The Modern Data Stack has become the default choice for analytics teams, often adopted early without a clear understanding of its real costs. While powerful, it frequently introduces unnecessary complexity and significant financial overhead.

In this talk, I’ll show when a full cloud-native data stack is overkill, both technically and economically. Through real-world examples and benchmarks, I’ll compare typical Modern Data Stack architectures with Python-centric alternatives built around tools like DuckDB, dbt-core, and minimal orchestration.

Beyond performance and maintainability, the talk focuses on cost drivers: infrastructure usage, SaaS pricing models, operational effort, and long-term ownership costs. Rather than advocating for a single “best” stack, I’ll present a practical decision framework to evaluate data volume, access patterns, team size, growth expectations, and budget constraints.

Attendees will leave with concrete criteria to choose simpler, cheaper architectures when appropriate and to justify the investment in more complex stacks only when the value clearly outweighs the cost.