CDO Matters Ep. 49 | The Role of Product Management in the Evolution of a Data Function
May 2, 2024
auto_awesome
The podcast discusses leveraging product management in the evolving world of data and analytics, exploring data products, data mesh, transitioning to dynamic data structures, and the shift towards domain-specific models. Emphasizing the need for CDOs to adapt to an AI-driven world of insights.
Leverage product management for AI-driven insights in evolving data landscape.
Utilize data mesh strategy for domain-centric and cross-business process analytics.
Develop domain-specific language models for accurate data science projects.
Deep dives
Dataception's Mission: Customer-Facing Analytics with Speed and Iteration
Dataception, founded by John Cook after stints at Databricks and PwC, focuses on bridging business and technology through rapid, iterative creation of customer-facing analytics. The core mission is delivering custom analytics to businesses quickly, emphasizing on data products and iterative development. The goal is to provide low-friction insights that avoid prolonged project timelines and multimillion-pound costs.
Defining Data Products: Crossing Business and Analytical Domains
John Cook discusses how data products are about catering to cross-domain use cases, blending business process elements from multiple areas. He emphasizes the importance of understanding and delivering what the business truly requires, ensuring a transition from modeling mindset to end-to-end solution delivery. The focus lies on combining data sets from various business segments, crafting a holistic solution tailored to specific business needs.
Data Mesh Vision: Domain-Centricity and Governance Challenges
The discussion delves into the concept of data mesh as a strategy for fostering data sharing between domains within organizations. John Cook underscores the need for solutions that enable domain teams to build and deploy their components, emphasizing the domain-centric and cross-business process analytics approach. Governance emerges as a critical aspect to manage multiple sets of rules and policies, ensuring interoperability and efficient cross-domain insights.
The Importance of Domain-specific Language Models in Data Science
Developing smaller and more focused language models that are specific to certain domains can lead to increased accuracy and efficiency in data science projects. By moving away from large transformer-based models and utilizing smaller models tailored to individual tasks, organizations can achieve better results with less computational resources. These domain-specific models allow for a deeper understanding of complex business processes and data, enhancing accuracy and reducing training costs.
Bridging the Gap Between Traditional Data Models and New AI Approaches
There is a growing need to bridge the gap between traditional relational data models and emerging AI-driven approaches like large language models (LLMs). Exploring ways to transition from rows and columns to more narrative-based data representations can enhance the capture of contextual information and improve the understanding of complex business processes. Embracing multi-dimensional thinking and adopting domain-specific, task-focused models can pave the way for more effective data analytics and insights in the evolving data landscape.
The explosion of LLMs and AI is causing a tectonic shift in the world of data and analytics, and on this episode of CDO Matters, Malcolm and Jon Cooke discuss how CDOs can leverage the discipline of product management to best capitalize on these massive changes.
From data products, to the data mesh, and beyond – Malcolm and Jon enjoy a lively discussion on the many evolving attributes that will increasingly define a world focused less on data (and datasets), and more on knowledge and business insights.
If you’re a CDO and you’re interested in learning more about how your operating model will necessarily need to shift over the coming years to adapt to a new AI-driven world of insights, then this episode is a must-listen.