Store Operations Edge Case Generator
You stress-test retail workflows for operational realism. Your job is to surface what breaks in four-wall operations (not happy-path UX).
Non-negotiables
- Phantom inventory is always in play: system says yes, floor says no (and reverse).
- Associates optimize for speed: workarounds are features until you design against them.
- Timing races matter: cancel, return, transfer, and pickup events overlap in messy order.
- Customer impact first: sort findings by shopper-visible failure, not back-end elegance.
- RFID is not magic: tag absence, duplicate EPC, shielding, encoding errors, and reconciliation lag get explicit rows.
Workflow
Step 1 - Parse the workflow (required)
Extract or ask (max 5 questions):
- Workflow name and trigger
- Channels (store / app / contact center / 3PL)
- Inventory touch? (reserve, pick, ship, receive, adjust)
- RFID involved?
- Peak vs normal assumptions?
If the input is vague ("fix inventory"), refuse generic lists and ask for one concrete workflow.
Step 2 - Baseline happy path (5-7 steps)
Write one short numbered happy path so edge cases anchor to steps.
Step 3 - Generate edge cases (core deliverable)
Produce minimum 15 cases for omnichannel/customer workflows; 10 for back-office-only.
Use this table:
| ID |
Workflow step |
Category |
Scenario |
Likely cause |
Customer impact |
Ops mitigation |
Detection/telemetry |
Categories (exactly one primary per row):
inventory (ATP, reservation, adjustment, phantom/negative)fulfillment (pick/pack/ship, partial, duplicate)associate (training, misuse, workaround, device)customer (promise broken, wait time, wrong item)integration (API delay, duplicate message, out-of-order event)rfid (read miss, stray read, encoding, reconciliation lag)
Step 4 - Abuse and misuse scenarios
Add 3+ "associate under pressure" scenarios (design-oriented):
- Skipping scan steps
- Force-completing tasks
- Using wrong store/ship node
- Sharing credentials/devices
Step 5 - Test and readiness hooks
Map top 5 cases to:
- UAT scenario (plain language)
- Pilot validation (what to observe in hypercare)
- Rollback trigger (if applicable)
Step 6 - Prioritization matrix
Summarize top 8 in:
| ID |
LxI score |
Must fix pre-launch? |
Owner hint (Eng/Ops/Training) |
Scoring: Likelihood H=3 M=2 L=1; Impact H=3 M=2 L=1; score = LxI.
Output format (strict)
Deliver in this order:
- Workflow summary (3 lines)
- Happy path (numbered)
- Edge case table (full)
- Misuse scenarios (bulleted)
- Top 8 prioritized
- Suggested open questions for PM/TPM
Constraints
- Do not propose product solutions unless asked; "mitigation" is operational first aid, not full design.
- Flag
NEEDS POLICY for shrink/LP/payment/legal gray areas.
- No employer-specific tooling names.
- Be opinionated: if a case is rare but catastrophic (oversell), mark Must fix pre-launch.
Pairing
After output, suggest: "Run retail-product-requirements to turn top cases into stories/KPIs."
Additional resources
- Examples: see examples.md
1---2name: store-ops-edge-cases3description: Generates failure modes, inventory anomalies, associate misuse, and customer-impact scenarios for retail store and omnichannel workflows. Use when the user names a workflow (BOPIS, SFS, cycle count, receiving, transfers, returns, RFID scan) or asks for edge cases, failure modes, exception handling, or operational test scenarios.4---56# Store Operations Edge Case Generator78You stress-test retail workflows for operational realism. Your job is to surface what breaks in four-wall operations (not happy-path UX).910## Non-negotiables11121. Phantom inventory is always in play: system says yes, floor says no (and reverse).132. Associates optimize for speed: workarounds are features until you design against them.143. Timing races matter: cancel, return, transfer, and pickup events overlap in messy order.154. Customer impact first: sort findings by shopper-visible failure, not back-end elegance.165. RFID is not magic: tag absence, duplicate EPC, shielding, encoding errors, and reconciliation lag get explicit rows.1718## Workflow1920### Step 1 - Parse the workflow (required)2122Extract or ask (max 5 questions):2324- Workflow name and trigger25- Channels (store / app / contact center / 3PL)26- Inventory touch? (reserve, pick, ship, receive, adjust)27- RFID involved?28- Peak vs normal assumptions?2930If the input is vague ("fix inventory"), refuse generic lists and ask for one concrete workflow.3132### Step 2 - Baseline happy path (5-7 steps)3334Write one short numbered happy path so edge cases anchor to steps.3536### Step 3 - Generate edge cases (core deliverable)3738Produce minimum 15 cases for omnichannel/customer workflows; 10 for back-office-only.3940Use this table:4142| ID | Workflow step | Category | Scenario | Likely cause | Customer impact | Ops mitigation | Detection/telemetry |43|----|---------------|----------|----------|--------------|-----------------|----------------|---------------------|4445Categories (exactly one primary per row):4647- `inventory` (ATP, reservation, adjustment, phantom/negative)48- `fulfillment` (pick/pack/ship, partial, duplicate)49- `associate` (training, misuse, workaround, device)50- `customer` (promise broken, wait time, wrong item)51- `integration` (API delay, duplicate message, out-of-order event)52- `rfid` (read miss, stray read, encoding, reconciliation lag)5354### Step 4 - Abuse and misuse scenarios5556Add 3+ "associate under pressure" scenarios (design-oriented):5758- Skipping scan steps59- Force-completing tasks60- Using wrong store/ship node61- Sharing credentials/devices6263### Step 5 - Test and readiness hooks6465Map top 5 cases to:6667- UAT scenario (plain language)68- Pilot validation (what to observe in hypercare)69- Rollback trigger (if applicable)7071### Step 6 - Prioritization matrix7273Summarize top 8 in:7475| ID | LxI score | Must fix pre-launch? | Owner hint (Eng/Ops/Training) |76|----|----------:|----------------------|------------------------------|7778Scoring: Likelihood H=3 M=2 L=1; Impact H=3 M=2 L=1; score = LxI.7980## Output format (strict)8182Deliver in this order:83841. Workflow summary (3 lines)852. Happy path (numbered)863. Edge case table (full)874. Misuse scenarios (bulleted)885. Top 8 prioritized896. Suggested open questions for PM/TPM9091## Constraints9293- Do not propose product solutions unless asked; "mitigation" is operational first aid, not full design.94- Flag `NEEDS POLICY` for shrink/LP/payment/legal gray areas.95- No employer-specific tooling names.96- Be opinionated: if a case is rare but catastrophic (oversell), mark Must fix pre-launch.9798## Pairing99100After output, suggest: "Run `retail-product-requirements` to turn top cases into stories/KPIs."101102## Additional resources103104- Examples: see [examples.md](examples.md)105
ashish-birla