agentleFS
zerlinpizerlinpi/amazon-ads-skills/skills/search-term-analysis/SKILL.md

search-term-analysis

Skill8 starsChanged 53 days agoMITView on GitHub

Analyze Amazon Ads Search Term rows for origin, report coverage, query/traffic quality, winner and harvest-candidate evidence. Use when the primary task is interpreting Search Term evidence; route action-safe negative type/scope to negative-targeting and configured keyword/target lifecycle or structure to keyword-optimization.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 0 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When a conclusion depends on live MCP/API/connector data, load ../../references/connector-capability.md before interpreting missing or …

    line 18
  2. Very specific:

    When winner, waste, harvest, or scale interpretation depends on the campaign's job, load ../../references/campaign-objective.md.

    line 36
  3. Very specific:

    Do not infer campaign objective from Auto/Broad/Phrase/Exact or the displayed search-term string alone.

    line 46
  4. Very specific:

    Search Term report row 不一定等同于字面意义上的真实用户查询。Amazon 当前 Sponsored Products 文档说明,search term 可以是顾客用于搜索商品的词,也可以是在 non-search …

    line 50

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Search Term Analysis

Connector capability gate

When a conclusion depends on live MCP/API/connector data, load ../../references/connector-capability.md before interpreting missing or empty fields. If a machine-readable capability snapshot is available, identify the exact required capability IDs and run ../../scripts/evaluate_connector_capability_gate.py before metric interpretation.

  • Pass — continue with the Skill's normal evidence, sufficiency, lineage and safety checks.
  • Degraded — do not make a high-confidence dependent recommendation; keep the result to Directional, Hold, Alternate Source, Missing Data, or Manual Review.
  • Blocked — do not treat the dependent observation as action-safe until the capability is resolved or an allowed alternate source is verified.
  • missing_evidence_policy = never_zero — Partial, Unsupported, Unknown, or absent connector capability is never a numeric zero, unchanged state, or proof that the platform lacks the feature.

This gate is read-only and does not authorize live Amazon Ads mutation.

Routing boundary

  • 本 Skill 负责 Search Term row 的解释层:term_origin、report/row coverage、query/traffic quality、赢家/浪费证据,以及 harvest / negative 的候选筛选。
  • 如果用户的主要目标是选择 action-safe Negative Exact / Negative Phrase / Negative Product Targeting 的类型、scope 与误杀保护,路由到 ../negative-targeting/SKILL.md;本 Skill 提供证据,不复制否定动作的最终控制决策。
  • 如果主要目标是 configured keyword/target 的结构、match type、state、lifecycle 或 traffic ownership,路由到 ../keyword-optimization/SKILL.md。
  • 若同一任务同时需要解释 Search Term row 和选择否定动作,先完成本 Skill 的 origin/coverage/actionability gate,再把结论交给 negative-targeting;不要把两个 Skill 当成对同一问题的平行意见。

Campaign objective gate

When winner, waste, harvest, or scale interpretation depends on the campaign's job, load ../../references/campaign-objective.md.

  • Discovery traffic may remain useful while it produces decision-relevant query/target learning inside explicit guardrails.
  • Control traffic emphasizes clear ownership, isolation and stable measurement.
  • Growth traffic may use different short-term trade-offs only when the growth objective and guardrails are explicit.
  • Profit traffic should respect contribution economics and target efficiency.
  • Defense traffic requires strategic-coverage and cannibalization context.
  • Experiment traffic must preserve treatment integrity.
  • Unknown campaign objective prevents role-specific harvest/scale claims from becoming action-safe.

Do not infer campaign objective from Auto/Broad/Phrase/Exact or the displayed search-term string alone.

核心原则

Search Term report row 不一定等同于字面意义上的真实用户查询。Amazon 当前 Sponsored Products 文档说明,search term 可以是顾客用于搜索商品的词,也可以是在 non-search context(例如部分站外社交展示)中由系统推断出的 best match。因此先判断 term_origin,再决定能否把该 row 当成 literal shopper query 做意图、Exact 收割或 Negative 推理。

建议把 term_origin 标记为:

  • literal_query — 有可靠上下文证明来自 shopper search query;
  • inferred_non_search — 有可靠证据表明来自 non-search / inferred-best-match context;
  • unknown — 当前 report/connector 没有提供足够 origin/placement/context 证据。

unknown 不自动等于 literal shopper query。

如果结论依赖“报表是否代表完整搜索词总体”、缺失行、impression/CTR coverage、全账户 Top/Bottom 排名或跨报表 reconciliation,按需加载 ../../references/report-coverage.md。先确认 report 的 row-inclusion / eligibility contract,再解释缺失行或总体覆盖。

如果问题涉及 Search Term Impression Share (SIS)、Impression Rank、share-of-voice、竞争可见度或 query growth headroom,按需加载 ../../references/search-term-impression-share.md。SIS 用作市场可见度/份额证据,不单独证明增量销量或应该提高 bid/budget。

默认模式

Suggest。禁止自动添加 negative keyword/target。

输入

需要尽量包含:

  • search_term;
  • campaign_id / ad_group_id;
  • matched_target_id / match_type;
  • impressions、clicks、spend、orders、sales;
  • 日期、币种、Marketplace;
  • Target ACOS/ROAS 或业务目标;
  • 可选:term_origin、placement/context、是否为 on-Amazon search / non-search / off-Amazon context;
  • 可选:report_type、row-inclusion / eligibility 规则、分页/完整性状态;
  • 可选:Search Term Impression Share、Impression Rank、对应窗口/time unit 和 account scope;
  • 可选:历史窗口、品牌词列表、核心类目词、商品价格/利润。

覆盖边界

如果数据来自官方 Sponsored Products Search Term report,当前 Amazon Ads 文档说明该报告只包含所选期间至少产生 1 次广告点击的 search-term rows。因此:

  • 报表缺失的 search term 不自动等于 0 impressions / 0 clicks;
  • 搜索词报表中的 impressions 合计不要求与 Campaign Manager / campaign 总 impressions 完全一致;
  • 全账户 query impression coverage、无点击 query 浪费、完整 query CTR denominator 等问题需要兼容的额外来源;
  • top/bottom search terms 默认只代表当前报表已表示的 clicked population,除非完整总体覆盖另有证明。

另外,row 已出现也不自动证明它是 literal shopper query。对于 inferred non-search rows,点击和转化表现仍然可以用于流量质量/效率分析,但字面 query 意图、Exact keyword 收割和 Negative keyword 动作需要更强的 origin/context 证据。

分类

Winner / Harvest candidate

通常具备:

  • 有稳定转化;
  • 效率达到或优于目标;
  • 样本不是一次偶然订单;
  • term_origin 足以支持把该字符串解释为 literal shopper query;
  • 查询意图与商品高度相关。

若 term_origin = inferred_non_search 或 unknown,可保留为高价值 traffic/match signal,但不要仅凭字符串自动升级为 Exact keyword。优先输出 Manual Review 或请求可解释的 search-context / placement / targeting evidence。

建议:在 literal-query 证据成立时考虑 Exact 收割、单独控制 bid/budget、避免重复争抢。

Growth candidate

表现好但曝光/点击样本偏少。建议增加观察、提高可控性或测试扩量,不急于做大幅调整。若同时存在 SIS,则先判断 share headroom 与实际 binding constraint,再决定是 bid、budget、placement、routing 还是仅继续观察。

Explore

相关性合理但样本不足,或 row origin 尚不足以支持 literal-query interpretation。保留探索,不因为短期无单直接否定。

Waste candidate

花费/点击样本已足够且长期明显偏离目标,且没有品牌/战略价值。若 row 可能来自 inferred non-search context,优先检查实际 targeting/placement/traffic source,不把字面字符串直接当成应该屏蔽的 shopper query。

Negative candidate

同时满足:

  • term_origin / traffic context 足以证明 Negative keyword/target 与实际可控流量语义匹配;
  • 查询意图明显不相关,或已有足够证据显示长期低价值;
  • 否定不会误伤核心流量;
  • 已检查品牌词、历史赢家、归因延迟;
  • 已确认合适的 Negative Exact / Phrase / product-target 类型。

如果 row 为 inferred_non_search 或 unknown,不要仅根据显示字符串生成 action-safe Negative keyword。应先确认真正可控的 target/placement/context;证据不足时输出 Manual Review。

收割工作流

  1. 找到高价值 search-term row;
  2. 判定 term_origin;
  3. 只有 literal shopper query 证据足够时,检查是否已经存在 Exact keyword/target;
  4. 如不存在,再提出“Exact 收割”建议;
  5. 检查原流量入口是否需要保留探索;
  6. 只有在流量路由明确时,才考虑对源入口做 Negative Exact 以避免重复;
  7. 输出迁移动作顺序,避免先否定导致流量中断。

否词保护

在建议 negative 前必须回答:

  • 这是 literal shopper query,还是 inferred non-search match / unknown origin?
  • 是品牌词吗?
  • 是核心产品词/类目词吗?
  • 历史是否有订单?
  • 是否处于大促或归因延迟窗口?
  • 是否只是 CPC 高而非意图错误?
  • Negative Phrase 会不会覆盖有价值长尾词?
  • 实际可控对象究竟是 keyword、product target、placement/context,还是当前证据不足?

无法回答时只列为 Manual Review / 人工复核候选。

输出格式

Search-term summary

  • report coverage / row-inclusion status(当影响结论时);
  • term_origin / origin confidence(当影响意图解释或动作时);
  • spend / sales / orders;
  • top winning terms;
  • top waste concentration;
  • intent shifts;
  • SIS / Impression Rank 与 share-headroom classification(如果提供)。

Harvest candidates

每项:search term、term_origin、现有来源、证据、建议目标结构、confidence。

Negative candidates

每项:search term、term_origin、建议 negative type、证据、误伤风险、confidence、是否需要人工复核。需要形成 action-safe negative type/scope 时交给 negative-targeting 完成最终控制决策。

Keep exploring

列出样本不足、origin 不确定或相关性合理的词,解释为什么暂不动作。

Action proposals

使用 ../../schemas/optimization-action.json 的字段。

禁止

  • 不把所有 Search Term report rows 默认描述成 literal shopper queries;
  • 不把 inferred non-search / unknown-origin row 仅按显示字符串机械转成 Exact harvest 或 Negative keyword;
  • 不以固定“20 clicks 0 orders”之类规则作为唯一否词依据;
  • 不把所有高 ACOS search term 都否定;
  • 不在不知道品牌/核心词保护清单时批量输出可执行否词;
  • 不把 clicked-only / delivered-only 报表的缺失行自动补成 0;
  • 不把 selected report subset 无标注地描述成完整账户搜索词总体;
  • 不把低 impression share 单独解释为保证存在可盈利增量,也不从 SIS 直接推导固定 bid/budget 增幅。

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.