Many research reports on data usage echo the same message: Around 85% of analytics, big data, and AI projects will fail, despite massive investments of money. At Data Summit Connect 2021, Brian O'Neill, founder and principal, Designing for Analytics, explored why customers and employees are not engaging with data products and services and suggested that this is because, often, they weren't designed around user needs, wants, and behavior.
A "people first, technology second" approach can minimize the chance of failure and drive your analytics/AI/data/product team to create innovative and indispensable software solutions, he suggested.
Nobody actually wants analytics, machine learning, or AI, said O’Neill, noting the mission for data science and analytics teams should be “outcomes, not outputs.” A lot of the focus is creating output but what people want is outcomes and, as a result, there needs to be much more focus on the human experience.
The hurdle is adoption hurdle is adoption and innovative solutions are centered around humans and the way they work with technology, said O’Neill.
O’Neill gave examples of failed data scenarios such as an expensive fraud detection model that was ignored despite alerts that there were potential problems and a medical system that was designed to identify patients who will need to be readmitted but failed to consider how that information would be used to change processes in order to avoid that outcome. Data science and analytics teams must work on the last mile, he said, which is helping people to do their work and make improvements in their processes and decision making.
Key steps for getting better results from software development are:
- Human decision modeling before machine modeling: Teams need to be spend more time with their customers to understand their workflows and how data will be used.
- If model accuracy is all you care about, you might be happier in school. A relentless focus on increasing the accuracy of a model has nothing to do with delivering business value (although there are exceptions such as in healthcare, etc.)
- A data-driven solution is only that if it is used. If there is no usage, you have not provided value. You simply built something.
- Embrace product?not project— Especially with AI and machine learning, we need to be thinking long-term about products not projects. You are not building a one-off one-time solution when you're working with machine learning; most of the time you're building in a change, and that change needs to be treated as if it were almost a commercial product.
- Design is everyone ‘s job. Aim efforts at tasks and use cases. It is critical to understand what people need
O’Neill’s presentation was titled “Technically Right, Effectively Wrong: How to Avoid Creating the ML or Analytics Application No Customer Wants to Use.” More information about Data Summit Connect 2021 is available here.
Replays of all Data Summit Connect 2021 sessions will available to registered attendees for a limited time and many presenters have also made their slide decks available as well.