The original article called for seventeen names while giving uneven attention to fewer individuals and repeating broad claims about the future of data. Here the full seventeen are presented through the particular question each can help a learner investigate. Readers should judge claims by the published work and evidence, not by a title or an asserted salary premium.

Strategy, business value and communicating uncertainty

Bernard Marr helped popularize clear explanations of business analytics and the links between data strategy and organizational outcomes. His work is a useful starting point for asking whether an analytics investment changes a decision or merely creates another dashboard. Without an owner, a measurable decision and a feedback loop, the dashboard is unlikely to have a durable purpose.

Ronald van Loon was a prominent advocate for business applications of analytics and data-driven transformation. A reader can use that perspective to examine how a strategy connects data engineering, people and governance. Adoption is not the same as organizational value: a model should be evaluated against a real process and a defined baseline.

Merv Adrian contributed industry analysis of data-management platforms and architectures. His work is useful for comparing the constraints of storage, processing and governance rather than assuming every new engine makes earlier databases obsolete. Platform choices should follow workload and service-level needs.

Doug Laney emphasized treating information as an organizational asset. That perspective raises hard questions: who owns critical datasets, how is quality measured, and how should privacy or retention obligations limit reuse? Not every valuable dataset should be monetized or shared broadly.

Cindi Howson addressed business intelligence and the practical democratization of analytics. Self-service reporting can reduce delays, but uncontrolled versions of revenue or customer definitions can produce conflicting decisions. A governed semantic model and visible metric definitions are often more useful than an additional chart type.

Research, statistical thinking and data-science practice

Kirk Borne brought the perspective of scientific analysis and large datasets to wider data-science conversations. His work is a reminder that a convincing visualization must still be supported by sampling, measurement and validation. Scale amplifies both correct insights and systematic error.

Gregory Piatetsky built a long-running forum for data-mining and analytics discussion through KDnuggets. For a learner, the lesson is to compare methods and look for reproducible results rather than declaring a single technique the best for every dataset.

Vincent Granville was associated with data-science publishing and community building. Independent explanations and worked examples can make technical ideas more accessible, but their assumptions should remain visible. A reader should be able to reproduce a result before extending it.

Hadley Wickham influenced how analysts manipulate and visualize data through R tools and the tidy-data perspective. A consistent analytical workflow is valuable because it makes transformation steps easier to inspect and repeat. Reproducibility is a practical safeguard against silent errors.

Cathy O’Neil made the social risks of opaque and poorly governed models more visible. Her work encourages analysts to test whether a model’s errors fall unevenly across populations and whether affected people can challenge outcomes. Predictive accuracy alone does not settle questions of fairness or accountability.

Machine learning, education and responsible deployment

Hilary Mason contributed to industry discussions on applied machine learning and experimentation. A useful takeaway is that model development starts with a testable problem and sound data, not simply a fashionable algorithm. Evaluation must reflect how the output will be used.

Andrew Ng helped bring machine-learning education to a broad audience. His practical influence reminds beginners that sound fundamentals, error analysis and iterative experiments matter more than collecting model names.

Cassie Kozyrkov popularized decision intelligence and the importance of clearly stating what decision an analysis supports. A forecast may be numerically accurate but irrelevant if it cannot affect an action. Analysts should specify the available choices and the cost of errors before selecting metrics.

Rachel Thomas helped advance accessible machine-learning education and scrutiny of technology’s social effects. Her work is a reminder that deployment includes user context, impact assessment and ongoing monitoring, not only training a high-performing model.

Lillian Pierson focused on making data skills teachable to professionals from varied backgrounds. That is valuable because real teams need analysts who can explain their methods to domain specialists, not only experts who can write code.

Data platforms, leadership and inclusive professional communities

Joanna Schloss contributed to practical data and analytics discussions around management and integration. A useful theme is the relationship between operational data quality and downstream analysis. A sophisticated dashboard cannot compensate for inconsistent definitions or missing lineage.

Jen Stirrup was a visible advocate for business intelligence, responsible data practice and wider participation in technical communities. Her example encourages teams to include diverse domain perspectives when defining questions and evaluating results. Data quality is partly an organizational problem involving communication, ownership and incentives.

How to use the 2019 reading list

These voices differ in discipline, methodology and audience. To learn effectively, pick a problem such as customer retention, service reliability or quality measurement and read several perspectives on it. Pair an engineering view of ingestion and modeling with a statistical view of uncertainty, a decision-making view of usefulness and an ethics view of impact. Compare their assumptions and attempt a small reproducible analysis.

A 2019 influencer list cannot tell readers who is active or worth following in October 2026 without a new, separately verified review of each person’s publications. Its lasting value is a wider intellectual map: data science depends on trustworthy measurement, technical implementation, communication and responsibility for consequences. Follow the work, test its claims, and resist equating visibility with expertise.