The Yet Another Work ow Language (YAWL) language and environment has been used to prototype, verify, execute and analyse business processes in a wide variety of industrial domains, such as telephony, construction, supply chain, insurance services, medical environments, personnel management and the creative arts. These engagements o er the YAWL researcher community a great opportunity to validate our research ndings within an industry setting, as well as discovery of possible enhancements from the end user perspective. This paper describes three such industry projects, discusses why YAWL was chosen and how it was used in each, and re ects on the insights gained along the way.
The YAWL language and environment is well-known in the areas of research, for topics such as veri cation, con guration, exception handling, visualisation, riskawareness and cost-awareness, and teaching, having been taught in the courses of at least 36 universities across 20 countries. Originally intended as a reference implementation of work ow patterns, the open-source YAWL environment provides an ideal testbed for novel research ideas. In the decade since the YAWL language was rst developed, we, as YAWL researchers, have modelled numerous YAWL processes that re ect various scenarios in di erent application domains { ranging from structured transactional processes (e.g., insurance claim, travel booking, conference paper review processes) to manual and exible processes (e.g., scheduling of medical operations). In most cases, our main focus is on evaluating a speci c research artefact. Hence, we abstract from things that are less relevant to our purpose, for example, in a particular research project we might not spend much time on the development of intuitive user interfaces or we might not fully integrate a YAWL process with external systems. In addition, there may be limited stakeholder involvement in terms of user acceptance testing of the results of particular projects.
In recent years, a number of organisations have initiated YAWL deployments to prototype, verify and execute their business processes. This paper summaries the insights gained from our involvement in such projects in three di erent domains: construction, the creative arts, and telephony, whereby we demonstrated how YAWL can be used to model and automate processes within a particular domain in collaboration with domain experts who were actively involved in requirements analysis and design, process modelling, implementation and acceptance testing of the resulting system. 2
Each of the selected case studies was undertaken in close collaboration with domain experts, and each case delivered a YAWL environment tailored to organisational needs. The prototype process for construction is used for demonstration purposes, the YAWL4Film system was trialled and deployed in the real lm production setting, and the system developed by rst:utility was for ongoing production. A brief overview of the three case studies is given below. 2.1
Construction processes are complex, involving planning of activities from design to build, coordination between various teams (e.g. designers, contractors, builders, engineers, inspectors), and intensive use of building materials, equipment, and tools. In recent years, o -site manufacture (OSM) has been recognized as an e ective way to reduce cost at the procurement stage of construction. Currently, construction management practices are predominantly manual and founded on the project manager's experience. This case study aimed to demonstrate how business process automation techniques can assist in capturing OSM requirements within a construction project.
A collection of process models (using BPMN notation) that represent the
generic construction value chain was developed [
Processes in the eld of screen business are characterised with high demands for
creativity and exibility. A typical process is one that supports lm production,
an important phase that encompasses when the majority of cast and crew are
contracted and the majority of the equipment and resources are utilised. A lm
production is a highly manual process that requires handling large amounts
of heterogeneous data on a daily basis and coordinating many geographically
distributed stakeholders. The bene ts o ered by applying automation to such a
process are twofold: it may ultimately reduce the production costs by minimising
errors and delays; and by saving time otherwise spent in costly and tedious
procedural tasks, the production team can focus more on creative activities thus
increasing the quality of the nal product [
The development of a prototype system, namely YAWL4Film demonstrated
that the bene ts stated above can be brought to the eld of screen business
through the application of work ow technologies [
rst:utility
First Utility Ltd ( rst:utility) is an independent utilities group based in the
United Kingdom. Amongst other interests, it provides a variety of competitive
telephony services to its customers, through the leasing of wholesale telephony
lines from British Telecom (BT), which it on-sells to customers at
competitive rates [
An implementation of the YAWL environment was chosen by rst:utility to provide work ow support for both back-o ce automation and the humancontrolled work activities of their Customer Relationship Management (CRM) systems. For back-end automation such as the line transfer order process, YAWL provided a number of advantages. One was the ease with which a custom service could be developed to interface with the SPG. Another was the various mechanisms available to build-in exception detection and recovery into longlived processes by design. For CRM work ows, rst:utility developed their own worklist extension that generates web-based forms using runtime attributes and XSLT transformations.
For all three case studies, the YAWL language was found to be capable of
supporting all the required control ow behaviours for the processes, making use
of various complex control ow constructs including multiple instances,
cancellation, advanced synchronisation, composite and timer tasks [
Another advantage of the YAWL environment that made it attractive to rst:utility was its scalability. With many thousands of concurrent process instances, rst:utility was able to scale the environment across a number of servers, running several engine instances in parallel. They also created a `quarantine' server, where processes that had experienced a failure could be ported and closely examined to determine the root cause of the problem. Once recti ed, the process could be moved out of quarantine and back into the production environment. Although the auto-generated web forms provided by the YAWL environment are very useful for rapid prototyping purposes, we found that most end users prefer customised web forms in the nal product. The YAWL environment supports this customisation requirement in two ways. First, by using the extended attributes feature, the colour, font and label names of auto-generated forms can be customised at design time. This enables a YAWL consultant to quickly change the look-and-feel of auto-generated forms. Based on our experience in applying YAWL within industry settings, the key strengths of the YAWL environment can be summarised as follows. { The YAWL language is expressive and supports all necessary constructs for industry projects. { The YAWL environment is an excellent tool for rapid prototyping. { The extensibility of the open-source YAWL environment makes it possible for organisations to develop their own customised services. { The YAWL environment enables the development of customised user interfaces allowing organisations to ensure consistent look and feel across their applications.
Some of the key lessons learned from our experience to develop and deploy a YAWL system are as follows. First, although it is quick to develop a working YAWL prototype, a signi cant amount of project time is spent on developing customised user interfaces and custom services in most cases. Secondly, to develop a YAWL system that makes appropriate use of complex control ow constructs supported by the YAWL environment, a YAWL consultant should be well versed in the semantics of the YAWL language. The same is true for the data and the resource perspectives supported by the YAWL environment. Finally, a YAWL consultant should take into account the fact that the YAWL environment is an open-source research environment with functionalities being added by many participants over time and that some of these functionalities may not have been extensively tested before their release.
Work currently in progress, including a redesign of the YAWL Editor and an automated system update facility, may encourage more widespread adoption of the YAWL language and environment in industry. More extensive training materials have also been identi ed by designers and domain experts as desirable.