3 Ways To Drive Value When Design And Analytics Collide
Dan Feldman, Design Director, McKinsey, Sydney; Maksud Ibrahimov, Jr Principal Data Scientist, QuantumBlack, Melbourne; Justin Hevey, Expert Designer, McKinsey, Sydney; Cris Cunha, Partner, QuantumBlack, Perth and James Deighton, Partner, McKinsey, Melbourne
This is the final part of our three-part series, Exploring The Intersection Of Design & Advanced Analytics. Earlier articles examined the reasons why advanced analytics projects fail and honed in on one factor in particular, highlighting the need for collaboration between data scientists and designers. In this article we will provide practical advice on how multidisciplinary teams can thrive together throughout a project’s lifecycle.
Unlocking the value of analytics across an organisation requires building an environment where multidisciplinary teams thrive, where design experts collaborate with others to problem solve and see the stark differences between their disciplines as a strength. Our previous articles touched on the benefits this collaboration can bring. In this final installment, we wanted to share a selection of tips in how teams can facilitate this collaboration and bring the best out of design and data colleagues over the course of a project. This advice has been compiled during our experience in working across environments with a mixture of designers, data scientists, engineers and translators.
Build A Framework To Exchange Insights & Capabilities
One essential step that teams can take immediately is to begin regularly sharing insights across design and analytics as the disciplines are often attempting to solve similar problems from different angles. Building a knowledge sharing cadence into a project often leads to moments of clarity where insights shed light on new patterns, priorities, or purposes in the data. These can prove to be highly useful for rapidly refining the focus of teams and aligning the direction the project.
As an example, consider a recent analytics project overseeing the home loan churn process. A model’s impact may be influenced by understanding the reasons why people refinance their loans. Yet, there are many different reasons why someone would do this and data scientists may find it difficult to discern many of them from raw data alone. Looking at different customer journeys that lead to changing home loan providers uncovered new drivers for churn previously not considered in the model. A customer could be part of a separating couple, or someone who wants to build a swimming pool and therefore needs a larger loan. Perhaps the main driver behind the churn is the broker, persuading the customer to refinance in order to secure commission from another bank. Adding these new insights as features to the model increased predictive power and led to a more effective program.
Building structured opportunities to share insight and build capabilities benefits individuals and the wider team alike. It helps colleagues understand the value each of them brings, fosters better relationships, reduces the frequency and severity of misunderstandings and opens up chances for collaboration. Moreover, it’s necessary. By the end of the project, each specialism’s work will eventually need to integrate. By taking charge and creating a framework for collaboration, the disciplines work together to build knowledge, rather than simply going to their own respective corners to produce work independently.
Emphasise The Desirability Lens
Designers should take a proactive role in helping data scientists understand the user journey. Making time to walk colleagues through this early on will help generate better model later. If data scientists appreciate the end user’s mindset, needs and perspective, they can use this additional context to inform which hypotheses to test as features, reducing cycle times in feature engineering.
As an example, a previous project for a mining company focused on improving allocations of resources to a port, requiring individuals to make decisions based on analytics outputs. Bringing the analytics team along to research and map the user journey helped identify the priority capabilities the analytics models should contain and the order in which they should be analysed. This gave users everything required to make faster and better informed allocation decisions in their day-to-day work.
If designers walk data scientists through this customer journey and the various factors at play, the practitioners will better understand the process that led users to the tool in the first place and what they expect the tool to deliver. When this is appreciated in the early stage of a project, the model can best benefit from improved features and predictors.
Designers should be proactive in sharing this knowledge, but at the same time should not overlook the human resource available to make this process easier. Translators can assist with facilitating these walkthrough sessions — and in fact are probably the most important person a designer can align with, as they will become the glue that holds everything else together throughout a project’s lifecycle. Take the time to get to know them and sharing knowledge across the team later will be a far easier affair.
Harness Future State Experience Mapping
Future state mapping is a common tool used by designers which involves visualising how a user progresses through a process, typically used when considering how to develop interfaces. However, it can also be particularly valuable to involve a broader analytics team in this, as it helps everyone align on what value the solution generates for users and the end goal the solution is helping them achieve. Deploying this tool during the ideation stage will provide the team with a clear objective to work towards and cut down required iterations, uniting this mixture of specialisms under one goal.
Alongside team alignment, future state experience mapping can help reimagine processes when enhanced by models. The authors have observed the value this can add firsthand in a project tasked with optimise a specific subsection of an organisation’s sales process. The organisation initially wanted a model that would calculate the least expensive way for their salespeople to combine products that could then be sold, thereby widening margins.
By understanding the overall end goal and how model could change the future state of the user, practitioners were able to expand the impact of the model in a way that reimagined the entire sales process. New features were introduced that allowed salespeople to have live conversations with their customers about quality choices, saving the customers money and keeping them satisfied while still delivering a higher profit for the organisation’s sales process.
Thank you for reading this three-part series exploring the opportunities and challenges that arise when design meets data science. The insights and advice across these articles has been compiled while working across projects and diverse teams — however, we are always on the lookout for fresh perspectives and would like to hear any learnings, tips or tricks you have picked up.
Please do feel free to leave your own additions in the comments section and help contribute to an ever-more collaborative advanced analytics industry, where value is maximised when skillsets and perspectives collide.