¶ case file 6 of 7 · master inventory pipeline · filed 2025-06 · last verified 2026-08
status: concluded · role ended 2026-05
↳ consigned at the yard, ¶ 03 · 07:52 · waybill: the master inventory → the case file
Master Inventory Pipeline
Private proof from institutional ITSM data work: a Python and SQL pipeline that takes Workday exports and Tableau metadata, two systems that disagree, and files them into one master inventory of 35 fields keyed by a deterministic id.
[ problem ] · § as found
Asset identifiers, ownership fields, Tableau metadata, Workday exports: each system keeps its own version of the truth. Reporting built on records that disagree inherits the disagreement.
constraints
- Keep institutional details private while explaining the engineering shape.
- Normalize large operational datasets without losing auditability.
- Create deterministic inventory identifiers across heterogeneous sources.
- Produce outputs that downstream Tableau Prep flows and dashboards can trust.
fig. 1: two systems that disagree, filed as one table. every count is the checked-in ledger’s own.
a drawn plate, at rest, not a screenshot. all three row counts and the 35-field schema are the checked-in proof ledger’s; no institutional record is shown.
[ architecture ] · § fig. 2, inked
Workday exports and Tableau metadata feed Python and SQL transforms, deterministic IDs, timestamped run artifacts, and dashboard-ready outputs.
Workday exports
Operational report inputs
metadata
Tableau metadata
Report and asset metadata
report rows
Python/SQL
Cleaning and transforms
normalized records
Unified records
Unified schema and IDs
validation exports
Run artifacts
Timestamped validation outputs
trusted dataset
Tableau Prep
Dashboard-ready inventory
[ decisions ] · § as filed
d1: use deterministic inventory ids · accepted
Cross-system reconciliation needed stable keys instead of display names or inconsistent source IDs.1
1 tradeoff: ID logic adds up-front modeling work but makes downstream joins and audits more reliable.
d2: preserve timestamped run artifacts · accepted
Operational data changes over time and stakeholders need to inspect what a run produced.2
2 tradeoff: Artifact storage adds cleanup overhead but improves debugging and trust.
d3: keep the proof private but specific · accepted
The work is organizational and cannot expose raw institutional records.3
3 tradeoff: No one outside the institution can inspect the source data, so this file names the systems, constraints, and validation shape instead.
[ validation ] · § the receipts
validation
walks the receipts below, top to bottom, and marks each row this page can verify: a check where a pinned artifact resolves, a ring where the trail ends in an on-page capture, a dash where a claim is described only. walked once, the result settles here and stays.-
01claim: 3,731 Tableau rows and 6,743 Workday rows consolidated into a deduplicated master_inventory.csv of 10,453 rows.
- artifact: checked-in proof ledger, below
[private-safe]
-
02claim: The unified schema of 35 fields generates deterministic inventory_id values from row fields.
- artifact: checked-in proof ledger, below
[private-safe]
-
03claim: Runs write timestamped folders plus cumulative processed outputs, while raw institutional exports stay out of version control.
artifact: no linkable artifact, described only
[private-safe]
-
04claim: The source repo passed 3 extractor tests and critical ruff syntax/import checks in its local virtualenv during this audit.
- artifact: checked-in proof ledger, below
[private-safe]
outcomes
-
05claim: The unified schema feeds Tableau Prep and dashboard workflows without exposing institutional records.
artifact: no linkable artifact, described only
[private-safe]
ci rows link the public run · repo pins are the exact commits verified 2026-08.
what i’m NOT claiming
- This file shows counts, schema shape, and architecture only; raw CSV rows, owners, report names, PAT values, and institutional exports stay private.
[ ledger ] · § master inventory, processed-output ledger · checked in 2026-06
| entry | value | note |
|---|---|---|
| tableau output | 3,731 | processed metadata rows |
| workday output | 6,743 | processed custom-report rows |
| master csv | 10,453 | deduped master rows |
| schema | 35 | source fields defined |
| local tests | 3 passed | pytest, extractors |
| ruff check | passed | E9, F63, F7, F82 critical rules |
download the raw ledger (json) ⟶
sanitized local audit summary: no raw institutional rows, owners, report names, or PAT values.
[ corrections ] · § the register
erratum · 2026-09-23
The home page drew this work as one arrow chain: a legacy Laravel reporter into an ETL feed into a dashboard, and a compliance series rising across it. Every figure in that sentence was true and was read off Miami’s own systems, but the chain reads as a causal claim, and the rise is not mine to claim: most of it was already under way before the June 2025 start. Recorded here rather than quietly reworded because the defect was in the punctuation doing argumentative work, which is the kind a claim audit is least likely to catch. Superseded three days later, when the figures came off the site altogether: see the entry below, and the 2026-09-28 entry recording how this note was itself corrected.
erratum · 2026-09-26
The note above is withdrawn. The compliance figures and the reporting systems it names belong to the team, not to work this file can source, so they are off the site: the month count, the project count and both ends of the percentage range are retracted rather than restated in the run or its case files. A receipt that counted operational records in this role is removed as well: it had no artifact, no date and no source.
erratum · 2026-09-28
The 2026-09-23 note above has itself been corrected. It went on quoting the retracted compliance percentages as the honest form of the claim, one entry above the note that withdrew them, so this register was still publishing the figure it had retracted, and the built page carried it. The sentence that restated them is gone. What that note records, an arrow chain whose punctuation argued a cause the work did not earn, is unchanged, because that is the part it exists to keep. Written down rather than edited in silence, because an erratum corrected without a record is not an erratum.
[ appendix ] · § artifacts, none beyond the page
nothing to open here, deliberately: the ledger above is this file’s publishable artifact, shown in full, and the records behind it stay private.