Data Quality: Where to Start & What to Do | Ville Satopää, Tom Redman, Tom Kunz & Kinda El Maarry

Why It Matters

High‑quality data determines whether AI delivers value or amplifies risk, directly affecting decision‑making, regulatory compliance, and competitive advantage.

Key Takeaways

• Data quality is the foundation of all AI outcomes.
• Biased or incomplete datasets produce biased AI predictions and decisions.
• Organizations often lack confidence in their data’s accuracy and completeness.
• Governance, ownership, and proactive quality measures drive reliable AI adoption.
• Real‑world examples show costly errors when data quality is ignored.

Summary

The webinar, hosted by NSEAD professor Ville Satopää, centered on why data quality is the linchpin of any successful AI initiative. Panelists—including former Shell data manager Tom Kunz, data‑governance leader Kinda El Maarry, and data‑quality champion Tom Redman—explored how AI systems inherit the strengths and flaws of their training data, making data hygiene a strategic priority. Key insights highlighted the classic "garbage in, garbage out" principle, illustrating how biased or incomplete datasets lead to skewed image‑recognition outputs, gender‑biased hiring filters at Amazon, and sub‑par generative‑AI results. The discussion also covered modern AI pipelines—foundation models, Retrieval‑Augmented Generation, fine‑tuning, and autonomous agents—emphasizing that each step is vulnerable to poor‑quality source data. Concrete examples underscored the urgency: a live poll of 186 alumni revealed that 41% would refuse to bet their bonus on the data driving critical decisions, while only four participants trusted their data fully. Shell’s 2009 SAP rollout sparked a data‑governance overhaul, and Redman noted salespeople spend up to 30% of their day cleaning leads, proving data quality is a daily operational pain point. The panel concluded that organizations must shift from reactive data cleanup to proactive governance, establishing clear ownership, robust quality metrics, and cultural buy‑in. Without these measures, AI deployments risk reinforcing biases, eroding trust, and delivering costly business errors, ultimately undermining digital transformation goals.

Original Description

Data quality is often framed as an IT issue. In practice, it is a business problem that affects cost, decision-making, performance and, in some settings, even safety. The challenge for many organisations is not recognising the issue but knowing how to get started in a way that leads to real change.

In this INSEAD webinar, Ville Satopää, Associate Professor of Technology and Operations Management (INSEAD), will lead a discussion with Tom Redman, President (Data Quality Solutions), Tom Kunz, Former Data Manager (Shell) and Kinda El Maarry, Head of Data (GotPhoto.com) on what it takes to improve data quality in practice. Drawing on examples from Shell, HelloFresh, GotPhoto and Chevron, they will explore how organisations can start with a concrete business problem, build buy-in between data producers and data users and take practical steps to improve data quality over time.

Why watch?

• Learn where to start on data quality without turning it into a large transformation programme

• Understand why data quality is a management issue, not just a technical one

• Hear concrete examples of how organisations have made progress in practice