Field Notes on Building Trustworthy Analytics
Dispatches from active data warehouse and analytics engagements — the decisions, tradeoffs, and lessons as they happen.
-
When Two Numbers Reconcile, and They’re Both Wrong
A legacy margin report matched finance for years because two errors happened to offset each other. Reconciliation proved agreement, not truth.
Written by
-
Turnover Is a Data Quality Risk
Backing up data preserves the bits. It does not preserve the reasoning behind a critical model when its only knowledgeable owner leaves.
Written by
-
A Migration Shouldn’t Rewrite the Company’s History
A migration must distinguish corrections from reclassifications—or risk quietly rewriting the company’s past to resemble its present.
Written by
-
When AI Writes the Code, Clarity Becomes the Craft
AI can infer a rule from historical data in an afternoon. Responsible review must still distinguish a business decision from a constraint or a correlation.
Written by
-
The Mystery Measure
Field notes from an active migration: what happens when a trusted number outlives the reasoning behind it, and why the recovery is rarely only technical.
Written by
-
What Grain Arguments Are Really About
A fact table’s grain looks like a technical choice. Underneath, it is often a business argument about which events occurred and which decisions the model must support.
Written by
-
The Shadowy Spreadsheet
A shadow spreadsheet is not merely a discipline problem. It is evidence of the exact place an analytical system stopped representing the business correctly.
Written by
