Back to Details

amz-search-term-report-miner

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Search Term Report Miner

搜索词报告挖掘工具

A Search Term Report from a mature account is 50,000+ rows. Sellers download it, open Excel, scroll for 20 minutes, find 4 negatives, give up. Meanwhile real money leaks: terms with 30+ clicks and 0 orders, exact-match converters stuck in broad campaigns, and dead clusters that drain $200/day. This skill carves the file into 4 surgical buckets and outputs a Bulk Operations upload ready to submit.
成熟账号的搜索词报告（Search Term Report）通常包含5万+行数据。卖家下载后打开Excel，滚动查看20分钟，仅找到4个否定关键词就放弃了。但此时资金正在悄悄流失：有30+点击却0订单的关键词，精准匹配转化词被困在广泛匹配广告活动中，还有每天消耗200美元的无效关键词组。本工具可将文件精准划分为4个分组，并输出可直接提交的Bulk Operations上传文件。

When to use this

使用场景

  • Monthly PPC review and you want a clean negative sweep
  • Account spending $3K+/month on PPC with ACoS drift
  • New product 60 days post-launch ready for the first harvest pass
  • Pre-Q4 cleanup. Push wasted spend out before peak
  • 月度PPC复盘，需要清理否定关键词
  • 账号每月PPC花费3000美元以上，且ACoS出现波动
  • 新品上线60天后，准备首次进行精准匹配词挖掘
  • Q4旺季前清理优化，在峰值到来前减少无效花费

The framework. The Four Buckets

框架：四大分组

Pull the Search Term Report (60-day window). For every search term, route to exactly one bucket based on click and order thresholds.
BucketFilterAction
1. Negative ExactClicks >= 10, Orders = 0, ACoS = blankAdd as Negative Exact at campaign
2. Negative Phrase3+ terms share root, combined clicks >= 25, orders <= 1Add root as Negative Phrase
3. Exact HarvestOrders >= 2, ACoS < target ACoS, term not already exactMove to new exact-match campaign
4. Hero ScaleImpressions >= 1000, CVR well above your account average, ACoS < targetIncrease budget + bid +20%
Anything else: leave alone.
A note on the Hero gate: there is no universal "good" CVR. It varies hugely by category, price, and product. Do not hardcode a fixed number like 12%. Instead, compare each term to YOUR account or category baseline. A practical hero gate is roughly 1.5x to 2x your average CVR for that product. Compute your baseline first, then set the gate from it.
导出搜索词报告（60天周期）。根据点击量和订单量阈值，将每个搜索词精准分配至一个分组。
分组筛选条件操作
1. Negative Exact（否定精确匹配）点击量 >=10，订单量=0，ACoS为空在广告活动层级添加为Negative Exact
2. Negative Phrase（否定短语匹配）3个以上关键词共享词根，总点击量>=25，订单量<=1将词根添加为Negative Phrase
3. Exact Harvest（精确匹配挖掘）订单量>=2，ACoS < 目标ACoS，该关键词尚未设置为精确匹配转移至新的精确匹配广告活动
4. Hero Scale（核心拓展）曝光量>=1000，CVR远高于账号平均值，ACoS < 目标ACoS预算+出价提升20%
其他关键词：保持不变。
关于核心拓展的阈值说明：不存在通用的“优质”CVR，它因品类、价格和产品差异巨大。不要硬编码固定数值（如12%），而是将每个关键词与你的账号或品类基准值对比。实用的核心拓展阈值约为该产品平均CVR的1.5至2倍。先计算基准值，再设置阈值。

Step by step

操作步骤

  1. Download the STR. Advertising > Reports > Sponsored Products > Search Term Report. Date range: last 60 days. Format: CSV.
  2. Set the target ACoS. Usually 25-35% depending on margin. This is the threshold for bucket 3 and 4.
  3. Apply Bucket 1 filter. Clicks >= 10 AND Orders = 0. Export the search term column. Format as Negative Exact rows for Bulk Operations.
  4. Cluster for Bucket 2. Group remaining 0-1 order terms by shared 2-word root. If a root cluster has 25+ combined clicks, add the root as Negative Phrase.
  5. Apply Bucket 3 filter. Orders >= 2 AND ACoS < target. Check each term is not already an exact keyword in any campaign. Format as new exact-match campaign rows.
  6. Apply Bucket 4 filter. Impressions >= 1000 AND CVR well above your account average (roughly 1.5-2x your baseline for that product) AND ACoS < target. These are scale candidates. Generate bid +20% rows.
  7. Build the Bulk Operations file. 4 sheets, one per bucket. Use Amazon Bulk Sheet template columns.
  8. Run the quality check, then upload to Campaign Manager.
  1. 下载STR：广告 > 报告 > 商品推广 > 搜索词报告。时间范围：过去60天。格式：CSV。
  2. 设置目标ACoS：通常根据利润率设置为25-35%，这是分组3和4的阈值。
  3. 应用分组1筛选条件：点击量>=10 且 订单量=0。导出搜索词列，格式化为Bulk Operations所需的Negative Exact行。
  4. 分组2聚类：将剩余的0-1订单关键词按共享双字词根分组。如果某个词根组总点击量>=25，将该词根添加为Negative Phrase。
  5. 应用分组3筛选条件：订单量>=2 且 ACoS < 目标值。确认每个关键词尚未在任何广告活动中设置为精确匹配，格式化为新精确匹配广告活动的行。
  6. 应用分组4筛选条件：曝光量>=1000 且 CVR远高于账号平均值（约为该产品基准值的1.5-2倍）且ACoS < 目标值。这些是可拓展的候选词，生成出价提升20%的行。
  7. 构建Bulk Operations文件：4个工作表，对应4个分组。使用亚马逊批量表模板列。
  8. 运行质量检查，然后上传至广告活动管理器。

Output format

输出格式

undefined
undefined

STR Mining Output. [Account] [date range]

STR Mining Output. [Account] [date range]

Summary
  • Total search terms analyzed: [N]
  • Bucket 1 (Negative Exact): [N] terms, wasted spend last 60d: $[X]
  • Bucket 2 (Negative Phrase): [N] roots
  • Bucket 3 (Exact Harvest): [N] terms, projected exact-match orders/month: [N]
  • Bucket 4 (Hero Scale): [N] terms
Bulk Operations file structure Sheet 1: Negative Exacts Campaign | Ad Group | Negative Keyword | Match Type [rows]
Sheet 2: Negative Phrases [rows]
Sheet 3: Exact Match Harvest. New campaigns Campaign | Ad Group | Keyword | Match Type | Bid [rows]
Sheet 4: Hero Scale. Bid increases Campaign | Ad Group | Keyword | Old Bid | New Bid [rows]
undefined
Summary
  • Total search terms analyzed: [N]
  • Bucket 1 (Negative Exact): [N] terms, wasted spend last 60d: $[X]
  • Bucket 2 (Negative Phrase): [N] roots
  • Bucket 3 (Exact Harvest): [N] terms, projected exact-match orders/month: [N]
  • Bucket 4 (Hero Scale): [N] terms
Bulk Operations file structure Sheet 1: Negative Exacts Campaign | Ad Group | Negative Keyword | Match Type [rows]
Sheet 2: Negative Phrases [rows]
Sheet 3: Exact Match Harvest. New campaigns Campaign | Ad Group | Keyword | Match Type | Bid [rows]
Sheet 4: Hero Scale. Bid increases Campaign | Ad Group | Keyword | Old Bid | New Bid [rows]
undefined

Worked example (illustrative)

示例（仅供参考）

The figures below are an example to show how the buckets shake out, not a benchmark to expect. Your term counts, CVR, ROAS, and savings depend entirely on your account.
Account: home goods seller, 42 SKUs, $8,400/month PPC spend. 60-day STR has 38,200 unique search terms. After parsing:
Bucket 1: 184 negative exact candidates. Wasted spend = $1,847 over 60 days, or $924/month immediately saved.
Bucket 2: 12 phrase roots like "cat", "for dogs", "wholesale". 412 clicks, 8 orders. Adding negative phrases blocks 200+ wasted clicks/month.
Bucket 3: 67 exact-match harvest candidates. In this example the harvested terms convert well above the account average and move to exact campaigns at 1.2x bid. The incremental-orders and ROAS figures you would model here are account-specific. do not assume a fixed order count or a particular ROAS. run the projection off your own term-level CVR and AOV.
Bucket 4: 14 hero terms. +20% bid aims to move them from mid page 1 toward the top of search.
Total projected impact in this scenario combines the negative-keyword savings with the harvested-term upside. Both are illustrative. compute your own numbers from your account data rather than carrying these over.
以下数据仅用于展示分组结果，并非通用基准。你的关键词数量、CVR、ROAS及节省金额完全取决于自身账号情况。
账号：家居用品卖家，42个SKU，每月PPC花费8400美元。60天STR包含38200个独特搜索词。解析后：
分组1：184个Negative Exact候选词。60天内无效花费1847美元，即每月可立即节省924美元。 分组2：12个短语词根，如“cat”、“for dogs”、“wholesale”。总点击量412，订单量8。添加否定短语后，每月可阻止200+无效点击。 分组3：67个精确匹配挖掘候选词。在本示例中，挖掘出的关键词转化率远高于账号平均值，将以1.2倍出价转移至精确匹配广告活动。增量订单量和ROAS数据需根据自身账号的关键词层级CVR和AOV进行建模，不要假设固定订单量或特定ROAS，需基于自身账号数据进行预测。 分组4：14个核心拓展词。出价提升20%旨在将其从搜索结果第1页中部推至顶部。
本场景中的总预估影响结合了否定关键词的节省金额和挖掘词的收益提升，均为示例数据。请基于自身账号数据计算，不要直接套用。

Quality check

质量检查

  • Date range is 60 days, not 30 (need volume for cluster math)
  • Target ACoS set before bucketing, not after
  • Bucket 3 harvest terms confirmed not already exact in any campaign
  • Negative phrases checked for accidental brand cannibalization
  • Bulk file uses correct column headers per Amazon template
  • 时间范围为60天，而非30天（聚类计算需要足够的数据量）
  • 在分组前设置目标ACoS，而非分组后
  • 确认分组3的挖掘词未在任何广告活动中设置为精确匹配
  • 检查否定短语是否存在意外的品牌自损情况
  • 批量文件使用亚马逊模板指定的正确列标题

Common mistakes

常见错误

  • 30-day STR. Not enough volume per term. Bucket 2 cluster math breaks down.
  • No target ACoS. Bucket 3 and 4 lose their gate, you scale unprofitable terms.
  • Adding negative phrases without checking brand terms. Easy to accidentally negative-match your own brand variant.
  • Harvesting terms that already exist as exact. Creates internal bid wars.
  • Skipping Bucket 4. Most sellers focus on cutting waste and miss the scale upside, which is often the larger lever.
  • 使用30天STR：关键词数据量不足，分组2的聚类计算失效。
  • 未设置目标ACoS：分组3和4失去阈值，可能会拓展无利可图的关键词。
  • 未检查品牌词就添加否定短语：容易意外否定自身品牌变体。
  • 挖掘已设置为精确匹配的关键词：会引发内部出价竞争。
  • 跳过分组4：大多数卖家专注于减少浪费，却错过拓展收益的机会，而这往往是更大的增长杠杆。

Built by Jay GPT Pro

由Jay GPT Pro开发

Part of Amazon Pro Skills. Production-grade skills for serious Amazon sellers. Free and open. Built by Jay Margaliot.
I share a new AI play for Amazon sellers every week, free, in my WhatsApp group. Join here: https://chat.whatsapp.com/ILX65p1yWcaIG3c9WGHpTY
属于Amazon Pro Skills系列。为专业亚马逊卖家打造的实用工具。免费开源。由Jay Margaliot开发。
我每周会在WhatsApp群组中分享一个面向亚马逊卖家的AI玩法，免费加入：https://chat.whatsapp.com/ILX65p1yWcaIG3c9WGHpTY