Chris Hillman
Friday, October 30, 2026
Start Here
Eighty-odd posts is a lot to wade through. These six are where most readers land first, and together they cover the three things I keep writing about: how to model data so it can be rebuilt, how to run a pipeline nobody has to babysit, and how to lead the people doing both.
The Competitive Moat That AI Can't Replicate
A restaurant that refused online bookings, and what it taught me about the one advantage a data team has that no tool can copy. If you only read one post here, read this one.
Read it →
The methodologies
Named because my teams use them by name. Each one came out of a production incident, not a whiteboard.
Healing Tables: When Day-by-Day Backfills Become a Slow-Motion Disaster
Stop replaying history one day at a time. A six-step framework for Type 2 dimensions that rebuild deterministically from source, with the SQL and the tests.
Write-Audit-Publish with Iceberg Tables in Snowflake
Hands-on: stage the load, gate it on quality checks, publish with a zero-I/O branch swap. Production never sees a bad row.
Don't Go Dark: Visibility Is a Data Engineering Skill
Long migrations and invisible pipelines kill careers quietly. How to make your work legible to the people who decide whether it mattered.
Five Worlds of Data Engineering
Why half the internet is furious at the other half's advice. Work out which world you're in before you borrow anyone's best practice.
Modelling and strategy
For when the question is "what should we build" rather than "how".
The 2026 Data Engineering Strategy Nobody's Writing
Everyone's planning memo is about AI tooling and cost. The talent pipeline is what will decide the next five years. What to write instead.
Your Data Model Isn't Broken: Why Refactoring Beats Rebuilding
That 200-column fact table is reality encoded, not a bug. Part I makes the case; Part II is the strangler-fig playbook.
Your Data Platform Costs More Than It Should
Snowflake and AWS cost, measured and reduced, starting with the habits rather than the invoice.
Leading the team
The half of the job that doesn't show up in the architecture diagram.
The Style Isn't the Skill
What Bruce Lee's messy fights and broke years teach data engineers about mastery, and about writing your next self-evaluation.
Your Team Already Has Patterns. They Just Don't Know It.
A shared vocabulary for how work gets done, and the confidence to estimate it. How to build one without turning it into a bureaucracy.
Data Engineers: Just Do Code Reviews
The evidence was settled in 1993. Data engineering still hasn't caught up, and the cost shows up in incident reports.