In our joint research project RISE_BPM, we explored four information technology enablers, comprising Social Computing, Smart Devices, Big Data Analytics, and Real-Time Computing. Subsequently, we briefly present each enabler and some of its impacts on the BPM field.
Social Computing
For white-collar workers and customers alike, Social Media present an opportunity to network with each other and establish digital communities that foster communication, cooperation, and collaboration on a group level.
Social Media are means to make information, such as personal opinions, facts, recent experiences, and stories available at different levels of public accessibility. They enable users to communicate with a theoretically unbounded crowd of other people about products and the companies providing them. Based on these interactions, Social Media contain a partially unfiltered source of information that typically transcends the boundaries of a single organization, club, association, or company. Social Media can be as diverse as online forums, including blogs, company-sponsored discussion boards and chat rooms, consumer -to -consumer e-mail, consumer product or service ratings websites and forums, Internet discussion boards, and social networking websites, to name a few (Kaplan and Haenlein 2010).
User-Generated Content (UGC) has a significant impact on tools and strategies adopted by companies to communicate with their customers (Mangold and Faulds 2009). In Social Media, data are published with a direct attribution of the author and the exact time and date of publication. The main content of the message is conveyed through natural language, thus making published data semi-structured. Limiting their automated interpretation, user-generated content often contains abbreviations, idiomatic expressions, and emoticons. Tags and links enrich the semantics of a message, which is critical to conduct machine-driven information linkage.
Still, the extraction and analysis of this UGC can represent a valuable source of knowledge to companies. Examples of such sources of information include complaints via Instagram posts about the delivery of a defected product, or suggestions for improvements via the product user forum of an e-mail service provider, as well as tweets about a recent patent, publication, or released product from the creator. For instance, DELL has analyzed social media posts to identify more than 550 new ideas for their products based on analyzing UGC on their online community Idea Storm (Gardner 2014). The opportunities related to analyzing UGC have lead to a florescence of data mining techniques applied on customer information to ameliorate customer relationship management (Ngai et al. 2009).
Within their own boundaries, many organizations offer their workforce collaboration tools – including Groupware applications and Corporate Social Media – to enable them to perform knowledge-intensive processes and knowledge work. White-collar workers take advantage of the tools to communicate, cooperate, and coordinate their activities. Tools include, among others, instant messaging, e-mail (Geyer et al. 2006), and tools for designing and executing ad-hoc workflows. Taken together, Social Media represent a good deal of the communication and information sharing means used by employees to manage their day-to-day work and provide a valuable means to connect process actors, stakeholders, and clients on a shared public platform. The business processes conducted with these tools often represent rather informal, non-routine processes that do not fit well with the top-down design of mass transaction processes that are often implemented in a Business Process Management System (BPMS).
As communication tools, Social Media can also be used to perform follow-up work on standard processes that are conducted in enterprise systems. For instance, employees might be quickly asking for support during a process via, e.g., their private Skype accounts. Having so much important activity occur outside and beyond the awareness of an enterprise application degrades the application’s effectiveness and management value. For this reason, companies nowadays tend to offer their employees tailored Social Media Platforms to exchange process-focused information (Bernstein 2000) within their organization. Preserving the “soft knowledge” of the overall process is of critical importance, in particular in the area of knowledge-intensive processes (Di Ciccio et al. 2015) and artful processes (Di Ciccio and Mecella 2013; Hill et al. 2006), that is, processes whose conduct and execution are heavily dependent on white-collar workers performing various interconnected knowledge-intensive decision making tasks.
On a meta level, Social Media are repositories of recent relevant facts that the authors want to make available to their colleagues, friends, or acquaintances. Those facts could enrich, specify, or glue together events that are recorded by BPMSs or other intra-organizational IT systems by embedding a process into contextual information, e.g., to explain things that could otherwise be less explicable, very often articulated in the words of the people involved directly.
Smart Devices
The introduction and proliferation of Smart Devices is an earth-shattering event that will profoundly change information processing and business models in our world. In 2017, Gartner Inc. (2017a) stated “[...] that 8.4 billion connected things will be used worldwide in 2017 [...], rising up to 20.4 billion by 2020. Total spending on endpoints and services [related to the Internet of Things, IoT] will reach almost $2 trillion in 2017.” That said, in the Gartner HypeCycle, the IoT is still viewed as being at the (first) peak and/or sliding into the trough of disillusionment (Gartner Inc. 2017b).
Smart Devices are equipped with sensors that can detect their own status as well as physical and digital events in their proximity. They have build-in hardware to store and process data to reason autonomously about the data they collect. They feature actuators that can perform physical actions inside a device and/or in a device’s proximity, while they have connectivity to transmit and receive digital data to/from their environment (Beverungen et al. 2019), i.e., from other devices and information systems, including Workflow Management Systems (WfMS) and Enterprise Systems.
Smart Devices are expected to profoundly transform various industries, including transport and logistics, healthcare, and manufacturing as well as the individual domains of living and social interactions (Atzori et al. 2010). As artificial actors in their own right, myriads of Smart Devices – including smart meters, smart vehicles, smart machines, smart phones, and others – will be starting, conducting, influencing, and ending business processes. Their build-in features will make Smart Devices partially autonomous, such that their actions cannot be controlled by one central authority, such as a business process engine. This shift of control means that business processes will be conducted a lot more decentralized, which will render top-down process engineering unfeasible, shifting control from build-time to run-time.
Moreover, the emergence of Smart Devices adds a physical perspective to business processes; while faulty processes in digital execution environments might be rolled-back, it might be impossible to undo physical actions that have been performed. Therefore, business processes that lead to physical actions performed by Smart Devices must be fail-safe to prevent adverse consequences of business processes.
First industrial business processes have been transforming to incorporate the benefits of Smart Devices, many of them stemming from the machine tools industries, in which production technology has been equipped with automation technology for a long time. Continuing this tradition, connecting a machines’ internal data processing capabilities with the “world outside” seemed like the next logical step, such that many current cases and prospects (Atzori et al. 2010; Perera et al. 2010) focus on sensing events in the field and taking these events up in business processes. For instance, Oracle reports a case in which a smart equipment senses outages proactively – based on acquiring data on themselves and on their environment – and reports the outages as events to remote information systems (Acharya 2015). These information systems listen for events and start the execution of pre-defined business processes (for instance, maintenance processes aimed at fixing the equipment) as soon as these events have been thrown.
Another case that utilizes Smart Devices to perform physical actions is situated in Hamburg, where “300 roadway sensors were installed by the Port Authority in order to monitor, control and manage roadways traffic” (Ferretti and Schiavone 2016, p. 278). For instance, since movable bridges are being opened on arrival of a ship, the road traffic in the port can be diverted to alternative routes now. In addition, the “system also calculates the weight of vehicles in order to establish the volume of traffic on the 140 bridges available in the port for trucks and trains and provide useful information for the design, maintenance and restructuring of these infrastructures” (Ferretti and Schiavone 2016, p. 279), to improve the port’s “integration with customers, reduce direct contacts and formal information exchanges with them and, finally, made easier and shorter their decision-making process” (Ferretti and Schiavone 2016, p. 279).
Big Data Analytics
Increasing amounts of data have been recorded for decades now (Hilbert and Lopez 2011), many of them generated by the trends for Social Computing and Smart Devices. This development is often referred to as Big Data, which in general means that each of the “four V’s” is at play: Volume, Velocity (data grow quickly), Variety (data are heterogeneous), and Veracity (data quality varies). Big data as such does not always refer to large datasets, but could also indicate small but complex datasets.
In general, data are increasingly collected for general purposes and do not refer to a single goal or type of analysis. The main challenge is to make sense of the available data, using the right data and analysis techniques. In recent years, the field of Data Science emerged, which is an amalgamation of different sub-disciplines (van der Aalst and Damiani 2015): statistics, data mining, machine learning, process mining, stochastics, databases, algorithms, large scale distributed computing, visualization and visual analytics, behavioral and social sciences, industrial engineering, privacy and security, and ethics. Of these areas, process mining bridges the gap between big data and data science to BPM.
Process mining answers crucial BPM questions, based on analyzing data from event logs. An event log contains a collection of events, where each event corresponds to: a case or process instance (e.g., an order number), an activity (e.g., evaluate request), a timestamp to indicate when the activity was executed, and additional (optional) attributes, such as the resource executing the corresponding event, or the type of event (van der Aalst and Damiani 2015).
Based on the data provided in the event log, process mining covers three main aspects: discovery of a process model (e.g., BPMN model or Petri net) based on event data; conformance checking of event data with respect to a provided (or discovered) process model; and enhancement of a process model by using event data to project, for instance, time information on the process model in order to analyze the performance of the business process.
Extending the conventional approach to mine processes based on event logs, the analysis of Big Data allows putting data on business processes into a context of other events that are related to a process. These additional data might, e.g., be provided on Social Media or by Smart Devices, as sources of data that might extend, complement, or even contradict data stored in BPMSs. A crucial prerequisite for making these data usable is to assure data quality and an adequate degree of granularity (e.g., consistent process IDs), such that the data can be mapped to process data supplied in event logs.
Within our project, we investigated how contextual information about process instances and activities is causally related to process performance over time. For example, the resource executing a particular activity in the process can influence the overall case duration and/or quality, since more or less rework is required. Another question is how different schedules for different resources can have an influence on the waiting time for activities performed by those resources. This, in turn, can affect the total duration of a process.
Another example is the analysis of health care event data in order to identify how patients are treated in a health care organization. Questions like “what is the most common treatment process’, “among which persons are handovers performed in an organization”, or “how efficient are processes in a hospital” can be answered using health-care event data, as has been done for a Dutch hospital (Mans et al. 2009). However, the issue is that disease treatment is not structured, despite clinical guidelines and pathways, due to the combinations of diseases, patient characteristics and variability in medical staff. Providing insights into these processes, using the recorded event data, can result in re-designing and improving the business processes.
Real-Time Computing
Recent advances in data processing, allowing for higher data volumes due to distribution, have enabled the development of technologies that are capable of processing a huge amount of information in real-time. This means that organizations can leverage this information instantly and take immediate action to adapt operational processes and corporate strategies to the ever-accelerating pace of business. Note that when we talk about real-time, we do not refer to the classical meaning of real-time systems in which tasks have hard deadlines and timing faults may cause catastrophic consequences (e.g. car automated safety systems) (Stankovic 1988). Instead, in this context Real-Time Computing refers to the so-called near real-time, in which the goal is to minimize latency between the event and its processing so that the user gets up-to-date information and can access the information whenever required.
Amongst the technologies that have fostered the use of Real-Time Computing, we highlight four of them with a strong impact in a business context. Complex event processing (CEP) enables filtering, composition, aggregation and pattern-detection of events that come from multiple sources, such as customer orders or social media posts (Cugola and Margara 2012). In-memory analytics involves the use of Random Access Memory (RAM) to store and analyze data, in contrast to traditional analytics in which data are stored on disks. This results in significant performance gains that allow business users to experiment with customer data in real-time and hence, to make timely decisions (Acker et al. 2011). Big data stream analytics enable the real-time processing of streams of data that have high volume and velocity by relying on parallelization platforms like Apache Spark Streaming (Zaharia et al. 2013). Finally, data stream mining performs traditional data mining techniques with continuous rapid data records. This includes techniques that can produce acceptable approximate mining results to cope with the high data rate of data streams as well as capturing the changes of data mining results coming from the evolving nature of data streams (Maimon and Rokach 2005).
These Real-Time Computing technologies provide BPM with the necessary tools to leverage intelligence instantly and make evidence-based timely decisions. This means that the traditional division of on-line transaction processing (OLTP) and on-line analytical processing (OLAP) can be overcome, making real-time process execution viable. Doing so is critical in a digitized and globalized environment in which organizations must adjust their processes at maximum speed and, at the same time, they have to make sure that their decisions are based on proper data and analytics. Connecting with Social Media and Smart Devices, this implies that business processes can be started, conducted, influenced, and stopped from outside classic BPMSs.
There are many different situations in which real-time computing brings clear advantages to BPM. For instance, real-time business activity monitoring can support decision-making to react faster to different situations. For example, a movie streaming service company tracks instantly which films are most popular among its customer segments so that their content team knows which films they should promote (Oxford Economics 2011), or an airline company that uses real-time information to manage seat availability for its 2000 daily flights with the goal to put as many travellers on board as possible (Oxford Economics 2011). Another case in which Real-Time Computing brings significant advantages is the immediate detection of non-compliance situations or fraud. For instance, a payment platform leverages big data stream analytics to detect fraudulent credit card payments (Li 2017).