Writing
Practical perspectives on intelligence platforms, data modernization, AI analytics, and building data products that compound.
Customer Records Are Split Across Systems. Here Is What I Unify First.
CRM, POS, and ecommerce rarely share one customer key. I start with the decision you need, then the identity that decision depends on.
Read More →RFM Segmentation Fails Until the Customer Is One Person
Recency, frequency, and monetary value are simple. They fall apart when one person is three records, or when half the transactions were never tied to anyone.
Read More →What I Inspect Before an SSIS Estate Moves to Microsoft Fabric
A Fabric license does not migrate an SSIS estate. I start with what the packages actually feed, which ones are load-bearing, and what has to keep running during the move.
Read More →The Data Layer an Agency Needs Before It Promises Segmentation
Segmentation, lifecycle, and retention are promises about a person. They hold only when the client's systems can produce one customer history.
Read More →What to Put in the First Message
The intake and the contact form ask for the same thing: the outcome, what is in the way, and the systems involved. A useful brief does not require a data room.
Read More →Why Building a Vertical Intelligence Platform Beats Hiring a Data Engineer
A senior data engineer costs $150K/year and builds one thing. An intelligence platform build delivers a full SaaS product that runs itself.
Read More →The Schema Context Document: Why AI SQL Queries Fail Without It
Natural language to SQL works in demos and fails in production. The reason is almost always the same: the AI doesn't know what your tables mean.
Read More →Bronze, Silver, Gold: The Data Architecture Pattern That Works Every Time
After 20 years and 200+ production pipelines, I've tried most approaches. The medallion architecture is the closest thing to a universal pattern I've found.
Read More →