Summary: This blog post discusses the importance of designing digital experiences with the user in mind. With the rise of cutting edge technologies such as artificial intelligence, reality and virtual reality is beginning to blur the lines between our digital and real worlds. Integrating data science with service design is crucial for developing digital experiences that seamlessly blend into people’s daily routines and have a positive impact on our communities.
Table of Contents
- Data and Design Make a Whole
- Data-Informed Design Thinking
- Design-Driven Data Science
- Facing the Double-Edged Sword of Data
Our human experience of the world is constantly evolving. Digital applications have had a profound impact on our lives, changing the way we communicate, the way we work and even how we make connections and build relationships with other human beings. The rise of cutting edge technologies such as artificial intelligence, augmented reality and virtual reality is beginning to blur the lines between our digital and real worlds. Combining Service Design and data science is the key to creating digital experiences that seamlessly integrate into people’s day-to-day activities and positively impact our societies.
Data and Design Make a Whole
When it comes to quantitative and qualitative analysis, the whole is greater than the sum of its parts. While many of us were raised to believe that “numbers don’t lie”, the reality is that it is easy to manipulate statistical analysis and many people often do so. There is a world of podcasts, books and courses dedicated to debunking deceptive research practices. Analyzing numerical data in a vacuum will not lead to uncovering a universal truth.
A qualitative analysis that involves in-depth interviews and observations of human interactions, can provide a more holistic understanding of the social context, nuances and underlying perceptions and human motivations of complex problems. Combining qualitative and quantitative analysis can lead to a more comprehensive understanding of a complex problem and ultimately provide better insights for decision makers.
Data-Informed Design Thinking
Design thinking is a very effective approach to building human-centric products and services. Design thinking begins with gaining a deep understanding and developing empathy for users and their needs. It embraces a fail-fast and iterative process to quickly validate and adapt ideas.
The design thinking process can uncover root problems that need to be addressed as well as ideas for potential solutions. For example, churn is a common problem in subscription based services. User research may find that a poor onboarding experience causes members to churn before the end of a trial period.
Data science can validate and reinforce the research findings by analyzing uncovered pain points in a wider set of data, validating causal relationships between pain points and expected behaviors, identifying confounding factors, and modeling outcomes of potential solutions. In our churn example, statistical analysis can validate how frequently churn is happening before the end of the trial period, which groups are mostly affected and validate the relationships between poor onboarding and churn.
The quantitative data analysis on a larger and statistically relevant set of data will complement the qualitative human analysis on a small group of people to create new products and services that are widely accepted and desired.
Design-Driven Data Science
Data science and statistical analysis are the foundation of many AI products. Six years ago, most organizations experimenting with AI were focused on leveraging their data assets to find patterns and predict outcomes. Data scientists worked in silos to build very accurate predictive models. It is no surprise that very few of these models ever passed the experimental stage to be deployed in a live- real world application.
AI products, like all data products, should be designed to improve the human experience. Beginning the creation of a data product with design thinking will set data scientists and other team members on a better path to success. Data scientists can benefit from a clear understanding of the problem, how humans will interact with the solution and what are the target business outcomes of the model. The human-machine collaboration and operationalization of a model need to be considered in the early stages of design.
Referring to our previous example of churn, data scientists may build models to predict members that are most likely to cancel their plans before their trial period, and recommend personalized retention campaigns. It is essential to understand who will use this information (e.g. Customer Service, Sales or Marketing), when they need to access it (e.g. right after onboarding or one week before the trial period ends), and how they need to consume information (e.g. a report, CRM or risk score). Design thinking provides a framework for crafting a solution that fits the needs of the people it is meant to serve.
Following a design-driven approach for development and data science results in products that provide the information people need, when they need it, and in a way they can easily understand.
Facing the Double-Edged Sword of DataThe amount of data being collected today is incomprehensible to the human brain. The world stores zettabytes of data from location, purchase history and interactions, to images, audio and video. Extensive data collection creates a double-edged sword that increases efficiency, enhances public safety and provides more insights to humanity on its “good side.” Its “dark side” carries potential for adverse effects such as privacy invasion, information misuses, and racial profiling. Building human-centric products requires carefully examining the data being collected and its intended uses and designing safeguards to prevent adverse unintended consequences. Designers and domain experts should have an active role in the building of a model. They should be bakers of the cake. In collaboration with data scientists and machine learning engineers they should:
- Select the ingredients: Select data points and features that are relevant to the problem
- Understand the recipe: Be able to interpret the model and understand how it works
- Taste before serving: Validate that the results are desirable and will not cause adverse impacts on humans