platform-strategy（平台策略）

Platform Strategy

Scope

Covers

• Internal platforms (paved roads, shared infrastructure/services) treated as products
• External/hybrid platforms (APIs, extensions, partners) and ecosystem strategy
• Platform lifecycle strategy (when to open vs when to close for control/monetization)
• Platform surface-area design (interfaces, abstractions, governance) to reduce cognitive load for product teams
• AI platform defensibility (context repositories + integrated “toolkit” experiences) when relevant

When to use

• “Create a platform strategy for our developer platform / API.”
• “Turn our internal platform into a product with clear users, metrics, and roadmap.”
• “We want to open our platform to third parties—define incentives, governance, and a rollout.”
• “We’re building an AI platform—what’s the defensible system beyond a single feature?”

When NOT to use

• You don’t have a clear problem/job-to-be-done yet (use problem-definition first).
• You primarily need a product/company strategy and portfolio plan (use ai-product-strategy).
• You’re selecting a vendor/tool rather than defining a platform strategy (use evaluating-new-technology).
• You need an implementation design/architecture doc (use writing-specs-designs after this).

Inputs

Minimum required

• Platform type: internal / external / hybrid (and who “the platform owner” is)
• Primary users/consumers (e.g., developers, data scientists, partners) and their top jobs-to-be-done
• Current state: what exists today (surfaces, APIs, services, docs), and what’s broken/painful
• Business intent: why now, desired outcomes (speed, reliability, revenue, ecosystem leverage), time horizon
• Constraints: security/privacy/compliance, SLOs, regions, budgets, resourcing, dependencies
• Decision context: who decides, what decisions are on the table (open/close, pricing, governance), target date

Missing-info strategy

• Ask up to 5 questions from references/INTAKE.md (3–5 at a time).
• If still missing, proceed with explicit assumptions and present 2–3 strategy options (e.g., internal-only vs partner beta vs public API).
• Do not request secrets or credentials. Require explicit confirmation for any production changes or external outreach.

Outputs (deliverables)

Produce a Platform Strategy Pack (in chat; or as files if requested), in this order:

1. Platform Product Charter (users, jobs, non-goals, assumptions, outcomes)
2. Platform Surface & Interface Map (capabilities, owners, APIs/SDKs, “paved road” defaults, boundaries)
3. Lifecycle Stage & Open/Close Strategy (stage diagnosis, stage-appropriate moves, transition risks)
4. Moat & Ecosystem Model (compounding loops, incentives, seeding plan, investment gates)
5. Governance & Policy Plan (what’s open/closed, SLAs, deprecation, partner rules, pricing/packaging if relevant)
6. Metrics & Operating Model (platform-as-product operating cadence, intake, support, adoption + productivity metrics)
7. 12‑month Roadmap (milestones, bets, sequencing, dependencies)
8. Risks / Open questions / Next steps (always included)

Templates: references/TEMPLATES.md

Workflow (8 steps)

1) Define the platform as a product (users + jobs + outcomes)

• Inputs: Platform context, primary user groups, current pain.
• Actions: Write the platform’s “user promise” and top 3–5 jobs-to-be-done. Add 3–5 non-goals. Choose 2–4 outcome metrics (prefer developer productivity metrics like cycle time).
• Outputs: Draft Platform Product Charter.
• Checks: You can describe value without naming internal components (“We reduce X minutes of toil per deploy”).

2) Diagnose the platform lifecycle stage (and what decisions are truly on the table)

• Inputs: Market/organization conditions, competitive context (if external), timeline.
• Actions: Determine the most likely stage (Step 0→3) and list evidence. Clarify the “open vs close” decision(s) you must make now (not someday).
• Outputs: Lifecycle Stage & Open/Close Strategy (draft).
• Checks: The stage is justified with evidence, not aspiration (“we should be a platform”).

3) Map surface area and define boundaries (reduce decision complexity)

• Inputs: Existing services/APIs, teams, dependency graph, common failures.
• Actions: Inventory platform capabilities; define what becomes a paved road vs optional. Specify boundaries: what platform owns vs domain teams own. Draft interface contracts (APIs/SDKs/events) and “default decisions” the platform makes for others.
• Outputs: Platform Surface & Interface Map.
• Checks: A domain team can build without re-deciding foundational choices (auth, logging, deployment, guardrails).

4) Identify the moat and the compounding loop(s)

• Inputs: Unique assets, distribution, data/context advantages, ecosystem participants.
• Actions: Propose 1–3 moat hypotheses and at least one compounding loop (“if this works, it accelerates”). Define incentives for each participant and a small seeding plan. Define “investment gates” (signals that justify more spend).
• Outputs: Moat & Ecosystem Model.
• Checks: The loop has measurable leading indicators (activation, retained developers, successful integrations).

5) Decide what to open, how to govern it, and how to protect the core

• Inputs: Stage, risks, support capacity, security/compliance requirements.
• Actions: Specify what’s open now vs later, plus governance: access control, quotas, review processes, partner rules, SLAs, deprecation/backwards compatibility, and (if relevant) pricing/packaging. Include an “abuse/quality” plan (observability, enforcement).
• Outputs: Governance & Policy Plan.
• Checks: “Open” surfaces have a sustainability plan (support, docs, incident response, versioning).

6) (If AI platform) Build defensibility as a system, not a feature

• Inputs: AI use cases, context sources, data sensitivity, required integrations.
• Actions: Design a “Swiss‑army toolkit” system: shared context repository + multiple experiences (autocomplete, chat, agent workflows) with consistent policies. Define permissions, audit logs, eval/monitoring, and human-in-the-loop points.
• Outputs: AI section inside Platform Product Charter + Governance & Policy Plan updates.
• Checks: The plan improves outcomes while containing risk (least privilege, auditable access, measurable quality).

7) Define metrics + operating model (platform-as-product)

• Inputs: Target outcomes, resourcing constraints, stakeholder map.
• Actions: Define a metric stack (north-star + input metrics) and an operating cadence (intake, prioritization, roadmap reviews, documentation, support/on-call, feedback loops). Ensure a PM/owner exists for internal platforms.
• Outputs: Metrics & Operating Model.
• Checks: Metrics tie to user outcomes (not vanity counts like “# of services migrated” alone).

8) Sequence the roadmap and quality-gate the pack

• Inputs: All draft artifacts.
• Actions: Create a 12‑month roadmap with 3 horizons (Now / Next / Later). Add dependencies, resourcing, and rollback/exit paths. Run references/CHECKLISTS.md and score with references/RUBRIC.md. Always include Risks / Open questions / Next steps.
• Outputs: Final Platform Strategy Pack.
• Checks: A stakeholder can make a decision (owner, date, next actions) and understand trade-offs.

Quality gate (required)

• Use references/CHECKLISTS.md and references/RUBRIC.md.
• Always include: Risks, Open questions, Next steps.

Examples

Example 1 (internal platform): “Use platform-strategy to create a platform strategy for an internal ML platform used by 40 engineers. Goal: cut model deployment cycle time from 2 weeks to 2 days. Constraints: PII present; SOC2; 2 platform engineers; 6‑month horizon.” Expected: platform-as-product charter + paved-road interfaces + productivity metrics + governance for AI data.

Example 2 (external ecosystem): “Use platform-strategy to define an API platform strategy for opening our analytics product to partners. We want 20 high-quality integrations in 12 months without breaking core reliability.” Expected: stage diagnosis + open/close decisions + incentives + governance/versioning + roadmap.

Boundary example: “We should become a platform like Apple—make us a platform strategy.” Response: out of scope without specific users/jobs and a plausible compounding loop; ask intake questions and/or start with problem-definition.