Fictional training simulation. D2V Mock is an invented training organization. All people, data, incidents, and documents are fictional. Any resemblance to a real entity is coincidental.
MISSION 04 · INVENTORY DECISION SUPPORT
GUIDED LAB LIVE
D2V Mock
Retail
Design an inventory-transfer decision system that balances demand, margin, logistics cost, local knowledge, and adoption without pretending forecasts are certain.
Guided lab available: review the public mission specification here, then enter the paid workspace for staged evidence, decisions, artifacts, progress tracking, and portfolio export.
01 / CORE QUESTION
Which inventory should move, where, and why?
The simulated retail network operates a regional network of stores and two distribution centers. Some locations lose sales because popular products are unavailable while other stores carry excess inventory that later requires markdowns. A central planning team wants automated transfer recommendations, but store managers distrust forecasts that ignore local events, shelf capacity, labor, and customer behavior.
The mission is not solved by producing a fluent recommendation. The learner must define the decision boundary, reconstruct the evidence, build an inspectable workflow, and establish what would justify deployment.
DECISION TO IMPROVEWhich SKU-location transfers should D2V Mock Retail recommend this week, which should require local review, and what evidence supports the expected value?
CORE METHOD → MISSION
One operating decision. Multiple deployment boundaries.
The Core Labs isolate individual failure modes. This mission forces several of them to coexist under one customer timeline, one evidence room, and one final deployment recommendation.
$49STANDARD INDIVIDUAL PRICE · CORE MISSION · LAUNCH ACCESS IS CURRENTLY FREE
When should an inventory-transfer recommendation be trusted, overridden, or withheld when local demand and inventory truth are incomplete?
Forecasts, inventory snapshots, transfer economics, online promises, local events, and store knowledge each describe a different part of the decision.
Set recommendation scope, override rules, objective guardrails, approval validity, and pilot thresholds for a human-controlled transfer workflow.
CORE METHOD MAP
Review all 19 Core Labs →5 Core Labs are exercised here
02 / STAKEHOLDERS
Success has competing definitions.
The design must represent authority, incentives, operational constraints, and unacceptable outcomes rather than collapsing stakeholder needs into a generic requirements list.
Merchandise planning lead
Wants a repeatable network-wide process that reduces stockouts and markdown exposure.
Regional operations manager
Owns transfer execution, labor constraints, delivery windows, and store compliance.
Store manager
Knows local demand shocks and fears a central model will drain inventory before an event.
Finance and logistics partner
Requires contribution-margin impact after shipping, handling, and markdown risk.
03 / EVIDENCE PLAN
The operating truth must be reconstructed.
The guided evidence room combines structured data, policies, interviews, and operational records. Every material conclusion must remain traceable to source evidence and freshness.
SKU-store-day sales, lost-sales estimates, returns, channel mix, and promotion flags.
On-hand, reserved, in-transit, damaged, safety stock, and reconciliation timestamps.
Lane costs, lead times, minimum quantities, carrier capacity, and receiving calendars.
Price, unit cost, markdown schedule, carrying cost, and margin by channel.
Shelf capacity, backroom limits, labor availability, and local assortment rules.
Planned campaigns, regional events, school calendars, and weather-sensitive demand.
Prior recommendations, override reasons, actual outcomes, and follow-up notes.
Searches, abandoned carts, pickup demand, ship-from-store orders, and substitutions.
EXPECTED DATA DEFECTSLate inventory feedsFalse stockoutsPromotion leakageSKU substitutionsUnrecorded local eventsOverride selection bias
04 / REQUIRED BUILD
A thin slice that can change a real decision.
The build must connect evidence, logic, human authority, failure handling, and measurement. A model or dashboard alone is not a complete intervention.
Canonical SKU-location model
Reconcile inventory, sales, demand, economics, store constraints, and transfer lanes at a shared decision grain.
Demand range, not one forecast
Produce a central estimate with uncertainty, sparse-history warnings, and event-adjustment visibility.
Candidate transfer generator
Identify feasible source-destination pairs before optimization and explain why each pair qualified.
Constraint-aware recommendation
Balance expected recovered margin against transfer cost, lead time, safety stock, labor, and receiving limits.
Manager review workflow
Show evidence, confidence, tradeoffs, and structured override reasons without making local approval meaningless.
Pilot and monitoring layer
Track acceptance, execution, stockouts, markdowns, margin, reversals, and unintended concentration effects.
05 / STAGED COMPLICATION
A regional event appears nowhere in the history.
The complication is released only after the learner commits the first problem frame, architecture, and evaluation plan.
REVEAL CASE COMPLICATION+
A large youth tournament is announced near three stores after the first recommendation run. Search activity and manager reports indicate a likely demand spike, but the signal is noisy and the event has no historical analogue. The initial model recommends transferring relevant products out of one affected store because its recent sales were weak.
Required response: Revise the evidence model, uncertainty treatment, manager workflow, and pilot thresholds. Decide whether to accept a manual event adjustment, create a bounded demand scenario, pause the transfer, or require additional evidence.
Manager adjustment
Uses local knowledge quickly but can introduce inconsistent or self-serving assumptions.
Scenario range
Makes uncertainty visible but may produce a wider set of ambiguous recommendations.
Pause and observe
Avoids a damaging transfer but may miss the execution window.
External signal
Adds search or event evidence but raises reliability, licensing, and repeatability questions.
06 / EVALUATION
Value and harm must be measured together.
The final deployment decision must use predeclared technical, operational, adoption, financial, and risk measures. The strongest metric cannot erase a critical failure.
| Measure | What it tests | Target behavior |
|---|---|---|
| Recovered contribution margin | Incremental margin after transfer, handling, and markdown effects. | Improve against current planning baseline. |
| Stockout and lost-sales rate | Whether destination availability improves for relevant demand. | Reduce without draining source stores. |
| Transfer efficiency | Share of executed transfers that create positive net value. | Avoid movement that merely shifts inventory. |
| Override quality | Whether accepted and rejected recommendations are supported by later outcomes. | Learn without punishing justified overrides. |
| Adoption and execution | Recommendation review, acceptance, modification, and completion. | Measure workflow use, not dashboard visits. |
| Concentration risk | Whether the system repeatedly favors the same stores or regions. | Detect self-reinforcing allocation patterns. |
The recommendation is expressed as a decision with expected value and uncertainty, not a forecast alone.
Transfer feasibility is validated before optimization.
Local overrides require a reason and become reviewable evidence.
Margin calculations include shipping, handling, markdown, and lost-sales assumptions.
The pilot includes a current-process comparison and predeclared stop conditions.
The design can recommend no transfer when evidence is weak or value is negative.
HOW TO WORK
Bring your own agent. Keep the evidence chain visible.
The lab permits agent-assisted investigation and implementation. Strong work records material agent recommendations, checks them against authorized evidence, and documents what the learner accepted, changed, rejected, or left unresolved.
Give bounded context.
Provide the mission objective, approved evidence, constraints, and required output rather than asking for a generic solution.
Ask for alternatives.
Require multiple hypotheses, failure modes, and disconfirming evidence before choosing an intervention.
Trace every material claim.
Check source IDs, calculations, code behavior, policy constraints, and unsupported causal language.
Own the recommendation.
Record why the final decision follows from the evidence and what would cause it to change.
07 / REQUIRED DELIVERABLES
The complete record of the deployment decision.
The reviewed lab will score evidence traceability, technical judgment, implementation quality, risk handling, operating readiness, and the consistency of the final recommendation.
Inventory decision frame
Decision owner, baseline, value hypothesis, constraints, and unacceptable outcomes.
Canonical data and evidence map
SKU, store, inventory, demand, event, economics, and transfer lineage.
Recommendation design
Candidate logic, objective function, constraints, uncertainty, and human review.
Working thin slice
Reproducible recommendation pipeline and inspectable decision interface.
Evaluation and pilot plan
Offline backtest, live comparison, adoption measures, and stop conditions.
Executive recommendation
Launch scope, evidence limits, risks, and scale/revise/stop decision.
GUIDED LAB
Enter the D2V Mock Retail guided lab.
The guided workspace includes the evidence pack, eight deployment stages, a staged complication, notebook, artifact templates, and portfolio export.