The Data Product Conundrum

Let’s face it, as BI experts we are no masters of the data product mentality.? For the most part building products is hard for even the product team to do, so how do we get better at it??? I thought about and decided instead of talking about how to build them, how about we get real about how not to build them.? We all can talk at length on what went wrong so let’s dive in.? Data products can fail for various reasons, often rooted in strategy, execution, and alignment with user needs. Here are some key reasons why data products fail, along with examples:?

Lack of Clear Purpose and User Understanding?

Example: Google Flu Trends?

Google Flu Trends aimed to predict flu outbreaks using search queries. However, it failed to accurately model the complexity of human behavior and health data, leading to significant inaccuracies.

Lesson: Without a deep understanding of the problem and user needs, data products can produce unreliable results that undermine their credibility.? The dimensions for a model this expansive can be many, getting them exactly right well didn’t happen here.?

Poor Data Quality and Governance?

Example: NHS Care.data program?

The NHS Care.data program was an ambitious initiative aimed at integrating patient data from general practitioners (GPs) across England into a centralized database. The goal was to create a comprehensive data product that could improve healthcare delivery, support research, and drive policy decisions through extensive data analysis.?

Lesson: data from many GPs was inconsistent with differences in coding and patient information, database had data that was incomplete and unreliable.? Lack of data governance policies around data ownership and access control created distrust amongst stakeholders.? Finally a lack of communication from the Governance board meant a lack of knowledge and eventual skepticism.??

Inadequate User Experience and Accessibility?

Example: Microsoft Tay?

Tay, an AI chatbot, interacted with users on Twitter. It quickly learned and started posting inappropriate tweets due to poor safeguards.?

Lesson: Failing to design for user experience and implement necessary safeguards can result in negative user interactions and product failure.

Scalability and Performance Issues

Example: HealthCare.gov Launch

The initial launch of HealthCare.gov faced technical issues, including slow response times and system crashes, preventing users from enrolling in health insurance.

Lesson: Lack of scalability and performance optimization can lead to system failures, especially under high demand.?

Misunderstanding the Business Objectives?

Example: Target’s Predictive Analytics

Target used predictive analytics to identify pregnant customers based on shopping habits and sent targeted ads. This led to privacy concerns and backlash when sensitive information was inadvertently revealed.?

Lesson: Data products must align with business objectives and consider the ethical implications to avoid negative repercussions.? To be fair I think this example is of something you might want to avoid but likely do not anticipate.?

Lack of Iterative Development and Adaptation?

Example: Nokia’s Mobile Mapping Service?

Nokia’s attempt to compete with Google Maps failed due to a lack of iterative development and responsiveness to market needs.

Lesson: Without iterative development and the ability to adapt to user feedback and market changes, data products can become obsolete quickly.?

Inadequate Collaboration and Cross-Functional Integration

Example: Quibi Streaming Service

Quibi, a mobile-first streaming service, failed to gain traction partly due to a lack of integration and collaboration between technology and content teams, leading to a disconnect with user expectations.

Lesson: Successful data products require collaboration across different teams to ensure alignment and integration of various aspects of the product.?

Overreliance on Technology without Business Context

Example: IBM Watson for Oncology

IBM Watson for Oncology promised to revolutionize cancer treatment with AI but struggled because the AI recommendations often lacked the nuanced understanding required for complex medical decisions.

Lesson: Overreliance on technology without deep business and domain context can lead to solutions that don’t meet real-world needs.?

Summary?

Data products fail for reasons such as lack of clear purpose, poor data quality, inadequate user experience, scalability issues, misalignment with business objectives, lack of iterative development, inadequate collaboration, and overreliance on technology. Successful data products require a deep understanding of user needs, robust data governance, iterative development, cross-functional collaboration, and alignment with business goals. By learning from past failures, organizations can better navigate the complexities of developing and deploying data products.