The cost of skipping proper problem definition

All thoughts expressed in this post are the views Daniel Alvarez, Data Science Lead, UNICEF Office of Innovation

Data innovation projects typically do not fail because the technology is wrong. They fail because the right questions were never asked at the onset.

In the development industry, technical teams may spend large resources building solutions for problems they've never properly defined, for users they haven't understood. This can be costly: analysts build dashboards nobody opens, data scientists build machine learning models that answer the wrong questions with precision, and engineers build complex features nobody asked for. To address the disconnect between what teams build and what people actually need, we need design thinking.

From defining the innovation problem to articulating the vision

Defining the innovation problem

Critically, as it relates to innovation, there is a question about what the innovation problem is to be solved. An innovation problem is a problem that requires innovative approaches to address and potentially solve for. Design thinking is a method to understand what the innovation problem is all about.

This starts with identifying the scope of the problem by measuring its horizons and parameters: Where does the problem fall in the strategic versus tactical horizon? In the policy versus technology horizon? In the people versus resource horizon? Then, design thinking delves into the urgency and importance of the problem — critically asking what the negative consequences of not attempting to solve it might be. Could the current situation get even worse?

For UNICEF's innovation portfolio, asking these questions is critical given the lack of data on the ground from the people solutions are meant to serve. In the context of child protection and humanitarian response, failing to define problems properly could mean children that go unreached and escalating crises. An example of this is the QOWA initiative, whereby UNICEF provided technology-enabled education to children affected by the conflict in Iraq but failed to address real user needs.

Design thinking presents the discipline for stakeholders to reframe from a solutions-first approach to one that begs first generating better questions.

Landscape analysis: learning from existing solutions

From the innovation problem definition, design thinking proceeds to a landscape analysis unpacking existing solutions. It asks: in what existing ways is the problem being solved, and what can be learned from these approaches? For the data scientist, it is important to understand what the data practices are at every stage — what works well, and what fails.

Actors, needs, and incentives

An analysis of actors and needs follows. Understanding what other organizations and partners are involved shapes the problem and potential solutions. This entails not just identifying the actors, but understanding their influence and impact — not just who is acting, but what they do or want to do and what they need to achieve their objectives.

As a data scientist, I might want to know what actors are sourcing data, how they are collecting it, and what incentives they might have to continue collecting it in the future. This raises questions about what's moving actors forward and what's holding them back.

Reframing the challenge as an actionable opportunity

Subsequently, design thinking means thinking through needs and gaps and reframing the challenge as an actionable opportunity. The human-centered aspect of this is critical: as problem-solvers, we need to understand the needs and challenges of the people identified. A good problem statement is open-ended enough to encourage creative thought, yet focused enough to be actionable.

Articulating the project vision

Finally, design thinking leads to a comprehensive description of the vision of the intended project. This includes a big-picture explanation of the idea, what success looks like, and how it will be measured. We would need to know what people, processes, and technological components are required to make the vision a reality — culminating with mapping primary actors and their motivations, and an articulation of the process with timelines, milestones, and accountability frameworks. Can we chart out the first six months of the project?

Design thinking in practice at the public sector

Design thinking is critical for properly addressing the real-world data problems we see at UNICEF: a lack of health record linkages across borders, lack of precise impact forecasting to prepare for imminent natural hazard catastrophes, lack of groundwater boreholes mapping, and lack of geographic data measurement of poor air quality. In these cases, design thinking presents the discipline for stakeholders to reframe from a solutions-first approach to one that first generates better questions.

Innovation units in advanced governments around the world have taken to these design thinking types of exercises to identify and address problems in government. One such approach is the path analysis method followed by 18F in the US Federal Government, which involves a technical assessment of current approaches and user-focused analysis to discover ways to improve and make meaningful, adaptive change in how government agencies work and deliver value for the public. Examples of path analyses include determining the best technical solution for disability case processing within the Social Security Administration (see here), transitioning Environmental Protection Agency’s Clean Air Markets Division towards human-centered design practices and cloud hosting of applications (see here), and the application of a technology acceptance model to trust in artificial intelligence in the commercial aviation industry (see here).