2. Related Work
Innovation Management (without connection to DSR) is a well-analyzed scientific field starting as early as the 1930s [
22]. Considerable research on innovation processes and their organizational management has been carried out in the related field of innovation management of which [
14,
23,
24] give an overview and examples of project-oriented innovation processes. These findings are integrated into the innovation engineering method as state-of-the-art. However, none of them integrates the design of IS artifacts by scientific standards from the field of Information Systems. Ref. [
13] acknowledges these shortcomings and provides a framework for strategic positioning of intracompany research organizations, but does not focus on methods or engineering to be applied by these. Ref. [
25] has described scientifically an approach where multiple industry partners find consensus in a moderated academic platform about joint research projects that are then carried out in a consortium. However, the created artifacts are not linked inherently to the innovation processes of the respective companies.
Innovation processes have been implemented in corporate environments over past decades where especially stage-gate oriented idea-to-launch processes, as proposed by [
23], are well established and continue to persist with modifications also for Digital Innovation [
26,
27].
DI has become an area that is researched in various application domains. One of the most focused areas is business management and researchers interested in this field are investigating general innovation processes in the organization. They also concentrate on novel innovative services and product research, fostering the organization’s improvement. Business models, processes, and product innovations are also in focus when it comes to research in DI, so it has been further investigated how digital technologies can contribute to these areas for improving the capability and quality of innovation processes [
28,
29]. Ref. [
30] takes into account the digitalization of the financial sector and assesses emerging business models and technologies. Ref. [
31] explores supporting tools, namely data analytics, for innovation processes and states that it is a must for firms to utilize big data analytics to stay innovative. Ref. [
32] develops a process model which eases analyzing the impact of potential digital innovations on existing business models and helps generate a new one (the authors also follow a DSR approach as research methodology). Furthermore, numerous researchers, including [
33,
34,
35,
36,
37,
38], have assessed the digital innovation capability of organizations by focusing on how organizations can produce innovations using digital resources, considering the organizational focus on businesses. Notable studies by [
39,
40,
41,
42,
43,
44,
45] have proposed a framework for digital innovation management which emphasizes the need for management of actual process innovation using digital technologies. The works of [
26,
46] further elaborate on IT artifacts that can increasingly facilitate the implementation process of digital innovation frameworks.
Although these studies have been conducted with respect to DI, the vast majority of them do not include DSR concepts to synthesize their findings in terms of DI.
On the other hand, regarding the joint synergies of DSR and DI, recent studies have been cordially conducted to identify the commonalities and cooperation of these fields in recent years. DSR is generally accepted as a contributor to DI from a research and practical perspective since it is mainly concerned with the design and development of innovative artifacts. However, we can see that there is still a scarcity of research regarding DSR/DI synergies, and this limitation is also related to the general realization of how digital innovation may be supported by DSR. Reference [
47] sees the possibility of inclusion of design science into digital innovation research; however, it also suggests some open-ended questions regarding whether to keep these research streams separate or joint. The authors of [
48], Ref. [
15] have combined these areas into a framework for getting the maximum benefit with practical and theoretical means, by conducting studies regarding the combination of DI and DSR concepts and building conceptualized frameworks. Ref. [
48] in their paper state that “DSR in the IS field is, at its essence, about DI”. They further introduce a matrix approach to DI based on DSR and consider innovation and entrepreneurship theories while building it. The paper defines a DI-DSR matrix-based approach that relies on the Knowledge Innovation Matrix, which in turn expands on the four strategies “Invention; Advancement; Exaptation; and Exploitation”. The research here contributes to the DI–DSR relationship in a considerable way by delivering an understandable process model, allowing people with various entrepreneurial backgrounds to perceive the concept as well.
Digital Innovation in the context of IS comprises both the use of IS to support (or even trigger) innovation processes and also the design of IS as an outcome of the innovation process. For example, Ref. [
26] have focused on the importance of Digital Technologies as a change agent for the discipline of Innovation Management. For DSR in the area of DI, this means that both the innovation method and process and also the innovation outcomes are IS artifacts. In a recent approach, Ref. [
33] have structured the field of DI into six roles. The roles that are important for our approach are Role 1 (
Design of a DI Technical Artifact), Role 2 (
Design of an Artifact for Deployment and Use of a DI Artifact), Role 4 (
Development of Design Theories Surrounding a DI Artifact), and Role 5 (
Use of a DI Artifact as a Creativity Tool in the DSR Solution Process). As a whole, this paper primarily describes how the central method artifact is composed (Role 5). We devise a method with a process model for the DSR solution process and the resulting method artifact fulfills this role. However, the artifact itself produces an output of Roles 1 through 3, as we will also document during the case study and lessons learned for the example of an automatic speech recognition system.
Ref. [
15] has introduced a combination of state-of-the-art, practice-driven stage-gate-innovation processes such as [
23] with Design Science Research Processes. This combination provides an elegant way of combining practical relevance and scientific rigor as demanded, e.g., by [
7] – provided that the degree of innovation allows for research. We will build on this work.
Information Systems is not the only field where design is relevant. The sciences of the artificial as described, e.g., in [
4] have influenced science in a great number of seemingly different disciplines: from the already introduced engineering and Information Systems domains to social sciences, medicine, architecture, design thinking, industrial design, and many more. It might seem far-fetched to compare the design science research approach in this paper to the design approaches of these heterogeneous domains. However, specifically, Design Thinking Research [
49] and Design Research stemming from—but not limited to—industrial design [
50,
51] offer some interesting parallelisms and also differences that are noteworthy for the discussion as they are often related to creativity in innovation, but this work has not been primarily created for Information Systems. A Unified Innovation Process Model for Design Thinking has been introduced by [
52]. While our method—including the process model—covers the whole life cycle of an innovation project and also the systematic contribution to science, the Unified Innovation Process Model focuses specifically on the creative phases of the design of an artifact.
6. Evaluation by a Case Study: Tailored Call Center Process
We will demonstrate the applicability of the approach in a case study. The evaluation is structured as proposed by [
69] in the following paragraphs: Situation Faced, Action Taken, Results Achieved, and Lessons Learned.
Innovation Process. An incumbent telecommunications company faced the problem of having excessive operational costs in the call center when compared to the competition. The decision was taken to set up an innovation project to introduce cost savings by automating part of the work of call center agents by means of Automatic Speech Recognition (ASR), resulting in an Interactive Voice Response system (IVR) and chatbot. The company used commercial off-the-shelf modules but also had the ability to differentiate itself from the competition by developing new own modules. A project team was set up, led by the innovation process manager of the company’s internal innovation lab and business stakeholders. Stakeholders were experts from the unit that runs the call centers and the business owners within this company. The innovation process manager from the innovation lab had been continuously in contact with the stakeholders. Tools for interaction and creativity stem from Design Thinking and (industrial product) Design Research. Part of the core team were also Research Engineers, Information System Developers, and End Customers (i.e., lead users of the call centers). During the workshop, it turned out that a major pain for the stakeholders was the high costs per call due to the tedious manual interaction of live call center agents. As a matter of fact, the automation of manual process steps would have yielded a high benefit (a business case for efficiency). Likewise, new value-added services for target groups (gender and age-dependent) would have offered additional marketing and business opportunities at the call center customer front end.
Research Process. Recognition of non-verbal features such as age and gender beyond speech-to-text from a speech signal has been a topic that had only recently emerged at the time of the project, with no commercial-of-the-shelf recognizers available. It was a goal of this project to perform a classification of such non-verbal features in parallel to recognizers that convert speech to text in a call center, to make skill-based routing and market analyses in call centers possible. These features were not commercially available.
Innovation Process. According to the process description proposed in [
15] a Gate 1 proposal was put together with stakeholders. At this phase, Design Thinking workshops based on iterative user-centric and participatory design with a multi-disciplinary team were carried out. Researchers of ASR knew about new research on non-verbal speech recognition. Thus, the practical relevance of the project was ensured when passing the four gates. In this context emerged the idea to tailor the IVR call process flow according to the age and gender of the caller. This would enable the call center agents to save time by pre-classifying the caller and automating part of the dialog script. This would then result in saving effort, thus reducing the cost per call as human interaction is the most cost-intensive.
Subsequently, a Gate 2 proposal was prepared by submitting the required documentation, the Project Scheme. For the Gate 3 proposal, a full project plan (work packages, Gantt chart, effort calculation) and a Business Case were finalized. After Gate 3 the project was successfully carried out and a prototype was presented. At Gate 4, decisions were taken by the stakeholders after evaluating the prototype along with the corresponding artifacts, and the decision was made to transfer the prototype to production, which required further project work with respect to the scalability, reliability, and resilience of the system. The project was finally put into production and met all of the planned goals, including higher efficiency in the call center.
Research Process. Before the Gate 1 presentation, the researchers identified the research design problem of age and gender recognition in the domain of speech recognition. Knowledge of the literature yielded several approaches, but none of them were ready for immediate use on the problem. The second research question was how to best tailor the IVR dialog, given knowledge of age and gender, in order to achieve the desired goals. The practical relevance of the project was formally approved by the Gate 1 meeting.
Thus, the research design strategy followed two steps. First, the team contributed to science with the design and comparison of four different approaches for age and gender recognition and the subsequent comparative empirical evaluation in a laboratory experiment [
70] on the same speech database “SpeechDat” [
71]. SpeechDat is a spoken language resources database of labeled audio files of multilingual telephone speech. It contains phonetically balanced sentences uttered by speakers of different ages and genders. The recognition task was to differentiate seven groups by age and gender: children of 13 years and younger, young people between 14 and 19 years (male/female), adults between 20 and 64 years (male/female), and seniors.
The best-performing method was an adapted design based on an existing Parallel Phone Recognizer (PPR) [
72]. For easy tasks, its precision is comparable to human performance [
70]. PPR was originally developed to recognize languages (such as English, German, Hungarian, etc.), not gender and age. For the adoption of the task, seven different phone models were trained using exclusively those parts of the audio files that had been labeled as belonging to one of the seven distinct speaker groups (differing by age and gender as described above), respectively. Instead of using PPR with different phone models for languages, the researchers adopted its principles for phone models of the seven different age and gender groups. According to the classification from
Figure 4 the design adhered to the strategy
Increase Scope as the scope of PPR was increased from language identification by recognition of the age and gender of the speakers. For more details of the recognizer, please refer to System A in [
70].
In the second design step, the team combined commercially available recognizers that are used for speech-to-text tasks with our own PPR classifier for non-verbal speech used in parallel on the same speech signal. By the classification exhibited in
Figure 4 the chosen innovation strategy is
Combine because the innovation is based on a novel combination of existing designs.
For the presentation held at Gate 2, the design strategy was stable according to
Table 3.
The corresponding abstract templates form [
73] are employed accordingly, as visible in
Table 4:
For the Gate 3 presentation, initial scientific experiments were built, as well as a first sketch of the evaluation by empirical recognition rates, user interviews, and process simulation. Researchers also participated in preparing Gate 4 as their innovative artifact was built into a live system and they needed to evaluate it also scientifically. This evaluation is published in [
74].
Refs. [
70,
74] answer the question of how to build a system, and are as a matter of fact eligible to describe it as an ISDT for design and action [
8]. Many Engineering publications fulfill this criterion, which is not surprising as engineering is also classified as a science of the artificial by [
4].
The empirical evaluation by a laboratory experiment shows improvements in mean opinion scores of live users and average ratings of users when compared to a conventional routing [
74]. Yearly efficiency gains could be quantified at 42 Million Euros.
Although the new artifact was originally not built using a "formal" DSR process, it was possible to align all its contributions in detail to the phases
Problem Identification,
Solution Design, and
Evaluation, as well as our research process from
Section 4.1.2. In fact, all the activities of the researchers could be mapped to the phases of the DSR process according to its embedding in
Figure 2. The process and the results fulfill all seven guidelines of [
7]:
Design as an Artifact: method artifact;
Problem relevance: ensured during the Gate process;
Design Evaluation: both empirical and qualitative in a case study;
Research contributions: contribution to IS Design Theories (ISDTs) as in
Table 5, following the innovation strategies
Increase Scope and
Combine;
Research Rigor: demonstrated by accepted peer-reviewed scientific publications;
Design as a Search Process: search iteration during Gate-process, four age and gender solutions were identified and tested;
Communication of Research: presentation to stakeholders, publication of results.
The Integrated Innovation Strategies Framework. The outcome of the case study is that the innovation strategies are very well applicable and they smoothly integrate into the combined process artifact. On the one hand, they perform well in classifying the actual design work and thus help teams to better document the actual innovation steps. The IISF could be used for the analysis of a wealth of past engineering projects, making their results available for the DSR body of knowledge (
Table 6).
Innovation Strategies Validation
We have performed interviews with researchers from the Institute of Industry-Academia Innovation at Eötvös Loránd University (ELTE), Budapest. Interviews were conducted with academic researchers to whom we asked questions regarding the projects they had been actively working on. After gathering empirical data, we selected four projects to validate three innovation strategies among the ones we point out in this paper. We will further validate the remaining innovation strategies in an upcoming research paper accordingly.
Mini-Link, developed by Ericsson, is a radio unit for microwave transmission in radio transport networks. In-network telemetry systems connected to this product require a vast number of configuration files, depending on the usage scenarios. Steps were taken to process these files into databases to help developers, testers, and customer support to focus their work on development and testing, and to be able to give advice to the customers about how to configure the nodes (e.g., Ericsson customers could obtain useful feedback when they were upgrading software on their nodes). Customers could more confidently upgrade their software because of the visualized data on the predictions based on logs. On the other hand, processing of this data in a relational database management system is very slow and can be hard to query; storing this data takes lots of disk space as well.
Hereby the authors [
75] present a better way to store the data produced by these nodes in graph databases by using a NoSQL environment instead of a relational database. With this approach, it is possible to easily represent and visualize a network of machines in its bigger picture. As a result, these machines achieve much better efficiency in several aspects including time, storage, performance, etc., when evaluating insertions, querying time, and storage size. Technologies used in this task are Apache Spark for the processing of the data, HIVE for storing data, and Tableau for creating visualizations.
Here, the researchers improve the existing design for storing and querying vast amounts of data on relational databases and come up with an improved design that makes use of NoSQL databases within the organization. The design strategy here is Improve since, after identifying shortcomings in the standing design, the researchers have proposed an improved design that overcomes such shortcomings.
RefactorErl is a static source code analyzer and transformer tool, which has also been made open source. The aim of the project was to create a product to support Erlang developers in their daily code comprehension tasks. The usefulness of the product has been proven in industrial usage. The tool has an Erlang source code analyzer and transformer [
76] which is able to handle real-world code. According to statistics, it was applied successfully on more than 1.5MM LoC. There are several helpful features which include support to analyze macro constructs, storage and fast retrieval of analysis results, source code layout improvement, and comment preservation during transformations. The results from various deep semantic analyses are expressed through a user-based semantic query language which can help Erlang developers in debugging, program comprehension, identifying relationships between parts of the program, and so on. It helps understand legacy code, aiding software restructuring, and checking code complexity and quality. The tool is capable of shortening the learning time of newcomers, increasing code quality by reducing faults, and facilitating effective teamwork in various ways.
Prior to the initiation of this project, there was no such tool for Erlang. That is the reason why researchers at Eotvos Lorand University, Budapest, started to build one (though, similar other tools did exist at that time for other programming languages). Moreover, elements such as the semantic analyzer framework and the incremental analyzer of the changes of the source code are still unique components of RefactorErl, if compared to other tools built for other scopes/languages/programming paradigms. Incremental analysis helps with source code that contains millions of lines of code, which are difficult to inspect and comprehend: if the analysis is incrementally applied, a few minutes are enough to obtain results. In addition to that, features such as variable binding and static analysis on data flow and control flow [
77] are currently not provided by other alternatives as well.
As mentioned earlier, the idea of performing source code analysis is not new, and tools to achieve this task existed for other programming languages, but not for Erlang. Therefore, in this project, an Increase Scope strategy has been followed, since an existing design was already available for a different purpose and scope, and the researchers extended it to be used in a new one.
Internet of Eyes is an object detection system capable of detecting and recognizing moving objects and determining their three-dimensional spatial position through real-time processing of video streams from multiple cameras. Although the system moves large amounts of data over the network and uses high computing capacity, it does all this with very low latency thanks to a high-performance “Edge” server placed at the edge of the network. The system is used in a simulation environment in which the goal is to detect and avoid possible collisions with vehicles. This project started as an Ericsson-supported project at ELTE, Budapest. The aim here was to simulate and illustrate possible scenarios in which the low latency and high reliability of 5G are fundamental. At the time of this project, there was already much positive talk/opinion on the benefits of 5G, but not many use cases that would effectively demonstrate why the technology and its properties would represent a breakthrough both in industry and in the daily lives of users. This project was meant to boost or create common ground to sell Edge Computing technology to telecom companies, and possibly their clients. Here, the design strategy is Explore New because the issue was not yet solved and the researchers proposed a solution to it.
CodeChecker is an open-source project developed in close collaboration with Ericsson. The tool applies static analysis to find potential software bugs in programs written using the C/C++ programming language. There are several issues that are likely not caught by compilers; herewith, to eventually increase software quality, static analysis tools are significantly important. CodeChecker does not run the program as in testing, it solely performs a static analysis. There are several known users of CodeChecker, which also include developers from companies such as Apple, Google, Sony, and Samsung. The tool provides command line C/C++ Analysis, web-based report storage, and incremental analysis which works by considering only the changed files and their dependencies, provides false positive suppression, and visualization of results in the command line or on a static HTML web application to allow viewing discovered code defects with a streamlined and easy experience. It has also been improved to become an ecosystem-independent, web-based multiplatform tool, and Cross Translation Unit Analysis was implemented [
78], which helps to find more bugs. The followed design strategy is
Improve because the proposed design brings improvements over the shortcomings of a previous one.