// Newsletter
Practical essays on data, AI, and adoption.
No noise. One sharp essay on data, AI, or the work of building systems people trust. That’s it.
Pick your most-argued-about report. Score it honestly. Know exactly where to start.
If any of these are true, you don't need a new tool. You need to fix the foundation.
Someone keeps a personal spreadsheet "just in case" the official dashboard is wrong.
Two people cited different numbers for the same metric in the same meeting - and it wasn't resolved in the meeting.
Nobody can name, in under a minute, who's accountable when your most-viewed report is wrong.
Pick one report - your most-argued-about one. Score each criterion 1 to 5. Three or more low scores means you already have the pattern, whether or not it's visibly broken yet.
Can you name the one system that's authoritative for this number - not the easiest to pull from, the actually correct one?
Does your core metric mean the exact same thing everywhere it appears - not approximately, exactly?
Can you name every downstream report that depends on a given field, before someone changes it?
When something changes upstream, is there an actual rule requiring a downstream check - or does everyone find out when the report looks wrong?
A shared name for where you are makes the conversation easier - in a budget meeting, a reorg, or a vendor evaluation.
| Level 1 | Level 2 | Level 3 | |
|---|---|---|---|
| Architecture | Fragmented - no single source of truth; nobody's mapped what feeds what | Mapped - documented for your most-argued-about report only | Unified - every core report has a documented source; new reports must declare theirs |
| Ownership | Unassigned - nobody's accountable in under a minute; seams between systems are nobody's job | Named - key reports have owners; seams still fall through | Covered - every report and handoff has a named owner that survives reorgs |
| Governance | Informal - rules are tribal knowledge; changes ship with no downstream check | Emerging - a changelog exists for one system, followed inconsistently | Enforced - the downstream check happens automatically, every time |
None of this needs a budget or a committee. Check them off as you complete them.
Five questions. A gut check on what stuck.
According to the framework, what's the real cause of dashboard mistrust?
What's the correct order to fix the three pillars?
What is a "seam owner" in the ownership model?
In the Starbucks AI inventory example, what actually went wrong?
What's the fastest first action to take in your own organization?
You've scored the report. Now fix it. Two ways to go further.
No noise. One sharp essay on data, AI, or the work of building systems people trust. That’s it.
