of (specific)transactionsper period

Number oftransactions

Transformationto activityquantities

(BPR factor)

Number of (specific)transactionsper period

Transformationto personnel

capacity used(BPR factor)

Activitiesperformed

Log files

es for the realization of the proposed infra- project plan.

h projects that aim at combining data ware- management systems. The existing sources

ries: Those who apply workflow concepts to those who describe the analysis of workflow

epts to Data Warehouse Designology for the design of data warehouses aim data warehouse. A notable exception is the at designing a generic data warehouse for]. Since the retrieval and transformation ofarehouse is a frequent process, whose steps

andidate for workflow automation.

the modeling of the data warehouse refresh-. They use an event-driven approach to work-s parts of the refreshment process either afterter a certain condition in the data warehousey distinguish between Client-driven refresh-es an update to the data warehouse structure,n the changes to the source data of the data

nt process, and ODS-driven refreshment,

low and data warehouse concepts and identi-arehouse design process that could benefit

hnology, such as handling update anomaly

it Trail Datause concepts to workflow technology can befirst group of publications in this area dealsessary for the logging of the audit trail data papers describes the reverse engineering oftrail data [1]. Finally, a number papers can be

who executed activities, this table is required to allocate cost factors to theworkflow participants.

The realization of the process-oriented data warehouse will enable theinsurance company to evaluate its operational business processes in a timelyfashion and with a much higher level of accuracy than currently possible. Itshould be noted, that the existence of a workflow infrastructure is the majorprecondition for the realization of such an information system. At the com-pany in question, a project was initiated to lay the foundation for a futuredata warehouse by selecting and implementing a combined document man-agement and workflow infrastructure. Even though the financial benefits ofan improved process controlling infrastructure are hard to determine, thefinancial savings from the introduction of the document management andworkflow infrastructure were sufficient to provide a return on investmentjust three years after the deployment of the new system. The ability to per-form detailed process analyses is perceived as an added benefit of the new

infrastructure and work packagstructure have been added to the

5 DiscussionThere are relatively few researchousing concepts with workflowcan be grouped into two categothe data warehouse domain andaudit trail data.

5.1 Applying Workflow ConcSources that use workflow technmainly at the design phase of awork by LIST ET AL. who aimbusiness process evaluation [15operational data into the data ware structured, it is a potential c

BOUZEGHOUB et al. describement process as a workflow [3]flow modeling to trigger varioua certain timer has expired or afhas occurred. In particular, thement, when a user activity causSource-driven refreshment, whewarehouses trigger a refreshme

PATTERSON discusses workffies several areas in the data wfrom the use of workflow tecproblems [21].

5.2 Analyzing Workflow AudThe application of data warehodivided into three groups. The with the technical facilities nec(e. g. [27]). A second group ofworkflow processes from audit

Figure 9: Controlling Data Flow (To-Be)

DataWarehouse

Activity-basedCostingSystem

InternalAccounting

System

Internalaccounting

sheets

WfMS Audit Trail

ProcessQuantity

ResourceQuantity

TransactionProcessing

Systems andDatabaseSystems

Customer andContract Data

Cost AccountingCost Accounting Structure

RelationshipWorkflow-Roles/Cost Accounting

Structure

RelationshipProcess-ID /

Customer

WfMS RuntimeData

Process andWorkflowModeling

Processand ActivityStructure

WorkflowModels

ManualActivities*

Imaging/DMS

Feed-back

Date/Time Received

ABC-data

submitted to the Coalition, but so far noge the interface from a data format specifica-.

isting WfMC standard with regard to a pro-n to generate the process perspective of a bal-k aims at the development of a closed-loopormation as an input and delivers advice tots of the workflows analyzed could be opti-

workflow based controlling can be found inverview of the analysis of historical processn of workflow history data as workflow met-s described are statistical evaluations as well cases and overdue tasks.

ethodology for the identification of processs of work practices [24]. Even though theirR”, it does not require the use of a workflowe basic idea of determining the knowledge

kflow audit trail data can be perceived as aght. For example, the experience of a work-uted taking the number of times into account

ed a certain activity type. When this infor-kflow management system, new staff alloca-nted, such as “assign this activity to the most

esign of a controlling tool called WorkFlownctional prototype was implemented on then audit trail data from the workflow manage-mens Nixdorf and to-be data from the busi-RIS and Bonapart, the prototype providesmethods. However, the combination of theusiness data was not realized using this

he controlling of workflow-based processesanufacturing domain. In his example, thecess, which he enhances with cost informa-

identified, that focus on the ex-post analysis of workflow audit trail data withmethods known from enterprise controlling ([5],[22],[19],[28],[32],[33]).

5.2.1 Technical Facilities for Audit Trail GenerationKOKSAL et al.discuss the management of audit trail data in the distributedworkflow management system Mariflow [11]. The paper focuses on techni-cal issues regarding the storage of audit data and economic queries on dis-tributed data sources, but the authors do not address the business value of thehistory data.

An approach for the tracking of history information in a distributedworkflow management system is presented by MUTH et al. [18]. Within theprototype Mentor-lite, data about current and past workflow instances arekept in a temporal database that can be queried either at runtime or for ex-post analyses.

5.2.2 Reverse Engineering Using Audit Trail DataAGRAWAL et al. use data mining techniques to create workflow models fromaudit trail data [1]. The purpose of their project is to use data from the ad-hocexecution of processes, to subsequently identify common rules and proce-dures and to create workflow models using a bottom-up approach. Theauthors have tested their approach against artificial data sets and audit traildata from a live workflow installation. However, the practical use of themethodology presented requires the existence of a flexible workflow tool,that records processes while they are being created on the fly. Despite theobvious demand for such a tool, the current workflow market is lackingproducts with this kind of execution flexibility.

5.2.3 Workflow-based Controlling in the LiteratureWhereas the controlling of processes using workflow audit trail data hasbeen analyzed in the German literature to some extent (for example[5],[22],[28]), there are relatively few English sources dealing with this topic(for example [19],[32],[33]).

The WORKFLOW MANAGEMENT COALITION Interface 5 specifies the ele-mentary information a workflow management system should record the exe-cution of workflow instances [29]. The existing standard provides a dataformat for the audit trail data as well as guidelines, which events should berecorded. However, the evaluation of this information is not addressed in theWfMC standard. After the publication of the current standard document, a

proposal for an API has beenprogress has been made to chantion to a functional specification

MCGREGOR discusses the extotype, that uses this informatioanced scorecard [16]. Her worsystem that takes audit trail infprocess designers, which aspecmized.

One of the first sources on [19]. MCLELLAN provides an odata. He discusses the evaluatiorics. The controlling applicationas the run-time detection of late

SHOHAIEP ET AL. present a mknowledge through the analysimethod is called “Workflow*BPmanagement system. Taking thincorporated in a process, wor“knowledge store” in its own riflow participant could be compthat this participant has performmation is fed back into the wortion methods could be implemeexperienced person available”.

DERSZTELER discusses the dAnalyzer in [5]. A partially fubasis of Forest&Trees. Based oment system WorkParty by Sieness process modeling tools Aseveral quantitative evaluation audit trail information with bapproach.

In [22], RAUFER discusses tusing a case study from the mauthor focuses on a specific pro

cess into the center of attention. Workflow-us entirely on information contained in thentitative analysis of the event-driven history

orkflow-based controlling approaches toat is not stored in the traditional audit trail. evaluations is widened to include qualitativeion of business processes through their asso-

or a more detailed segmentation of the analy-

Adapted from [12], pp. 425 ff.

kflow-based Controlling

tion as well as target work- and cycle-times. The presented prototype, whichis based on the workflow management system COI, is targeted specifically atthe process analyzed by the author. As a result, the system architecture is noteasily generalizeable.

A similar prototype is presented by WEISS in [28]. He developed a work-flow-driven activity-based costing system for the commercial workflowmanagement system Staffware. The focus of this approach is the realizationof a single evaluation method, therefore, the resulting system is not designedto be extended by additional evaluation methods.

The COPPA project (Computer-based Process Performance Measure-ment) conducted at the University of Fribourg, Switzerland, deals with thedesign of a performance measurement system ([12], [13]). Using a threestage approach, the authors first surveyed the market and corporate practiceof performance measurement. During a second stage the architectural andfunctional requirements of a performance measurement system were out-lined, and during a third phase, a prototype of the performance measurementsystem was implemented. In relation to workflow-based controlling, theauthors position process performance measurement at a higher level ofabstraction, that includes information about the strategic positioning of anenterprise, whereas a workflow-based controlling system is mainly focusedon the analysis of operational data.

6 Summary and OutlookWorkflow-based controlling provides companies with the ability to accu-rately measure the operational performance of business processes. Com-bined with data warehouse technology and operational business data,complex evaluations can be performed, that help enterprises to assess theircurrent situation more precisely than the sole use of traditional key perfor-mance indicators. The increasing maturity of workflow and data warehouseproducts provide an important precondition for the implementation of work-flow-driven controlling systems.

It should be noted that workflow-based controlling does not replace othercontrolling mechanisms of the enterprise, but enhances them significantly.Figure 10 shows the positioning of workflow-based controlling in relation toother controlling techniques. While strategic controlling instruments like theBalanced Scorecard focus on qualitative aspects of the enterprise, and tradi-tional controlling methods rely on financial information, few techniques

exist, that put the business probased controlling tools that focaudit trail are limited to the quaof process execution.

Our research enhances winclude business information thThis way, the scope of potentialinformation, and the differentiatciated business objects allows f

Figure 10: Positioning of Wor

l. Conference on Data Engineering, New 996).F. and Matulovic-Broqué, M., “Modeling ment Process as a Workflow Application”, ternational Workshop on Design and Man-uses (DMDW ’99), eds. Stella Gatziu, Man-Staudt, Yannis Vassiliou, Heidelberg, 6.1-

le.comanagement auf Basis von Workflow-Syste-ohmar et al. (2000).

w Market: A Global Perspective for 2000”, , Volume IV, Innovation and Excellence in e Management, ed. Layna Fischer, Future int, 41-47 (2000).W., “Workflow-Projekte: Vom l zur unternehmensspezifischen Workflow-ation Management, Vol. 10 No. 1 (1995), 20-

the Data Warehouse”, in: Communications . 9, (1998), 52-60.ration der Aufbauorganisation in Workflow-n: Information Engineering, ed.: Heidi Heil- Munich, Vienna 147-165 (1996).. and Sousa, R., “Corporate Information Fac- New York et. al., (2001).. and Dogac, A., “Workflow History Man-cord 27 (1998) 1, 326-348 (1998).PR though a Process Performance Measure-ss Information Technology Management,

r-Anand Publications, New Delhi, 422-434

rmance Measurement System: a tool to sup-izations”, in: Total Quality Management, 000), 67-86 (2000). D., “Production Workflow - Concepts and ll, Upper Saddle River (2000). A. M. and Quirchmayr, G., “The Process ehouse Approach for Business Process Man-

sis domain. Since this approach relies on the workflow-enabled businessprocesses of an enterprise, there will be a number of blind spots at the enter-prise level (i. e. manual processes, managerial processes without workflowsupport etc.). However, the development of enhanced workflow-based con-trolling systems can serve as a data-feed to other controlling approaches, likethe Process Performance Measurement Systems discussed by KUENG in [12]or the process perspective of the Balanced Scorecard.

We have discussed the technical and organizational aspects of building aprocess-oriented data warehouse and used a case study to illustrate the prac-tical relevance of this type of application. Nevertheless, many aspects ofworkflow-based controlling systems require further research.

So far, existing papers focus on the realization of single evaluation meth-ods, like activity-based costing [28]. An analysis, which existing evaluationmethods can be applied in the context of a workflow-based controlling sys-tem based on the attributes recorded in workflow audit trail data would pro-vide further insight about the coupling of audit trail data and business objectinformation.

Also, the handling of changing process models variants over time resem-bles the problem of schema evolution in databases. Further research in thisdirection would help to open workflow-based controlling systems to incor-porate audit trails from workflow management systems that support ad-hocprocesses.

In the context of the insurance project described in section 4, we are cur-rently working on a prototypical implementation of the concepts presentedin this paper using a java-based workflow management system which will beevaluated using operational data from the business processes of this com-pany.

7 References[1] Agrawal, R., Gunopulos, D., Leymann, F., “Mining Process Models

from Workflow Logs”, in: Advances in Database Technology - Pro-ceedings of the 6th International Conference on Extending Database Technology, ed. Fèlix Saltor, Isidro Ramos and Gustovo Alonso, Valencia, 23-27 (1998).

[2] Alonso, G., Agarwal, D., Abbadi, A. E., Kamath, M. and Gunthor, R., “Advanced transaction models in workflow contexts.” in: Proceed-

ings of the 12th IEEE IntOrleans, LA, February (1

[3] Bouzeghoub, M, Fabret, Data Warehouse Refreshin: Proceedings of the Inagement of Data Warehofred A. Jeusfeld, Martin 6.12 (1999).

[4] Candle Corp.: www.cand[5] Derszteler, G., “Prozessm

men”, Josef Eul Verlag, L[6] Emery, P., “The Workflo

in: Excellence in PracticeWorkflow and KnowledgStrategies, Lighthouse Po

[7] Galler, J. and Scheer, A.-GeschaeftsprozessmodelAnwendung”, in: Inform28.

[8] Gardner, S. R., “Buildingof the ACM, Vol. 41, No

[9] Heilmann, H.: “Die IntegManagement-Systeme” imann et al.. Oldenbourg,

[10] Inmon, W. H., Imhoff, Ctory”, 2nd Edition, Wiley,

[11] Koksal, P., Arpinar, S. Nagement”, in: Sigmod Re

[12] Kueng, P., “Supporting Bment System”, in: Busineed. P. Banerjee et al., Ha(1998).

[13] Kueng, P., “Process Perfoport process-based organVol. 11, No. 1 (January 2

[14] Leymann, F. and Roller, Techniques”, Prentice Ha

[15] List, B., Schiefer, J., TjoaWarehouse - A Data War

ds. Bengt Wangler and Lars Bergman, Stock-

Monitoring: Queries on Logs or Temporal er at HPTS'95. Pacific Grove, (1995).

nrechnung und Workflow Management”, Verlag, Wiesbaden (1998).oalition, “Audit Data Specification. Version FMC-TC-1015, Winchester (1999).oalition, “Terminology and Glossary” 3rd

ber WFMC-TC-1011, Winchester (1999).en, R., “Embedded vs. Autonomous Work- into Perspective”, in: Excellence in Practice, d Excellence in Workflow and Knowledge

Fischer, Future Strategies, Lighthouse Point,

semann, M., “Workflow-based Controlling”, th Hawai’i International Conference on lph Sprague Jr., Wailea (2000).flow-based Process Controlling - Or: What an Control”, in: Workflow Handbook 2001, Strategies, Lighthouse Point, 61-78 (2000).

agement”, in: e-Business and Intelligent Management - Proceedings of the International Conference on Management of Information and Communication Technology (MiCT 1999), Copenhagen, Denmark (1999).

[16] McGregor, C., “The Impact of Business Performance Monitoring on WfMC Standards”, presentation at the Workflow Management Coali-tion meeting, New York, NY, May 2001.

[17] Moore, C., “Workflow Goes Mainstream”, Giga Information Group Planning Assumption, Cambridge, MA, April 21, 2000.

[18] Muth, P., Weissenfels, J., Gillmann, M. and Weikum, G., “Workflow History Management in Virtual Enterprises Using a Lightweight Workflow Management System”, in: Proceedings of the Ninth Inter-national Workshop on Research Issues on Data Engineering, Sydney, 148-155 (1999).

[19] McLellan, M.: “Workflow Metrics - One of the great benefits of workflow management” in: Praxis des Workflow-Management, eds. Hubert Oesterle and Petra Vogler, Braunschweig, 301-318 (1996).

[20] Naef, A., Schuler, Ch. and Schuldt. H., “Monitoring komplexer Dien-ste in unternehmensübergreifenden Prozessen am Beispiel von SAP R/3 Business Workflows”, in: Proceedings of the 9th GI-Conference Datenbanksysteme in Buero, Technik und Wissenschaft (BTW 2001), 85-94 (2001)

[21] Patterson, J. E., “Workflow and Data Warehousing Taxonomy”, Mas-ters Thesis, University of Texas, Arlington (1996).

[22] Raufer, H., “Dokumentenorientierte Modellierung und Controlling von Geschäftsprozessen”, Deutscher Universitaets-Verlag, Wiesbaden (1997).

[23] Sassone, P. G., “Cost-benefit methodology for office systems”, in: ACM Transactions on Information Systems, Vol. 5, No. 3, (1987), 273-289.

[24] Shohaiep, G., Housel, T. J. and Kanevsky, V.: “Workflow-BPR White Paper: Business Process Auditing with Workflow-BPR”, White Paper, University of Southern California, (1997). URL: http://www-rcf.usc.edu/~housel/IOM499/wp.html [2001-07-15].

[25] Tivoli Corp.: www.tivoli.com[26] Vassiliadis, P., Quix, C., Vassiliou, Y. and Jarke, M., “A Model for

Data Warehouse Operational Processes”, in: Proceedings of the 12th International Conference on Advanced Information Systems Engi-

neering (CAiSE 2000), eholm, 446-461 (2000).

[27] Weikum, G.: “Workflow Databases?” Position Pap

[28] Weiss, D., “ProzesskosteDeutscher Universitaets-

[29] Workflow Management C2”, Document Number W

[30] Workflow Management CEdition, Document Num

[31] zur Muehlen, M. and Allflow - Putting ParadigmsVolume IV, Innovation anManagement, ed. Layna 49-58 (2000).

[32] zur Muehlen, M. and Roin: Proceedings of the 34Social Sciences, ed. Rap

[33] zur Muehlen, M., “WorkYou Can Measure You Ced. Layna Fischer, Future