Kwik Trip — walk-in card (Mon 2026-07-13)
Layers on [[2026-07-13-kwik-trip-meeting-prep]] (company snapshot, logistics, base questions) and [[2026-07-11-kwik-trip-jobs-tech-signal-read]] (hiring corpus, full 7-probe list). Source: Build Manifest 2026-07-11T111042 + contracts 2/4, cellar/assessments/kwik-trip/.
Posture
This is a pre-discovery candidate portfolio (Contract-4 landed 2026-07-11, meta-council ratified, founder C1-approved): every claim rests on public data plus our own 07-10/11 corpus — zero client-internal artifacts. Platform target is frozen bespoke, honestly: Databricks verified in production (DAIS 2024), the Snowflake-migration signal is a single uncorroborated LinkedIn source, and all six structured bespoke constraint fields are to_be_captured_at_discovery. Contract-2's verdict is proceed_with_named_constraints — readiness to discover, not to build. Monday is for capturing those six fields, testing the capacity model, and confirming which candidates they already cover — walk in holding hypotheses, not pitching builds.
Portfolio — 12 candidates (EA = evidence-anchored · GA = general-applicability, founder-directed breadth, tier-labeled not inflated)
| # | Candidate | Tier | Why (one line) |
|---|---|---|---|
| Must | |||
| KT-01 | SAP S/4HANA → lakehouse reconciliation | EA | The multiplier thread: every downstream use (incl. KT-03's forecast) inherits this pipeline's reliability; rank-1 corpus signal. |
| KT-03 | Store-scale demand-forecast expansion | EA | Extends their verified in-production hourly app — strongest substrate we hold; ROM 1–3% waste/stockout, percentage only. |
| KT-02 | Eagle Eye loyalty reverse data flow | EA | Integration-unlock: no member-linked analytics on 5.25M Kwik Rewards members until loyalty outcomes land in the lakehouse; value deliberately unquantified. |
| KT-05 | Governance & access posture | EA | Fronts the access-grant/vendor-diligence gate every other thread routes through, under their live security push (4 of 9 IT reqs). |
| Should | |||
| KT-04 | Data-team capacity relief (self-serve) | EA | Demoted must→should at C4: value case void under 2 of 3 capacity-model readings (CR-KT-5); promotion path recorded. |
| KT-07 | Fresh-food waste/shrink reduction | GA | Highest-signal general item (forecasting-app adjacency); ROM 10–20% waste cut accruing at both production and store ordering. |
| KT-09 | DC + fleet optimization | GA | ROM 3–7% distribution cost on the self-operated network; already-covered check runs first (TMS is standard at this fleet size). |
| Could | |||
| KT-06 | Store labor scheduling | GA | ROM 1–3% labor across ~36–38k coworkers; predictive-scheduling statutes unreviewed; below-threshold confidence the gap exists. |
| KT-10 | Fuel price optimization | GA | Cents-per-gallon framing, no margin baseline; commonly vendor-covered — honest default is extend/validate, not greenfield. |
| KT-08 | Promo effectiveness (own-brand) | GA | ROM 1–2% margin lift on promoted categories; hard-sequenced behind KT-02 landing. |
| KT-12 | Predictive maintenance | GA | Only fuel-pump telemetry is verified substrate; everything else is pattern-level. |
| KT-11 | Site-selection analytics | GA | Loss-avoidance vs the $151M expansion anchor; two stacked gates (bespoke capture + external-data licensing) — sequences last. |
Top 5 discovery probes (ranked; deduped vs the tech-signal read's 7)
- The six bespoke constraint fields (decomposition unit, native artifacts, forbidden patterns, validation hooks, deployment, observability) — Monday's hard deliverable; uncaptured = whole portfolio stays hypothesis and any SOW conversation waits (kill criterion #1).
- SAP → lakehouse transport (CDC / batch / vendor tool?) — extends signal-read probe 1; a third-party tool already reconciling = KT-01 descope/kill, and get the Snowflake signal corroborated or retired in writing (CR-KT-4).
- Capacity model of the 12-person Data Services team — extends probe 2; the answer arbitrates KT-04's demotion (capacity-capped promotes it back; off-portal hiring or flat demand voids it).
- Eagle Eye contract terms on data export — extends probe 4; an export bar blocks KT-02 and KT-08 falls with it (hard precondition).
- Existing-capability sweep for the general tier (G-KT-7: WFM, TMS, fuel-pricing, promo measurement, predictive maintenance, site analytics) — each positive finding re-scopes that item to extend/validate or retires it with recorded evidence.
Cautions
- Forecasting-collision trap. They HAVE an hourly food-sales forecasting app in production — KT-03/06/07 are framed as extensions of their muscle, never "you lack forecasting." Their app may already consume exogenous signals: ask, don't assume.
- No dollar theater. 11 of 12 items carry not-estimable value sentinels (0/0 ≠ zero value — no defensible dollar base exists pre-discovery). KT-04's $35–95K is the lone dollar figure and it's a probability-discounted outside-in ROM — never client-deck-ready. Percentages stay percentages Monday.
- Stakeholder map is honestly empty (CR-KT-2): AE unconfirmed, SFDC null. Matt Korger is a research-identified public figure, NOT a stakeholder record — don't name-drop him as a contact; names come from the client docs/system-access exchange expected this week.