Inventory HealthPillarsPatternsDangersOpportunitiesWisdomCuriosityAction

TM-203: Curiosity — Standing Questions Log


Structural Component #6 — Between Wisdom and Action

Last updated: 2026-10-27 (Design v2 — Output Levels + Seasonal CQ Mode + Item-Store Companion)



What this file is


A living log of questions the system cannot yet explain. Not findings — questions. Each entry represents a gap between what the system expected (from Pillars, Patterns, Wisdom) and what it observed, where the gap matters commercially and the answer is findable.


Output format: "Expected X, observed Y, in context Z — worth looking at?"



Output Levels (Design v2)


Every CQ entry produces output at two levels. Chain summary alone is not sufficient.


LevelWhatExample
Chain summaryThe aggregate signal — total items, total value, scope of the gap"24,495 fast-selling items with zero inbound supply — $1.1M at risk"
Item clusterTop 5-10 specific UPCs behind the question, ranked by (velocity x retail value)"UPC 661127171679 PLUSH PNUT SNPY — 22 units/30d, OH=0, no PO. $40 retail. Site 165."

Chain summary provides orientation. Item cluster is what travels to the DM for investigation. Both are required.


Item-store companion questions: where a specific transfer row has a structural anomaly (no transaction history, post-window seasonal, dead item with active reorder), a paired Curiosity question appears beside that Action row. It does not block the transfer — it flags the open question alongside the decision.



How entries move


`

OPEN → INVESTIGATING → CLOSED (outcome noted)

→ GRADUATED (promoted to Wisdom candidate)

`


Entries suppressed from other pillars (Dead Stocks, Stockout Risk, Uneven Distribution) are excluded — Curiosity only covers what the other components have not already caught.



Open Questions


CQ-001: Replenishment System Blind to Stockouts

Status: OPEN

Detected: 2026-10-27 (Mode A — Standing Question Layer)

Sources: Stockout Pillar (already_stocked_out) × stockout_items (on_order)


Expected: When 2,188 items stock out with confirmed daily sales velocity, the replenishment system generates on-order activity to refill them.

Observed: Zero of the 2,188 stocked-out items have on_order > 0. None. The replenishment system is generating no orders in response to active stockouts.

Context: These items have average 10.1 days at OH=0 with real sales activity. The system is aware — it classified them. But no order was triggered.


Curiosity question: Is the replenishment system disabled for these sites/categories, or is there a structural override preventing auto-reorder on stocked-out items? The answer would change how Transfer Options is prioritized — if replenishment is broken, transfers are the only near-term fix.


Resolution pathway: Check Hallmark Connect vendor order settings, or query on_order fields at the vendor level. DM field knowledge may surface whether specific sites have had replenishment paused.


Item cluster: Top stocked-out items with highest velocity and no PO — to be populated on next run.

(Ranked by: daily_velocity × retail_price, OH=0, on_order=0, days_at_zero > 7)



CQ-002: Dead Stock with Active Reorder Points ($995K)

Status: OPEN

Detected: 2026-10-27 (Mode A — Standing Question Layer)

Sources: Dead Stocks Pillar (items where reorder_point > 0)


Expected: Items classified as dead stock (91+ days no sale) have reorder_point = 0 or suppressed. The system agrees they are dead and does not order more.

Observed: 16,244 dead stock items carry reorder_point > 0 — the system's own reorder logic is set to trigger restocking of product it simultaneously classified as dead. Total OH value: $995,594.


Curiosity question: Are reorder points being updated when an item transitions to dead? Or is the reorder_point set once at item creation and never cleared? If it's never cleared, the system may be passively queuing reorders for $995K of dead product.


Resolution pathway: Check if reorder_point reflects a live system setting or a historical configuration. If live, this is a systemic mis-configuration that would compound the $10.8M dead stock problem rather than resolving it.


Item cluster: Top dead items by retail value with active reorder points — to be populated on next run.

(Ranked by: OH_value DESC, dead_days DESC, reorder_point > 0)



CQ-003: 893 Dead Items with Active On-Order (Ordering More of What's Not Selling)

Status: OPEN

Detected: 2026-10-27 (Mode A — Standing Question Layer)

Sources: Dead Stocks Pillar (on_order_units > 0)


Expected: No active on-order for dead stock items. Buying is paused or cancelled when product hasn't sold in 91+ days.

Observed: 893 dead stock items have on_order_units > 0, totaling 10,668 units inbound to locations where the same UPC is already classified dead. Current dead OH value at those locations: $17,106.


Curiosity question: Were these orders placed before the product went dead (pipeline lag), or is the system actively re-ordering items that stopped selling? Either way — 10,668 units of confirmed dead product are inbound with nowhere to go.


Resolution pathway: Check order dates vs. last_sold_date for each affected item. If order date > last sale date, the ordering happened after the product died. That is the actionable case. Sites most affected: 1789 (64 items), 1822 (61), 231 (49), 166 (46).


Item cluster: Top dead items with on_order — to be populated on next run.

(Ranked by: on_order_units × retail_price DESC, filtered to order_date > last_sold_date)



CQ-004: 24,495 Fast-Selling Items with Zero Incoming Supply ($1.1M at Retail Risk)

Status: OPEN

Detected: 2026-10-27 (Mode A — Standing Question Layer)

Sources: Stockout Pillar (quick_sellouts where on_order = 0)


Expected: Items classified as quick sellouts (selling fast, low DOS) have on_order > 0 — replenishment is running ahead of demand.

Observed: 24,495 of 24,686 quick sellout items (99.2%) have on_order = 0. Total retail value at risk: $1,113,245.


Of these, 20,155 items have less than 7 days of stock remaining with nothing inbound.


Curiosity question: Is vendor replenishment simply not available for these items (out of stock at vendor, discontinued), or is the ordering function not running? The scale (99.2% of quick sellouts have no supply coming) is too high to be a vendor availability issue. Something systemic is suppressing orders.


Resolution pathway: Cross-reference against Hallmark Connect availability data. If vendors have stock, this is a store-level or system-level ordering failure. If vendors are out, that is a separate but equally important signal heading into holiday.


Item cluster: Top fast-selling items with no supply path — to be populated on next run.

(Ranked by: daily_velocity × retail_price DESC, DOS < 7, on_order = 0, seasonal_category prioritized first)

Note: seasonal items (Christmas, Thanksgiving) in this cluster are Mode D candidates — flagged for urgency escalation.



CQ-005: Kathy Carter Territory Has 84% More Supply-Starved Quick Sellouts Than Vanessa's

Status: OPEN

Detected: 2026-10-27 (Mode A — Standing Question Layer)

Sources: Stockout Pillar (quick_sellouts by DM, on_order = 0)


Expected: Supply starvation on fast-selling items would distribute roughly proportionally across DM territories (Kathy 14 sites, Pascale 13, Vanessa 9).

Observed: Kathy Carter: 10,175 items. Pascale: 8,796 items. Vanessa: 5,524 items.

Adjusting for site count: Kathy 727 per site, Pascale 677 per site, Vanessa 614 per site.


Curiosity question: Is Kathy's territory carrying higher-velocity product across more categories, or is there a territory-level replenishment disparity? The per-site difference (727 vs 614) is meaningful in a holiday season context. If Kathy's stores are higher volume, the starvation impact is amplified.


Resolution pathway: Kathy Carter as DM would have direct visibility. Compare Kathy site sales velocity to Pascale/Vanessa sites. If Kathy's stores are genuinely higher-traffic, the replenishment system should be calibrated accordingly.


Item cluster: Top Kathy territory items with per-site starvation rates — to be populated on next run.

(Ranked by: items_starved_per_site by site within Kathy territory, to isolate whether 1-2 sites are driving the gap or it is uniform across all 14)



CQ-006: 5,776 Transfer Candidates with No DM Assignment

Status: OPEN

Detected: 2026-10-27 (Mode B — Divergence Detector)

Sources: Transfer DB (donor_dm = "" or NULL)


Expected: All 39 active stores have a DM assigned in the stores table. Transfer candidates should carry DM labels on both donor and recipient sides.

Observed: 5,776 transfer candidates have empty donor_dm and recipient_dm fields — 5,553 from the Dead Stocks pillar, 223 from Expiring Stocks. These represent $133,297 in transfer value. Additionally, 4,901 transfer candidates have DM fields populated but they differ between donor and recipient (cross-territory transfers), which the same-DM constraint should have excluded.


Curiosity question: Are these from sites that exist in the transfers pillar but are absent from the MySQL stores table DM mapping? Or did the DM enrichment step miss them? If the former, these sites are effectively invisible to the territory-based workflow.


Resolution pathway: Query MySQL stores table for the specific site numbers appearing in empty-DM rows (sites like 357 appear in sample). If they're in the stores table, the enrichment logic has a bug. If not, these are sites without DM assignment — a master data gap.


Item cluster: All sites appearing in empty-DM transfer rows — to be populated on next run.

(This is a data integrity check, not a velocity ranking — list all affected site numbers with their transfer counts and total value)



Seasonal Open Questions (Mode D — Active: Christmas window 59 days, Thanksgiving 29 days)


These fire because Season Context is in CRITICAL/HIGH urgency tier as of 2026-10-27. To be populated with item clusters on next run.


CQ-007: Christmas Category Items at SR4/SR5 with No Supply Path

Status: OPEN (pending next run)

Mode: D — Seasonal Gap Detector

Season context: Christmas window open, 59 days remaining, CRITICAL tier


Expected: Seasonal category items (Christmas cards, ornaments, seasonal decor) with SR4/SR5 status would have active POs or confirmed transfers in queue — given that the holiday is 59 days out.

Observed: To be measured on next run.


Curiosity question: Which Christmas-category items are at stockout risk with no inbound supply during the highest-demand window of the year? This is the highest-urgency variant of CQ-004.


Resolution pathway: Filter stockout_items to seasonal_category = Christmas + SR4/SR5 + on_order = 0. Cross-reference Season Context urgency tier.


Item cluster: Top 10 Christmas-category SR4/SR5 items with no PO, ranked by (daily_velocity × retail_price × days_remaining_in_window).



CQ-008: Seasonal Category Sites with Flat Velocity During Expected Surge

Status: OPEN (pending next run)

Mode: D — Seasonal Gap Detector

Season context: Christmas window open — October velocity should be trending up for seasonal categories


Expected: Sites with Christmas card and ornament assortment would show accelerating velocity as the holiday approaches (October > September > August trend).

Observed: To be measured on second run (requires baseline).

Note: Until a second same-period run exists, this question is structural — cannot be answered until velocity baseline is built.


Curiosity question: Are there specific sites where seasonal category velocity is flat or declining while chain-wide seasonal velocity is rising? Flat where surge is expected could mean assortment mismatch, markdown needed, or store-level execution gap.


Resolution pathway: Compare site-level seasonal category velocity week-over-week. Sites deviating -20% from chain average seasonal trend are flagged.


Item cluster: Sites with lowest seasonal velocity relative to chain average for Christmas category, with OH > 0 (not explained by stockout).



CQ-009: Halloween Stock Still On-Hand After Oct 31

Status: PENDING (activates Nov 1)

Mode: D — Seasonal Gap Detector + E — Item-Store Companion

Season context: Halloween sell window closes Oct 31 — 4 days from current run date


Expected: Halloween-category items reach OH=0 or near-zero at time of window close. Any remaining stock is an immediate markdown or inter-store transfer candidate (if any sites have sell window still open).

Observed: To be measured on Nov 1 run.


Curiosity question: Which sites will have Halloween inventory on Nov 1 that should have cleared? What is the markdown path?


Resolution pathway: On Nov 1 run, filter dead_stocks or stockout to Halloween category + OH > 0. These are post-window dead — Wisdom rule applies: markdown, not transfer.


Item cluster: All Halloween-category items with OH > 0 on Nov 1, ranked by retail value, with site-level breakdown.



CQ-010: Dead Stocks Run Missing 9 Sites (Oct 27)

Status: OPEN

Detected: 2026-10-27 (Mode B — Divergence Detector)

Sources: Dead Stocks Pillar site coverage comparison (Oct 26 vs Oct 27)


Expected: The Oct 27 dead stocks run (dead_stocks_v5.py) covers all 36 active stores, consistent with Oct 26.

Observed: Oct 27 run produced data for only 27 sites. 9 sites present in Oct 26 are absent from Oct 27:

Missing: 1403, 157, 160, 162, 165, 166, 1825, 1940, 1948

Notable absences: 157 (N Point Village Reston), 160 (Dulles Town Center), 165 (Virginia Gateway), 1948 (Leesburg Plaza), 166 (Oak Springs Plaza)


Curiosity question: Did these 9 sites fail to pull data during the Oct 27 run? If so, why — connection error, no data for the period, or were they filtered by the isActive=1 store query? Missing 9 sites means today's dead stock totals ($8.5M, 319K rows) are understated vs. the true chain position.


Resolution pathway: Re-run dead_stocks_v5.py or check MySQL stores table for isActive status of the 9 missing sites. If they are active and the pull failed, schedule a re-run. The transfer candidates and pattern comparisons for this run should be treated as partial until confirmed.


Commercial impact: The 9 missing sites contributed ~$2.3M in dead stock value on Oct 26. Today's run is missing ~21% of chain dead stock value. Stockout and transfer data may also be incomplete for these sites.



Investigated


(None yet)



Graduated to Wisdom


(None yet)



Curiosity Modes — Operating Reference


Mode A: Standing Question Layer (fires during every run)

After each Pillar executes:


Mode B: Divergence Detector (fires between runs)

After comparing current run to prior run:


Mode C: Human-Facing Open Prompt (always available)

Standing question when reviewing any output:


Mode D: Seasonal Gap Detector (fires when Season Context is active)

When a holiday window is open (CRITICAL or HIGH urgency tier), additional seasonal CQ questions fire:


Each Mode D question fires at item cluster level — specific UPCs, specific sites, not chain totals.


Mode E: Item-Store Companion (fires alongside Action output)

For each transfer row with one of the following flags, a companion CQ question is generated: