How one analyst turns a business question into insights a customer can act on: the eight-part problem statement, the same structure applied twice to build a base then answer from it, and the verification loop that decides when a finding has earned confidence.
Working with LLMs
Research Process & Methodology
When Even the Best Prompt Isn't Enough
We use these tools well and still leave most of their power on the table. The case for a durable platform and process behind serious research and writing, rather than a better one-shot prompt.
Claude for Research — Learn to use the Stick
Most chat tools are automatics; the most capable one asks you to drive. Why learning to operate Claude deliberately, like a manual transmission, reaches places one-shot prompting cannot.
Data engineering
The Data Lifecycle Series
A three-part series on why a single retention rule pulls in opposite directions at once, and the architecture that resolves it.
-
Paper 1
Designing for the Data Lifecycle is not easy
One retention rule usually hides four distinct mechanisms and two different problems; conflating them is where implementation breaks.
-
Paper 2
The Best Pattern for a New Source
The Day Zero fork: one ingestion point splitting into a compliance vault and an analytical store, each governed under its own rules.
-
Paper 3
When the Request Is "Delete Me"
Why a subject-erasure request and routine analytical removal are the same operation, and the platform conditions that make it a managed step rather than an investigation.
The Data Access Platform
Two decades running analytics platforms at a large healthcare company, told through the Brooklyn Bridge: why organizations tolerate "good enough" data access for years, and what it finally takes to commit to the platform they actually need.
Leadership
The Interns — Red Team
A narrative series on leading the team nobody else knew what to do with: a dozen summer interns handed to one manager, and what happens when you turn them loose to stress-test the work.
A Team You Have to Build and Lead
Most writing about these tools stops at productivity — faster drafts, quicker summaries. The more useful question is whether they can help a researcher, analyst, or leader do the deeper work, once you build and lead them like a team rather than prompt and walk away.