Bad Data? Where to Start - Tackling CFO Tech Legacy III
Why this is in the vault
Part 3 of the Tech Legacy arc - the most directly MAC-relevant of the four. The Secret CFO names a 2x2 Data Quality Diagnosis Matrix: Data Provenance (source-capture quality) on one axis, Data Fluidity (movement to right places) on the other. The four quadrants - Data Dawg (good/good), Tech Poverty (good provenance, poor fluidity), All The Gear But No Idea (poor provenance, good fluidity), Dark Ages (poor/poor) - are an immediately portable diagnostic. The thesis turn: start with data-provenance fixes before investing in new systems, because AI's emerging capability to handle messy unstructured data eliminates the historical requirement for clean data BEFORE implementation. The killer line on AI failure modes: the most dangerous error AI can make isn't one that looks wrong, it's one that looks right.
Mapping against Ray Data Co
This is the single highest-value Tech Legacy piece for MAC's editorial direction. Specific maps:
- The 2x2 matrix is portable straight into MAC content. The "All The Gear But No Idea" quadrant - poor provenance, good fluidity (modern data stack with bad data flowing fast through it) - is the exact failure mode MAC sells against. There's a MAC piece titled "All The Gear But No Idea: Why Your Modern Data Stack Is Lying To You Faster" with this article as the citable provenance.
- The "AI errors that look right" warning is the harness-engineering thesis stated in CFO-buyer language. AI without harness = confident-sounding wrong answers. The harness IS the data provenance layer + the validation layer that catches the look-right-but-actually-wrong outputs. MAC should make this its anchor argument.
- The "Buys → Makes → Sells" three-cycle frame with the explicit prioritization on Makes (because that's where margin insight lives) is portable into how MAC sequences its content: cover the makes-cycle data quality first, then buys, then sells. That's a content roadmap.
- RDCO's own data discipline: the matrix applies to the vault. Provenance (where each note came from, how it was captured) is mostly good. Fluidity (does the right note surface at the right moment via QMD / graph) is the weaker axis. RDCO is in the "Tech Poverty" quadrant on its own data, and the [[graph-reingest]] discipline is the move to push it toward Data Dawg.
- Cross-link to existing CFO Secrets material: the "AI can do a lot with messy data, but it can't work with no data" line directly extends the [[06-reference/2026-04-28-cfosecrets-finance-stack-of-the-future-unbundled-erp]] argument and reinforces the [[06-reference/concepts/2026-05-10-harness-moat-two-layers-portability]] case from a data-provenance angle.
⚠️ Sponsorship
Sponsored by Maximor AI (recurring). Topic-matched - Maximor's pitch is exactly the AI-handles-messy-data direction the article advocates. Disclosure clean. The argument-vs-sponsor alignment is tighter here than in Parts I-II so worth flagging when re-citing.
Related
- [[06-reference/2026-03-07-cfosecrets-shelfware-shenanigans-tech-legacy-i]] - Tech Legacy I, the diagnosis
- [[06-reference/2026-03-14-cfosecrets-how-you-got-here-tech-legacy-ii]] - Tech Legacy II, the 7-eras history
- [[06-reference/2026-03-28-cfosecrets-unbundling-the-erp-tech-legacy-iv]] - Tech Legacy IV, the unbundled-stack architecture
- [[06-reference/concepts/2026-05-10-harness-moat-two-layers-portability]] - the harness-moat thesis; data provenance is the harness foundation
- [[06-reference/2026-04-30-mac-bet-architecture-audit]] - MAC bet architecture; the 2x2 matrix is MAC's most portable piece of the corpus