Amazon Writing Assessment — 3份模拟练习¶

状态: 📝 草稿

创建日期: 2026-02-18
最后更新: 2026-02-18

背景：Amazon TPM Writing Assessment 通常在 Loop 面试前 2 天发出，要求 48 小时内提交 1-2 页文档。考察的不是文采，而是结构化思维、数据驱动决策、风险识别能力、和清晰沟通。 Amazon 偏好叙事体（6-pager 风格），不是 bullet-point PPT 风格。

通用写作框架¶

1. Problem Statement — 问题定义（用数据说明为什么这是个问题）
2. Proposed Approach — 方案概述（2-3个备选，推荐1个，解释为什么）
3. Milestones & Timeline — 里程碑（不超过5个关键节点）
4. Risks & Mitigations — 风险（至少3个，每个有具体缓解措施）
5. Success Metrics — 成功指标（可测量，有baseline和target）
6. Dependencies & Assumptions — 依赖和假设（明确列出）

写作风格指南¶

禁用模糊表达："significantly improved" → "reduced from 16.5s to 5.4s"
每段开头是结论，后面跟支撑数据（倒金字塔结构）
用"We"而非"I"（Writing Assessment不是面试，是团队视角的文档）
长度控制：1.5-2页，单倍行距，11pt 字体
不要用bullet point列表代替叙述。Amazon 6-pager 文化厌恶 bullet-heavy 文档

Practice 1: 支付系统迁移方案¶

Prompt¶

You are a Senior TPM on the Amazon Payment Products team in Vancouver. The team has been running a legacy payment processing system built 8 years ago that handles $2B+ in monthly transaction volume. The system has become increasingly difficult to maintain — deployment frequency has dropped to once per month, and the mean time to recovery from incidents has risen to 4+ hours (vs. the team's target of 30 minutes).

Leadership has approved a migration to a new microservices-based architecture, but requires the migration to happen with zero downtime and no degradation in transaction success rates (currently 99.97%).

Write a 1-2 page document proposing your program plan for this migration, including timeline, key risks, and how you would measure success.

练习要点¶

这是一个高风险、零容错场景，Amazon最关心的是稳定性
需要展示渐进式迁移思维（strangler fig pattern），而非big-bang cutover
重点突出风险管理（回滚方案、金丝雀部署、流量灰度）
可以借鉴你的双11经验：强弱依赖梳理、1359个预案、灰度+功能开关

你的优势素材¶

双11：0 P1/P2 故障，1359个预案
手淘：自动化性能卡口（CI/CD Gate）防止回退
设计家：渲染成功率 65%→99.95%+

Practice 2: 跨团队 AI 功能发布计划¶

Prompt¶

You are a Senior TPM on the Amazon Catalog team. The team is launching a new GenAI-powered product description generation feature that will automatically create and update product listings for 3rd-party sellers. The feature touches multiple teams: the ML model team (building the LLM), the Catalog platform team (integration), the Seller Experience team (UI), and the Trust & Safety team (content moderation).

The VP wants this launched for the holiday season (November 15th). It's currently June 1st. The ML team has a working prototype but hasn't started production-readiness work. The Trust & Safety team has raised concerns that AI-generated content could violate marketplace policies.

Write a 1-2 page document outlining your program plan, how you would manage the cross-team dependencies, the key risks, and your recommendation on whether the November 15th date is achievable.

练习要点¶

需要展示诚实的时间线评估（可能推荐延期或分阶段发布）
跨4个团队的依赖管理是核心
Trust & Safety 的顾虑不能被忽略——这体现 Customer Obsession
展示分阶段 GA策略：先内部 → beta sellers → 全量

你的优势素材¶

每平每屋：跨业务线（设计师/商家/消费者）三方协同
双11：25个项目集、109个子项目的依赖管理
价值交付机制：DORA指标和交付节奏管理

Practice 3: 事故复盘 + 改进方案¶

Prompt¶

You are a Senior TPM on the Prime Tech team. Last week, a deployment to the Prime membership renewal service caused a 47-minute outage during which 230,000 members could not renew their subscriptions. The root cause was a database schema migration that was not backwards-compatible, combined with a rollback procedure that had never been tested in production.

The VP has asked you to lead the Correction of Errors (COE) process and write a document that: 1. Analyzes the root cause chain (not just the immediate trigger) 2. Proposes systemic improvements to prevent similar incidents 3. Includes a prioritized action plan with owners and timelines

Write a 1-2 page COE document.

练习要点¶

Amazon COE 文化非常严格，不是"写个事后总结"，而是深入到系统性根因
需要做 5-Why 分析，不能停在"工程师犯错"层面
改进措施必须是可验证的机制（mechanism），不能是"加强培训"这种软措施
展示你理解 Amazon 的 "mechanisms over good intentions" 理念

你的优势素材¶

双11故障归因：从"深夜误操作"追踪到"缺乏配置变更审批机制"
手淘：CI/CD性能卡口 = 将人工依赖转化为系统化机制
设计家：故障从2周3次降到半年1次的系统性改进

练习节奏¶

周次	任务	产出
W1	完成 Practice 1 初稿（限时3小时）	1.5页文档
W2	自我review + 修改 Practice 1；开始 Practice 2	定稿 + 初稿
W3	完成 Practice 2 + Practice 3	两份定稿
W4	找 Mock Interview partner review 所有3份	收到反馈

自我检查清单¶

每段开头句是否是该段的核心结论？
是否每个claim都有数据支撑？
风险列表是否包含"非显而易见"的风险？（面试官在看你能不能想到别人想不到的）
时间线是否现实？（过于激进=红flag；过于保守=缺乏urgency）
是否有清晰的success metrics？（可测量、有baseline）
文档是否在2页以内？
是否用了叙事体而非bullet-point？