Digital Economy Dispatch #08020th March 2022

Three Case Studies in Data-Driven Innovation

We are increasingly living in a data-driven world. With a vast collection of information sources being used to record the characteristics and behaviour of people, artifacts, and processes, we now see increasing use of algorithms fed by this data to produce a variety of insights from these sources. Across all aspects of our lives, decisions are being taken based on reviewing past events and observing the current state of the world around us. Every step we take has the possibility of generating a digital footprint that is recorded, amalgamated, analyzed, shared, and traded by someone somewhere.

At a societal level, all around us data is being gathered to understand the status and performance of everything from the operation of complex mechanical devices such as trains and aircraft engines to the behaviour of natural systems such as the movement of the oceans and current weather patterns.

At work, operating characteristics of the organization are being monitored with continuous collection of data recording variations in sales performance, changing customer demand, the contributions of each employee, market trends and fluctuations, and much more.

At an individual level, our online choices, actions, preferences, and desires are captured to be used as input to sophisticated algorithms that attempt to understand past behaviours, interpret current activities, and predict the future.

Understanding the Impact of Data

However, taking advantage of the opportunities afforded by this abundance of data raises several concerns. In particular, the deep impact of data-driven decision making demands a broad, inclusive approach to how we view the capture, management, and use of data. Unfortunately, as a key concept that has received increasing attention in recent years, data-driven innovation is often narrowly defined in terms of the technologies used in data acquisition or the details of the algorithms that form the heart of how the data is analysed. In any data- driven approach, this focus is necessary but not sufficient to grasp the full implication of the disruptive impact this shift creates.

In practice, a much broader perspective is essential. It has been found to be particularly useful to view data-driven innovation as the alignment of three critical perspectives:

• Desirability. Successful data-driven innovation solves a problem that matters to a client in a way the client can readily adopt into their operating environment. Whether these are internal or external clients, any innovation must address a problem that someone cares enough about to invest the (often considerable) time and effort to change behaviours, and provide a solution deployed in a way that the client can readily adopt at scale and in the timeframe in which it can have the required effect.

• Viability. Data-driven innovation must address a product or service need in a way that meets the constraints of the operating environment, organizational structures, and economic conditions in which it will be deployed. A myriad of issues must be considered including cost of production, strategic fit, impact on market and ecosystem, scalability of deployment, environmental sustainability, and maintainability.

• Feasibility. The opportunities afforded by data-driven innovation often seem to be infinite. However, any new idea must be able to be realized, introduced, operated, maintained, and disposed within current market and engineering constraints. Such concerns evolve rapidly with continual advances in all aspects of technology capability and performance, efficiencies in production processes, and breakthroughs in materials science. Furthermore, not only must it be practically possible to deliver a solution, it must also generate valued impact for potential clients and ensure aligned incentives for all stakeholders involved in its creation, management, and maintenance.

Consequently, success in data-driven innovation must be considered through an alignment across these three perspectives. Sounds good. But what does this mean in practice. Let’s look at 3 real examples of how this broader view of data-driven innovation can bring deeper understanding of how it changes our view of the world.

Case 1: Identifying Data-Driven Behaviour Changes and their Impacts

In a recent project for a large UK-based utility company, our activities involved gathering data from widely deployed in-home IoT devices providing readings of environmental conditions at 10 second intervals. Using this data, we were tasked with the problem of performing time series analysis to understand changes in behaviour that correlated with energy usage patterns in the home.

The initial hypothesis from the client was that early signals of changes in energy use could be used to optimize their energy trading practices to gain efficiencies, drive down costs, and smooth supply in response to variations in demand.

In carrying out this analysis using the Diagnostic Framework, we soon identified a number of critical concerns:

Feasibility: The company had out-of-date estimates for the amount of data that was to be ingested, stored, managed, and analyzed. Their technology infrastructure was unprepared for the level of cost and effort that would be required to perform this work. Consequently, the technical solutions being planned required significant enhancement to take advantage of the latest approaches to data handling, move processing closer to the data sources, and reduced data traffic across their data infrastructure.

Desirability: Existing approaches to connect with homeowners and tenants were based on infrequent manual interventions with simplified overview information based on quarterly billing periods. A different approach was required that brought more frequent updates to customers using data that could be readily used to drive actions to amend their energy use based on feedback they received.

Viability: Existing cost models for data processing and information management were based on 2-year-old estimates that proved to be wildly inaccurate based on rapidly changing technology availability and variations in energy prices. Executing those plans would have resulted in major financial losses for the company. Alternative approaches were proposed that involved simpler technology deployment options with clearer upgrade paths to deal with the volatility expected over the life of the project.

Case 2: Accelerating Data Flows Across Complex Ecosystems

A project carried out with a consortium of organizations in the food supply chain required investigation into the data flows across various players in this ecosystem to improve information flow and ensure alignment. A particular concern was the slow speed of response to food security alerts initiated by accidental or malicious introduction of food contaminants at some point in the chain.

The path from “farm to fork” involves a variety of steps crossing domain and sector boundaries within several industries. The project sponsor was interested in how data created at various stages in this process could be more effective in identifying, managing, and intervening to reduce the health issues that arise.

Feasibility: A variety of technological approaches are possible to support data flows across this scenario. However, much of the effort so far has been aimed at describing common data definitions to support meaningful synchronization between previously disparate systems. The resulting ontology will provide a base for future data sharing across all parties.

Desirability: A great deal of the effort has been spent on understanding the incentive models surrounding the different organizations in this data sharing ecosystem. Despite the technical capability for data sharing (even if cumbersome), little data sharing was occurring due to concerns about who had access to data, how that data would be used, proprietary issues of data ownership, etc. Agreement on an appropriate data sharing mechanisms (e.g., data trust) is now considered an essential part of any solution.

Viability: The costs for developing an integrated food safety system can be significant. Building models of different costing approaches has been undertaken. However, more significantly, there is uncertainty surrounding ownership of these cost models and who has responsibility for them. To assist with this, a number of cost recovery schemes are under investigation (e.g., outcome-based costing).

Case 3: Changing Digital Health Outcomes Using Wearable Technologies

Working with a large multi-national pharmaceutical company, we have been supporting efforts over several years to understand the use of wearable technologies in assisting with early disease detection and intervention. By using such devices, patients can receive early warning of changes in medical conditions, the efficacy of interventions such as drug treatments can be improved, and largescale epidemiological studies can be enhanced using massive datasets gathered from large populations over significant periods of time.

To enable these studies, there is wide availability of such devices from a variety of technology companies. These offer a multitude of data streams from simple step counters and heart rate monitors to sophisticated stress management capabilities. Many people use these devices every day to understand more about their daily activities.

However, for large pharmaceutical companies these devices introduce a number of challenges. While the data gathered may be of use, it is unclear whether the data is accurate, complete, relevant, and actionable. For this reason, our client was investing in its own specialist wearable device and required insight into its operation and utility.

Feasibility: The use of wearable sensor technology brings a number of challenges. Validating their use requires considerations about their design, robustness, and fitness for purpose. Data gathering, storage and management can be resource heavy and make excessive demand on the battery life. Re3viewing the device and the data streams allowed us to ensure that it operated as required across a range of usage scenarios.

Desirability: Much of the attention in our study was to consider how to gather more data form users more often. Encouraging users to adopt wearable technology and keep using it on a regular basis is often found to be challenging. After overcoming concerns about surveillance and misuse of personal information, the devices can be intrusive and awkward to configure. Furthermore, in practice much of the data generated from wearable devices is unusable. Gaps in data and erroneous recording of information is common due to misuse.

Viability: Many studies have considered the use of physiological data for diagnosis and treatment monitoring. A consistent finding is that patients often find them empowering while medical experts are concerned about the disruptive role they play in the health care system. In particular, they doubt their value in making better use of their time and ensuring better long-term outcomes for patients. Our studies were able to bring new insights into their economic impact on medical practitioners.

Digital Economy Tidbits

The Future of Software Engineering. Link.

I have been taking a look at this research agenda for “The Future of Software Engineering” and relating this to my previous work at the SEI over 15 years ago. I am interested to see how things have moved forward in those years, and what problems remain the same vs which are new. How much of what we believed important for largescale software development remains critical?

Software is vital to our country’s global competitiveness, innovation, and national security. It also ensures our modern standard of living and enables continued advances in defense, infrastructure, healthcare, commerce, education, and entertainment. As part of its work as a federally funded research and development center (FFRDC) focused on applied research to improve the practice of software engineering, the Carnegie Mellon University Software Engineering Institute led the community in creating this multi-year research and development vision and roadmap for engineering next-generation software-reliant systems.