Combine advanced analytics with design to identify the right problem, and ensure your solution leads to the desired outcomes

An abbreviated version of this article has been published on the QuantumBlack blog.

When we recently held a 24h hackathon aimed at helping an NGO fight human trafficking, the composition of the competing teams could not have been more different: There were a number of teams with a heavy background in analytics consulting. And then there was my team: One data engineer, and three designers. The goal of the event was to generate insights from data provided by the client. The game was on. The client brief was the nail. Each team had its hammer.

It might have been hard to see what such a design heavy group had to do with what seemed like an obvious data analytics challenge, but we did not despair. Instead, the challenge allowed us to illustrate some valuable lessons about how combining data- and design-driven approaches can generate unique insights that previously had been overlooked. The key to getting there was in looking for the stories that came to the surface at the intersection of analytics and ethnography.

Ignore the hammer - Start with the problem, not the data

One reason for the success of Design Thinking is that it is solution agnostic, which lets problem solvers explore a much broader solution space than when using other approaches such as hypothesis-driven problem-solving. This still holds true when working in a data-centric environment. When we are asked to look for the answer to a question in data, we are obligated to first make sure the question is the right one. This is often considered an additional drain on resources on any project, especially when Design is only brought into a project when the team is already in delivery mode. However, when brought in early, Design helps drive the problem definition process together with the business. Potentially while others on the team are occupied with logistics - especially the infamous step of ‘getting the data’. This way, once a team hones in on a problem to work on, there can be certainty that solving it will actually lead to impact for business and users.

In our project room, we made sure the data was available and usable for when we needed it down the line, but then left it at that. Most of us didn't even take a peek at the data dictionary for the longest time so as to not bias our thinking, leading to the subconscious elimination of potentially rich research areas. Instead, we started with two familiar questions: What is the underlying problem really? Whose problem are we actually solving? Trying to answer these questions became the actual preliminary objective which sent us on a journey of discovery along the first slope of the famous double diamond.

Use data as a constraint

There are many tools at disposal to define the problem space and target personas: Stakeholder mapping, service blueprints, how-might-we's, five whys, expert interviews via video conference, and lots and lots of Post-Its (come to think of it, you really can fit an awful lot of discovery into one evening - kick-off for us was at 4 pm). Where an analytics engagement differs from a different type of transformation project is that the focus on specific data and information - and the flow thereof - provides additional constraints; and creativity loves constraints.

Think about how this realisation influences any design activities. Service blueprints gain a data layer, interview questions on this topic are added to the script, focus areas and ideas are prioritised not only by business, technical, and human constraints but by those related to data as well. At this stage, data becomes an enabler to drive research and ideas forward and adds an additional lens to focus on what can be achieved, and on what's missing.

Look for human stories in the data

Around the middle of the last century, the US Air Force was struggling with an unexplainable series of crashing pilots. Pilots could not maintain control of their newly designed planes. They later discovered that the reason for this problem was that, when designing the cockpits, they had done so by using averages of the various dimensions of the human body - rendering cockpits extremely unergonomic for many individuals. Once this was discovered, they changed cockpits to be adjustable to individual pilots and the crashes stopped.

When faced with data about humans, the risk of de-humanisation is very real. A scatter plot here, a bell curve there, and before you know it you have made generalised statements that don't really hold true for anyone. Instead, think about what can be learned by looking at data points as humans and try to imagine what they might mean in the real world. Often, you will find elements of human stories which can be used as research input and investigated further. Go through the list of key parameters and ask yourself what it might mean for a person if <metric> was <x>. Further, if you do this for a series of metrics, what story does this tell you about an individual? What you might start to see coming from this is the foundation of data infused personas that can be built on with further research.

During our hackathon to combat human trafficking, the distribution of survivor's ages gave us the most striking and tragic example of such a story. When looking at correlations between age and effectiveness of mitigating interventions, we could see that the youngest victims were only a few months of age. Not only is this a horrible insight, but it also informed our research as we now knew to have a very different set of conversations with our client that would include the special need of the youngest victims.

Listen to stories about data

While you are centering some of your research around data and people’s interaction with it, their data needs, and so on, pay special attention to where it is missing. While this might not be hugely relevant for the analytics engagement at hand, it is invaluable when defining a data strategy moving forward.

By shining a light on what else is presently being worked on or thought about at an organisation in respect to data, you can build a road map of intersections of various initiatives, leveraging connections rather than building yet another thing in complete isolation.

During research, both these points lend themselves to be captured in a service blueprint or process map and take centre stage when designing a multi-touch-point intervention down the line.

On the first evening of the hackathon, the client was already impressed with the range of questions we had come up with, saying that this half day of our work had already yielded exciting and very different insights, to those that a purely analytics lead team had found after two months of work. They were also excited to get working on a number of them - eventually leading to a road map proposal. We further learned about ongoing efforts seeking to enable the collection of entirely new data, and were prepared to develop a strategy that would make full use of (but not rely on) these new data streams, once they were ready.

During the hackathon, one might have been worried that all of these design activities wasted a lot of time before our team actually got started on ‘the real work'. However, by the time we had generated a range of ethnographic, data-guided insights and set aim at a specific high-value business problem, other teams were neck-deep in the data, still broadly exploring what was available, some still struggling to get their environment set up on their laptops. Having arrived at an idea for a concept that would let us illustrate those insights, we felt good about going home and getting some sleep before starting fresh the next morning.

Drive actions through insights

The most advanced analytics outcomes are worthless if they don’t find adoption and drive actions that lead to outcomes. Indeed, as the tech around analytics and AI becomes more widely understood and moves towards the mainstream of businesses, it is model adoption and other human factors that surface as key blockers - and opportunities. It is thus vital to understand a user’s context, needs - their own stories, if you will - and deliver analytics insights within that narrative in a way that empowers them to act. The work of an outcome focused analytics team can not be considered done at the point of delivering a model.

By applying Design Thinking and Human Centred AI methodologies, teams can invent compelling solutions that help users understand the real implications of analytics insights and gives them the right information in the right context. It is worth noting that this may not require the most advanced technology.

For our project, we created a concept that illustrated the future state of an entire user journey during which a case worker came to interact with a range of interventions to help them do their best work - some of which were low-tech and paper based. Crucially, this was not a blue sky art-of-the-possible concept, but one very much grounded in reality and executable today, while - at the same time - highlighting future potential.

Wrapping up

Of course, at the end of a hackathon during which many talented people from mixed backgrounds dealt with a very real and emotional topic, everyone was a winner. We went away not only happy about the potential for impact from all our newly generated insights, but also excited to pocket a picture-book case study of how a combination of different problem solving approaches can multiply a team’s effectiveness without being a drain on resources.

***

There are more humans living in slavery today, than there have ever before. If you’d like to help combat human rights abuses, donate to Amnesty International.