# Case-study library

Research baseline: **July 28, 2026**

This directory is the shared, longitudinal case layer for the Product
Management Mastery System. A dossier supplies durable context, actors,
evidence, exhibits, decision nodes, and an after-action reveal. Question-bank
files can then test different decisions without repeating or quietly changing
the underlying facts.

The planned v1 library contains **40 dossiers: 20 real documented cases and 20
synthetic or composite cases**. The split is deliberate. Public cases teach
evidence discipline and hindsight control; synthetic cases make it possible to
practice decisions for which companies do not publish the necessary raw data,
including India-specific, regulated, enterprise, vulnerable-user, failure, and
sunset situations.

Current authored seed: **7 of 40 dossiers**, all independently `Reviewed` and
none yet `Verified`; the remaining 33 matrix rows are still `Planned`.

## How to use a dossier

1. Choose a decision node and stop at its stated evidence cutoff.
2. Read the prompt and the initially available exhibit only.
3. Ask for evidence by name: metric definition, segment, timeline, qualitative
   evidence, architecture, economics, policy, or operational capacity.
4. State what you infer and what remains unknown.
5. Make a reversible or irreversible decision with thresholds.
6. Only then read the after-action reveal and retrospective.
7. Score the answer with the primary bank's rubric and record one transfer
   drill in a different product context.

A case can support several banks only when each use has a distinct decision
atom: a different decision verb, evidence state, required output, key
constraint, and rubric. For example, sizing a marketplace, choosing its health
metric, designing a switchback experiment, diagnosing cancellations, and
writing a rollout memo are not the same question even if they share context.

## Evidence rules

### Real documented cases

- A real case is marked **REAL — PUBLIC RETROSPECTIVE** or **REAL —
  HISTORICAL DECISION-FORCING** at the top.
- Every company-specific fact must resolve to a direct first-party engineering
  account, official incident report, product documentation, filing, regulator,
  standard, or peer-reviewed paper by identified participants.
- A retrospective source published after the decision cutoff is withheld from
  the candidate and appears only in the after-action section.
- An author's conclusion is labeled **Interpretation**, not presented as the
  company's motivation or result.
- Missing numbers stay unknown. A dossier never manufactures sample size,
  revenue, user sentiment, statistical significance, internal debate, or
  causal certainty.
- First-party accounts can still be incomplete or self-interested. The evidence
  ledger records that limitation rather than treating publication as proof of
  every possible claim.

### Synthetic and composite cases

- A fictional case is marked **SYNTHETIC — ALL COMPANY, PEOPLE, DATES, AND
  EXHIBITS ARE HYPOTHETICAL** at the top.
- Scenario numbers are internally consistent but are not attributed to a real
  company, regulator, bank, model provider, or population.
- Standards, regulator guidance, and public market anchors may be real. They
  are separately labeled **External anchor** and never used to imply that the
  fictional company complied.
- A composite may borrow a general failure pattern from several documented
  incidents, but it must not use a real company's name or imply that the
  invented sequence happened there.
- A synthetic after-action reveal is one plausible designed outcome, not a
  prediction or industry benchmark.

### Evidence labels used inside every dossier

| Label | Meaning | Candidate treatment |
| --- | --- | --- |
| Public fact | Directly supported by a cited source | May be asserted within the source's scope and date |
| External anchor | Real standard, regulator rule, or public statistic | Apply only after checking jurisdiction and date |
| Interpretation | Author's reasoned reading of public evidence | Challenge it; do not attribute it to the company |
| Hypothetical exhibit | Invented, internally consistent case data | Treat as true for the exercise only |
| Unknown | Evidence is absent from the public record or prompt | Ask for it, bound it, or make a labeled assumption |

## Forty-case coverage matrix

Status is `Drafted` only when a full dossier exists. `Reviewed` additionally
requires an independent evidence, reasoning, arithmetic, source-scope,
hindsight, and duplication pass; it still does not mean `Verified`. A planned
real case names the public event or product account that must be verified
before authoring; its row is not evidence that the interpretation has already
been checked.

| ID | Working title | Form | Archetype / sector | Region | Lifecycle and decision center | Primary bank links | Status |
| --- | --- | --- | --- | --- | --- | --- | --- |
| [CS-001](CS-001.md) | Uber marketplace balance and switchback evaluation | Real public retrospective; Uber first-party sources only | Two-sided mobility marketplace, ML | Global | Mature optimization; matching policy, interference, rollout | [PLT](../04-question-banks/PLT-platforms-marketplaces-ecosystems/BLUEPRINT.md), [EXP](../04-question-banks/EXP-experimentation-causal-inference/BLUEPRINT.md), [AIP](../04-question-banks/AIP-ai-ml-generative-ai-product/BLUEPRINT.md), [MET](../04-question-banks/MET-metrics-kpi-product-analytics/BLUEPRINT.md) | Reviewed |
| [CS-002](CS-002.md) | Cloudflare's July 2, 2019 WAF outage | Real public retrospective; official postmortem | Global infrastructure, security, developer platform | Global | Reliability incident; containment, recovery, systemic prevention | [RCA](../04-question-banks/RCA-root-cause-product-diagnosis/BLUEPRINT.md), [TEC](../04-question-banks/TEC-technical-systems-apis-data/BLUEPRINT.md), [EXE](../04-question-banks/EXE-execution-delivery-product-ops-lifecycle/BLUEPRINT.md), [COM](../04-question-banks/COM-product-communication-writing/BLUEPRINT.md) | Reviewed |
| [CS-003](CS-003.md) | NadiPay multilingual UPI launch and failure diagnosis | Synthetic; official NPCI/RBI anchors | Consumer fintech, bank-partner ecosystem | India | Launch plus live RCA; staged rollout, reconciliation, redress | [GTM](../04-question-banks/GTM-go-to-market-launch-product-marketing/BLUEPRINT.md), [RCA](../04-question-banks/RCA-root-cause-product-diagnosis/BLUEPRINT.md), [TSP](../04-question-banks/TSP-trust-safety-privacy-security-accessibility/BLUEPRINT.md), [TEC](../04-question-banks/TEC-technical-systems-apis-data/BLUEPRINT.md) | Reviewed |
| [CS-004](CS-004.md) | NorthstarDesk enterprise support copilot | Synthetic; NIST/PAIR evaluation anchors | B2B SaaS, generative AI | North America and Europe | Discovery to limited beta; evals, human control, enterprise readiness | [AIP](../04-question-banks/AIP-ai-ml-generative-ai-product/BLUEPRINT.md), [B2B](../04-question-banks/B2B-enterprise-product-management/BLUEPRINT.md), [DIS](../04-question-banks/DIS-discovery-research-validation/BLUEPRINT.md), [TSP](../04-question-banks/TSP-trust-safety-privacy-security-accessibility/BLUEPRINT.md) | Reviewed |
| CS-005 | Netflix experimentation and recommendation learning | Real historical/public retrospective; Netflix first-party research required | Consumer subscription, recommendations | Multi-region | Scale optimization; metric choice and long-term effects | EXP, MET, GRW, AIP | Planned |
| CS-006 | GOV.UK service transformation under the Service Standard | Real historical decision-forcing; GOV.UK evidence required | Public digital service | United Kingdom | Discovery-to-migration; public value, inclusion, service quality | DIS, PDS, TSP, EXE | Planned |
| CS-007 | GitHub public incident and service recovery | Real public retrospective; GitHub status/engineering evidence required | Developer platform, B2B SaaS | Global | Reliability incident; dependency, communication, prevention | RCA, TEC, COM, B2B | Planned |
| CS-008 | Stripe API evolution and compatibility promises | Real public retrospective; Stripe documentation required | Payments API and developer platform | Global | Mature platform; versioning, migration, developer trust | TEC, PLT, EXE, TSP | Planned |
| CS-009 | Figma multiplayer architecture and product constraints | Real public retrospective; Figma engineering evidence required | Collaborative B2B/consumer software | Global | Scale; real-time collaboration, quality, technical trade-offs | TEC, PDS, B2B, STR | Planned |
| CS-010 | Spotify Backstage from internal tool to ecosystem | Real public retrospective; Spotify first-party evidence required | Developer platform, open ecosystem | Global | Zero-to-platform; adoption and governance | PLT, STR, GTM, LDR | Planned |
| CS-011 | Etsy search or recommendation experimentation | Real historical decision-forcing; Etsy engineering evidence required | Commerce marketplace | North America / global | Mature optimization; buyer-seller outcomes and experiment guardrails | EXP, PLT, MET, GRW | Planned |
| CS-012 | Airbnb marketplace experiment interference | Real public retrospective; Airbnb first-party research required | Travel marketplace | Global | Scale; network interference and geographic randomization | EXP, PLT, MET, STR | Planned |
| CS-013 | Duolingo product metric and learning-outcome tension | Real public retrospective; Duolingo first-party evidence required | Consumer education subscription | Global | Growth optimization; engagement versus learning quality | GRW, MET, TSP, MON | Planned |
| CS-014 | Slack infrastructure migration and customer continuity | Real public retrospective; Slack engineering evidence required | B2B communication SaaS | Global | Migration; reliability, sequencing, enterprise communication | TEC, EXE, B2B, COM | Planned |
| [CS-015](CS-015.md) | GitLab's public product operating model | Real public retrospective; GitLab handbook evidence required | B2B developer SaaS | Distributed/global | Operating model; decision rights, planning, product craft | [LDR](../04-question-banks/LDR-product-leadership-portfolio-operating-model/BLUEPRINT.md), [EXE](../04-question-banks/EXE-execution-delivery-product-ops-lifecycle/BLUEPRINT.md), [COM](../04-question-banks/COM-product-communication-writing/BLUEPRINT.md), [BEH](../04-question-banks/BEH-behavioral-stakeholder-leadership/BLUEPRINT.md) | Reviewed |
| [CS-016](CS-016.md) | Google's satellite decommission postmortem | Real public retrospective; Google SRE source | Edge infrastructure | Global | Incident and recurrence; blameless learning and automation | [RCA](../04-question-banks/RCA-root-cause-product-diagnosis/BLUEPRINT.md), [TEC](../04-question-banks/TEC-technical-systems-apis-data/BLUEPRINT.md), [LDR](../04-question-banks/LDR-product-leadership-portfolio-operating-model/BLUEPRINT.md), [BEH](../04-question-banks/BEH-behavioral-stakeholder-leadership/BLUEPRINT.md) | Reviewed |
| CS-017 | Mozilla telemetry and privacy choices | Real historical decision-forcing; Mozilla evidence required | Open-source browser | Global | Measurement architecture; consent, data minimization, trust | MET, TSP, STR, COM | Planned |
| [CS-018](CS-018.md) | Kubernetes API deprecation and ecosystem migration | Real historical decision-forcing; official Kubernetes sources | Open-source platform/API | Global | Deprecation; compatibility, governance, partner migration | [TEC](../04-question-banks/TEC-technical-systems-apis-data/BLUEPRINT.md), [PLT](../04-question-banks/PLT-platforms-marketplaces-ecosystems/BLUEPRINT.md), [EXE](../04-question-banks/EXE-execution-delivery-product-ops-lifecycle/BLUEPRINT.md), [COM](../04-question-banks/COM-product-communication-writing/BLUEPRINT.md) | Reviewed |
| CS-019 | US Digital Service recovery of a public digital service | Real historical decision-forcing; official government evidence required | Public service / health | United States | Turnaround; delivery, reliability, organizational coordination | EXE, RCA, LDR, TSP | Planned |
| CS-020 | Wikimedia product change and community governance | Real historical decision-forcing; Wikimedia evidence required | Nonprofit knowledge platform | Global | Mature ecosystem; community consent, metrics, rollback | PLT, BEH, TSP, COM | Planned |
| CS-021 | CoWIN public vaccination platform at national scale | Real public retrospective; official Government of India evidence required | Public health digital service | India | National launch and scale; access, operations, public value | EXE, TSP, MET, LDR | Planned |
| CS-022 | ONDC open-commerce network launch and governance | Real historical/public retrospective; official ONDC and Government of India evidence required | Open commerce network / marketplace | India | Cold start and ecosystem scale; governance, liquidity, trust | PLT, STR, GTM, TSP | Planned |
| CS-023 | Low-bandwidth school learning app | Synthetic | Public/education mobile product | Sub-Saharan Africa | Zero-to-one; offline design, evidence, public value | PDS, DIS, TSP, MET | Planned |
| CS-024 | Creator platform take-rate migration | Synthetic | Creator marketplace | Latin America | Monetization change; fairness, churn, communication | MON, PLT, COM, STR | Planned |
| CS-025 | Smart-meter hardware recall | Synthetic | IoT hardware and field operations | European Union | Reliability/recall; safety, inventory, service continuity | RCA, EXE, TEC, TSP | Planned |
| CS-026 | Climate-accounting SaaS data-quality crisis | Synthetic | B2B SaaS, regulated reporting | Multi-region | Trust turnaround; lineage, auditability, customer renewal | B2B, MET, TEC, COM | Planned |
| CS-027 | Neobank account-scam controls and appeals | Synthetic | Consumer fintech | United Kingdom | Mature risk intervention; fraud, false positives, redress | TSP, MET, EXP, PDS | Planned |
| CS-028 | Food-delivery city-entry portfolio | Synthetic | Three-sided local marketplace | Southeast Asia | Market entry; sizing, density, incentives, operations | STR, GTM, GUE, PLT | Planned |
| CS-029 | Warehouse picking copilot for frontline workers | Synthetic | Enterprise AI, physical operations | North America | Discovery and rollout; worker safety, latency, change management | AIP, B2B, DIS, BEH | Planned |
| CS-030 | Payroll platform pricing and package migration | Synthetic | B2B SaaS / fintech | India and Southeast Asia | Monetization; value metric, contracts, migration | MON, B2B, COM, GTM | Planned |
| CS-031 | Public-transit disruption assistance | Synthetic | Public mobile service | India | Product redesign; accessibility, multilingual alerts, operations | PDS, TSP, MET, EXE | Planned |
| CS-032 | Developer API breaking-change negotiation | Synthetic | B2B developer platform | Global | Migration; partner governance, deprecation, enterprise commitments | TEC, PLT, B2B, BEH | Planned |
| CS-033 | Subscription wellness app retention turnaround | Synthetic | Consumer subscription | North America | Turnaround; cohort diagnosis, ethical growth, pricing | GRW, RCA, MON, TSP | Planned |
| CS-034 | Insurance claims document agent | Synthetic | Regulated enterprise AI | European Union | Limited automation; evals, human appeal, audit, regulation | AIP, B2B, TSP, TEC | Planned |
| CS-035 | Community moderation policy and appeals redesign | Synthetic | Social/content platform | Multi-region | Trust crisis; enforcement quality, appeals, governance | TSP, PDS, MET, COM | Planned |
| CS-036 | Last-mile electric fleet capacity bet | Synthetic | Marketplace plus hardware/operations | India | Scale investment; capacity estimate, unit economics, reliability | GUE, MON, PLT, STR | Planned |
| CS-037 | Hospital scheduling system migration | Synthetic | Enterprise health software | Middle East | High-stakes migration; workflows, reliability, stakeholder alignment | B2B, EXE, TSP, BEH | Planned |
| CS-038 | Portfolio reset after acquisition | Synthetic | Multi-product B2B company | Global | Portfolio strategy; resource allocation, org design, sunset | LDR, STR, PRI, BEH | Planned |
| CS-039 | Children's voice assistant safety redesign | Synthetic | Consumer AI / hardware | Multi-region | Trust redesign; parental control, evals, data minimization | AIP, TSP, PDS, DIS | Planned |
| CS-040 | Responsible sunset of a nonprofit messaging service | Synthetic | Nonprofit communication service | Emerging markets | Sunset; continuity, export, partner transition, communication | EXE, LDR, COM, TSP | Planned |

## Coverage audit

### Form and evidence

- 20 real documented cases: CS-001, CS-002, CS-005 through CS-022.
- 20 synthetic cases: CS-003, CS-004, and CS-023 through CS-040.
- Real dossiers default to Tier A first-party engineering, official
  documentation, government, standards, filings, or participant research.
- Synthetic dossiers must contain at least one authoritative external anchor
  when the exercise involves law, safety, payments, accessibility, security,
  or AI governance.

### Product archetype

The matrix includes consumer subscription, B2B SaaS, marketplaces, platforms
and APIs, AI-native and AI-assisted products, public and nonprofit services,
fintech, health, education, hardware/IoT, physical operations, open-source
ecosystems, and mixed models. No archetype is treated as a cosmetic brand tag:
each changes participants, incentives, failure modes, or required artifacts.

### Geography

The plan includes global and multi-region systems; India and South Asia; other
emerging markets in Africa, Latin America, Southeast Asia, and the Middle East;
and jurisdiction-specific North American, UK, and EU decisions. Every legal
claim must name the applicable jurisdiction and evidence date.

### Lifecycle and outcome balance

The library deliberately includes zero-to-one discovery, launch, scale,
experimentation, pricing, migration, reliability incidents, trust crises,
turnarounds, portfolio resets, recalls, and sunsets. It includes negative,
neutral, and ambiguous outcomes—not only celebrated launches.

### Competency coverage

All twenty bank codes appear in the matrix. `TSP`, `MET`, `COM`, and `TEC`
appear across domains rather than being quarantined as specialist topics.
`BEH` and `LDR` have system-level decision nodes, while `GUE`, `RCA`, `PDS`,
and `EXP` retain their bank-specific reasoning requirements.

## Dossier quality gate

A dossier is ready for question-bank use only when it has:

- stable ID, form label, evidence cutoff, verification date, and source tier;
- actors, incentives, value flow, architecture boundaries, and constraints;
- a dated timeline and at least three genuine decision nodes;
- an evidence ledger that separates fact, anchor, interpretation, hypothetical
  exhibit, and unknown;
- an exact metric tree and definitions, not a list of dashboard labels;
- alternatives including do-nothing and non-product responses;
- failure modes, trust/inclusion review, economics and operations where
  material;
- an initially available exhibit, requested evidence, and after-action reveal;
- a retrospective distinguishing documented outcome from author inference;
- interview prompts mapped to distinct bank decisions;
- direct links, access date, evidence tier, and source limitations.

Planned rows do not pass this gate. `Drafted` means the structure exists and
facts are source-traceable; `Reviewed` and `Verified` require independent
content and citation review under the corpus delivery contract.