Headline: the cube delivery is strong on transactional history (shipments, POs, market data) but the four highest-leverage items are still open: F-01 frozen forecast submissions (marked available, undelivered — blocks the V2 real baseline and the V3 K1 measurement), T-01 plan rows + F-12 launch calendar (Brian dependency — blocks V1 cube-only ingestion and V2 lifecycle forecasting), T-03/T-04 open orders (marked available, undelivered), and F-04 snapshot history (starts 2026-08-18, needs run-length).
Degree of readiness by version (core datasets, priority-weighted)
Score per dataset: 3 Ready · 2 Partial · 1 Promised/at-risk · 0 Missing. Version readiness = Σ(score ÷ 3 × weight) ÷ Σweight over that version's core datasets, weighted P0=3, P1=2, P2=1.
Where the gaps sit
Count of core datasets by readiness state for each version. The V2/V3 orange-and-red block is dominated by items the client marked "available" in the data-request sheet but has not yet delivered.
Dataset × version readiness
3 · Ready2 · Partial
1 · Promised / at-risk0 · Missing
— · Not required
Axis 1 columns: Forecasting = how the V2 build uses it today (used / proxied / simulated / not wired). Data-sheet Excel = the client's own availability flag in DSA_Data_Request_vShare.xlsx. Data cube = what actually landed in 03_Data/data_cube. Axis 2 columns: requirement and readiness per version — chips marked ● are core requirements, ○ supporting.
Ranked gaps — what to chase, in order
| # | Item | Why it matters | Owner / path | Unblocks |
|---|---|---|---|---|
| 1 | F-01 frozen forecast submissions (B1–B4) | Marked available, never delivered. The only clean way to measure K1 baseline accuracy at planning grain, and V2's real (non-reconstructed) baseline. One dataset, two projects. | Client BI / My Mai | V2 · V3 |
| 2 | T-01 plan rows + F-12 launch calendar | Both live with Brian's team, outside JDE. T-01 blocks V1's governed plan ingestion; F-12 (plus empty LAUNCH/EOL in T05) blocks lifecycle forecasting — the dominant error driver (SF13→SF14). | Brian | V1 · V2 |
| 3 | T-03 / T-04 open order book with snapshots | Marked available, undelivered. V1's current-month logic ("expected incoming", ASAP) still leans on manual sheets; open-order aging is also the best K5 stockout proxy. | Client BI (Haley open-PO dataset) | V1 · V3 |
| 4 | F-02 / F-03 full-portfolio weekly sell-through & KA inventory | Pilot covers Soft Feel only (2023–2026, 11 accounts). Full portfolio turns proven pilot regressors into production signals. | Client BI — can ship as one combined feed | V2 |
| 5 | F-04 snapshot cadence | 2026-08-18 file must become a standing weekly/monthly extract; K3 (inventory baseline) quality is purely a function of run-length. History never existed — every week waited is a week lost. | Client IT (Kevin) — confirm schedule | V2 · V3 |
| 6 | T-07 with effective dates | Current price list only. K6/K7 (clearance, markdown) need price history; interim proxy = T02 realized price vs T07 wholesale. | Client BI | V3 |
| 7 | Cheap fixes in delivered files | T02: add header row · T05: populate LAUNCH_DATE / EOL_DATE · T06: quote names with commas · F07: multi-level BOM. | Client BI — trivial | V1 · V2 |
| 8 | Non-client items | Generate T-08 calendar (done in pipeline) · stand up planner effort log (K8/K9) · get carrying rate from finance · lead time & unit cost now measurable from F05/T07 — retire two guessed assumptions. | Viaduct / DSA planning team | V3 |
What is already strong
- T02 shipments 2020–2026 — exceeds the 36-month ask, richer than spec (line types, return reasons, DOM/INT, Open/Closed) — the actuals side of every accuracy measure is done.
- F13 Golf Datatech full portfolio 2007–2026 — retires the simulated product-share file in the V2 pipeline.
- F05 purchase orders 2017–2026 — supply signals ready, and makes lead time measurable (replacing the 28-day guess in the value model).
- T06 at customer grain (ship-to added as offered) — enables Canada split (F-16 partial) and future customer-level channel mapping.
- T07 cost rollups — makes unit cost measurable (replacing the $35 guess in the value model).
Version definitions
| Version | What it is | Core data dependencies |
|---|---|---|
| V0 — Excel planner (as-is) | Current planner tool fed by manual upload of the SRX Live File; replicates the NASOP build-up. | Plan tabs, SHIP DATA, OPEN Thru, ASAP sheets — all inside the live workbook. |
| V1 — Governed planner | Multi-source pipeline + persistent DB; the planner reads governed extracts (cube) instead of manual files. | T-01…T-08 (plan rows, shipments, open orders, dims, pricing, calendar) + F-04 snapshots. |
| V2 — Forecasting engine | ML baseline + exogenous signals (sell-through, inventory, market share) at full portfolio, replacing pilot extracts and simulated data. | T-02/T-05/T-06/T-08 + F-01, F-02/03, F-04/05, F-12, F-13; F-06…F-16 enrich phases 2–3. |
| V3 — Deployed & measured | Live forecast cycles with write-back, plus K1–K10 impact tracking in measured mode (the value case). | F-01 + T-02 (K1/K2 at planning grain), F-04 history (K3/K4), T-03/F-06 (K5), T-07 history (K6/K7), effort log (K8/K9), adoption telemetry (K10), finance params. |
Readiness scoring
| Score | Meaning |
|---|---|
| 3 · Ready | Delivered in the cube (or generated internally) with usable quality and sufficient history for the version's purpose. |
| 2 · Partial | Delivered or available with material defects/limits (missing fields, no history, pilot scope only, proxy via another file). |
| 1 · Promised / at-risk | Client marked it available in the request sheet (or a design exists) but nothing has landed yet. |
| 0 · Missing | No source identified, or explicitly not available (Brian-dependency, doesn't exist in JDE, log not yet built). |
| — · n/a | Not required for that version. |
● core requirement — counted in the version readiness score. ○ supporting — shown but not scored.
Axis-1 source columns
- Forecasting (V2 build) — from 03_DSA_forecasting: the five processed inputs (sales_history, ka_sell_thru, ka_oh_inventory, gdt_market_history, market_share_simulated), what is real, what is a proxy, and what is simulated.
- Data-sheet Excel — DSA_Data_Request_vShare.xlsx (Dataset Register, 24 requests T-01…T-08 / F-01…F-16) with the client's own "Available?" flag and history notes.
- Data cube — the nine files delivered to 04_DSA_excel/03_Data/data_cube, profiled for grain, history, and quality defects (headers, null fields, encoding).
Full profiling detail lives in Docs/DSA_Data_Assessment_20260819.md; the plan-vs-actual accuracy companion is 05_Impact_Tracking/01_Docs/DSA_Baseline_Accuracy_Report.html.