Background: To solve complex analytics problems, data analysts often engage in a series of problem-solving activities, including extracting meaningful information from the data, synthesizing information to form higher-level concepts, creating a mental depiction of the problem, and imagining the impacts of possible scenarios. However, existing data analytics systems often focus exclusively on low-level data exploration and fail to effectively support these problem-solving activities. Consequently, the analysts have to contend with the gaps between the low-level technical analytic results and the high-level conceptual understandings which are required to solve the complex analytics problems. As the result, data analysts often find it is challenging to determine how the analytic results can be used to inform their decision-making and solve their real-world problems. In other words, existing data analytics systems often fail to deliver the actionable insight.

Objectives: The goal of this study is to develop a design of data analytics systems that can explicitly support the analyst´s problem-solving activities. This study theorized that when the problem-solving activities are supported, analysts are more likely to produce higher-level insights that can more readily inform decision-making. In order to achieve this design goal, this study asserts there is a need for 1) systematically understanding and defining actionable insight, 2) understanding the problem-solving activities and outcomes required to achieve actionable insight, and 3) proposing system features that can effectively support the problem-solving activities.

Method: This study employs design science research as the methodology to guide the overall design of this study. Through an integrated understanding of relevant theories, namely situation awareness (SA), sensemaking, and complex problem solving, this study conceptualizes actionable insight as a multi-component construct. Based on the way actionable insight is conceptualized, an explanatory framework is developed to provide a holistic explanation for the complex analytics task. This framework is specifically contextualized in the field of data analytics to explain the information processes, user behaviours, cognitive states, and information artefacts in different phases of a complex analytics task. More importantly, this explanatory framework provides systematic and theoretically-grounded design requirements which can be leveraged to improve user performance in the problem-solving activities.

A design framework was then developed to provide a set of prescriptive design principles for how the design requirements can be addressed. The design framework also acts as the blueprint for translating the conceptual design into tangible system features. To evaluate the effectiveness of the proposed design, a user study involving 30 participants was undertaken in a controlled setting. A prototype system was developed based on the design framework. The prototype system was evaluated against a conventional data analytics system in the user study. The user study requires the participants to analyse stock markets and to develop stock portfolios that will maximize the returns of investment.

Findings: This study categorizes the problem-solving activities into three phases: 1) data exploration, 2) information synthesis, and 3) knowledge actualization. The result shows that the proposed design is capable of enhancing the participants performance in the information synthesis and knowledge actualization phases, but not in the data exploration. Additionally, the proposed design was found to increase the perceived quality of the analytical result, implying that the results are more likely to be deployed into the physical world through decision making. Lastly, mixed results were found on the effects of the proposed design on actual decision performance. Specifically, the qualitative aspect of the participants decisions has been significantly improved, but the quantitative aspect of the decisions was not improved over the conventional data analytics system.

Overall, the findings suggest that the proposed design can support users to be more effective in integrating low-level technical findings into a holistic understanding of the problem situation and predicting and assessing the plausible impacts of the problem situation´s future development. Such understandings that are meaningful at the problem-solving level reduce the gaps between the low-level technical analysis and high-level understanding required for solving analytical tasks. In comparison with conventional data analytics systems, the proposed design enables the users to derive analytics results that can more readily be used to inform decision making and to solve the complex analytics problem. In other words, the proposed design improves the chance of deriving actionable insight from the data.

Conclusion: The contributions of this study include the explanatory framework which provides systematic understanding of analyst’ workflow, behaviours, and information needs, as well as other design considerations, in three different phases of the data analytics process. The framework can be used by data analytics researchers to understand the design considerations and requirements without repeatedly integrating the scattered knowledge from different domains. Additionally, the design framework can provide useful guidelines for practitioners to build the data analytics systems that can effectively support the problem-solving activities of the users. As a further implication, it is hoped that the proposed data analytics system can help practitioners to harness greater value out of their data, and thus can turn their IT investments into value-creation assets.