Consider whether activities may be executed in parallel
Reijers, H., & Liman Mansar, S. (2005). Best practices in business process redesign: an overview and qualitative evaluation of successful redesign heuristics. Omega, 33(4) All sources
category: Business Process Behavior class: Routing Rules frameworkAspect: Behavioral view redesignOperation: Parallelize perspective: control-flow
Many processes are designed and executed as a strict sequence of activities, even though only some of these activities depend on the output of the previous one. This pattern asks you to check, for each sequence, whether a real logical dependency exists and, where it does not, to execute the activities in parallel. A case then no longer waits for activity A to finish before activity B starts, and its throughput time is determined by the longest parallel branch rather than by the sum of all activities.
Parallelisation requires that the activities share no data dependency, that they can be performed by different resources, and that their results can be merged with reasonable effort. Technology often makes this possible in the first place: two people cannot work on a paper file at the same time, which is why the pattern is frequently enabled by Apply Integral technology. In processes with knock-out checks there is a tension with Fail Early, because parallel checks shorten throughput time but spend effort on cases that a single check would have stopped. Resequence activities is the more general pattern, and parallelisation is one of its strongest forms.
The heuristic is mentioned by Rupp and Russell (1994), Berg and Pottjewijd (1997), and Van der Aalst and Van Hee (2002). Van der Aalst (2000b) provides quantitative support for this heuristic.
■ This pattern ■ Baseline
The obvious effect of applying this heuristic is that the throughput time may be considerably reduced. The applicability of the heuristic in workflow redesign is large. In practical experiences we have had with analyzing existing workflows, tasks were mostly ordered sequentially without the existence of hard logical restrictions prescribing such an order.
A drawback of introducing more parallelism in a workflow that incorporates possibilities of knock-outs is that the cost of workflow execution may increase. The management of workflows with concurrent behavior can become more complex also, which may introduce errors (quality) or restrict run-time adaptations (flexibility).
On balance, the pattern is rated as improving all four dimensions of the Devil’s Quadrangle, with time by far the strongest. Shorter throughput reduces work-in-progress and status inquiries, and decoupled activities can be scheduled more freely. This assumes the drawbacks above are managed, e.g. knock-out checks before the parallel section.
Framed along the FLAC rule types — fitness, location, attribute, constraint.
Parallelism served as the worked example of the FLAC method. Fitness: the current degree of parallelism in the process is low (the example threshold was below 0.4). Location: pairs of activities that are always executed in sequence. Attributes: similar mean durations, so that parallel execution saves substantial time, and different executing resources. Constraints: no data dependency between the activities; the time saved must exceed the added merge time; enough resources must be available to work on both branches at once. In aBPR, parallelisation was among the patterns at the highest automation level (AL4), where alternative models are generated and simulated automatically.
Foundational free Patterns
Elevate physical constraints by applying new technology
Consider the division of a general activity into two or more alternative activities
Assign a responsible individual for handling each case type
Let workers perform as many steps as possible for single cases
Consider automating activities
Automate for environmental impact
Implement automation in a sustainable way
Buffer external information and subscribe to updates
Combine small activities into composite activities
Collect similar work items and work in batches
Constraint-based task assignment
Allocate tasks considering business process execution constraints