← 返回 Skills 市场
ivangdavila

SEM

作者 Iván · GitHub ↗ · v1.0.0
cross-platform ✓ 安全检测通过
1533
总下载
4
收藏
4
当前安装
1
版本数
在 OpenClaw 中安装
/install sem
功能描述
Search engine marketing strategy, keyword bidding, quality score optimization, and paid search campaign management
使用说明 (SKILL.md)

Keyword Strategy

  • Exact match isn't exact anymore — Google includes "close variants" (plurals, misspellings, implied intent). Review search terms report weekly
  • Broad match needs smart bidding — without automated bidding, broad match burns budget on irrelevant queries
  • Long-tail keywords have lower CPC but need volume — 100 low-volume keywords won't spend, consolidate into phrase/broad with negatives
  • Competitor brand bidding is legal but expensive — CPCs are 2-5x higher and Quality Scores stay low. Budget accordingly
  • Single Keyword Ad Groups (SKAGs) are outdated — Google's algorithm works better with theme-based ad groups of 5-15 related keywords

Quality Score Impact

  • Quality Score affects CPC exponentially — QS 10 pays ~50% less than QS 5 for same position
  • Three components: expected CTR, ad relevance, landing page experience. Check diagnostics to know which to fix
  • Historical CTR matters most — new keywords inherit account-level history. Poor performers drag down new campaigns
  • Landing page speed is a ranking factor — compress images, remove unnecessary scripts before scaling spend
  • Ad relevance requires keyword in headline — dynamic keyword insertion ({KeyWord:Default}) is cheap but effective

Bidding Mistakes

  • Manual CPC leaves money on table at scale — switch to automated bidding after 30+ conversions/month
  • Target CPA too aggressive kills volume — start 20% above actual CPA, tighten gradually
  • Maximize Conversions without cap spends entire budget regardless of efficiency — always set a target CPA or ROAS
  • Enhanced CPC (ECPC) is neither manual nor smart — commit to one strategy, ECPC is the worst of both
  • Bid adjustments stack multiplicatively — +50% device × +30% location = +95%, not +80%

Campaign Structure

  • Search and Display in same campaign is always wrong — different intents, different metrics, split them
  • Brand campaigns should be separate — protect branded terms, track true branded vs non-branded performance
  • Geographic targeting at campaign level, not ad group — easier budget control by region
  • Audience layering (observation mode) first — collect data before switching to targeting mode that restricts reach
  • Shared budgets hide which campaigns would spend more — use individual budgets to understand true demand

Ad Copy Rules

  • Include primary keyword in Headline 1 — ad relevance and CTR both improve
  • Responsive Search Ads need 8-10 headline variations — algorithm can't optimize with only 3-4 options
  • Pin critical messaging to position 1 — brand name, key offer, or compliance text that must always show
  • Call extensions increase CTR 5-10% — add phone number even if calls aren't primary goal
  • Test exactly one element at a time — headline vs headline, not headline + description + CTA simultaneously

Search Terms Hygiene

  • Review search terms weekly, not monthly — irrelevant spend compounds fast
  • Add negatives at campaign level for broad exclusions, ad group level for specific refinement
  • Negative keyword match types matter — "free" as phrase negative blocks "free trial" but not "trial free download"
  • Create a shared negative keyword list — apply learnings across campaigns automatically
  • Search term "other" category is hidden queries — if it's significant %, request detailed report from rep or use scripts

Conversion Tracking Errors

  • Last-click attribution hides assist value — at minimum, compare with data-driven attribution
  • Counting "All conversions" inflates ROAS — filter to primary conversion action for true performance
  • Phone call conversions need duration minimums — 60 seconds minimum filters spam and misdials
  • Micro-conversions (email signups) have different values than purchases — set values proportionally
  • Cross-device conversions are modeled, not tracked — factor uncertainty into ROAS calculations

Scaling Paid Search

  • Search volume has a ceiling — unlike social, you can't pay to reach more people than are searching
  • Impression Share shows headroom — if IS is 90%+, growth requires new keywords, not more budget
  • Diminishing returns are real — CPAs increase 20-40% when scaling past optimal volume. Know your efficiency/volume tradeoff
  • Bing Ads imports Google campaigns directly — 10-15% extra volume at lower CPCs, minimal extra work
  • International expansion multiplies complexity — separate accounts per market, don't just add countries to existing campaigns

Platform Automation

  • Auto-applied recommendations are opt-out, not opt-in — disable recommendations that conflict with strategy
  • Smart campaigns hide too much data — use regular campaigns for control and learning
  • Performance Max cannibalizes branded search — exclude brand terms or track incrementality carefully
  • Scripts automate repetitive tasks — bid adjustments, budget pacing, negative keyword mining run automatically
  • Rules-based automation is fragile — "if CTR \x3C 2%, pause" kills potentially good ads too early. Use with caution
安全使用建议
This skill is essentially a reference guide for paid-search best practices and is internally consistent. It's low-risk as provided because it contains only text and asks for no credentials or installs. Before using it to perform real changes, remember: (1) never paste account API keys, tokens, or passwords into a skill unless you trust the author and the connector's security; (2) if you pair this guidance with automation or an account-connector, grant the minimum scope (read-only where possible) and test on a small budget/account; (3) ask the publisher for provenance (homepage/owner identity) if you need stronger assurance — this package lists an anonymous owner ID and no homepage; and (4) validate critical recommendations against your account representative or official docs before applying large-budget changes.
功能分析
Type: OpenClaw Skill Name: sem Version: 1.0.0 The skill bundle contains a standard `_meta.json` file and a `SKILL.md` file. The `SKILL.md` file provides detailed, purely informational guidance on Search Engine Marketing (SEM) strategies. There is no executable code, no instructions for the AI agent to perform any actions beyond processing the informational content, and no evidence of prompt injection attempts, data exfiltration, malicious execution, or persistence mechanisms. All content is aligned with the stated purpose of providing SEM expertise.
能力评估
Purpose & Capability
The SKILL.md is marketing guidance about keyword strategy, bidding, quality score, and automation — exactly what the skill name/description promise. There are no extraneous requirements (no env vars, binaries, or install steps) that would be unrelated to SEM work.
Instruction Scope
The instructions are prescriptive advice and best practices only. The doc briefly mentions using 'scripts' and 'automation' for repetitive tasks, but it does not include any execution commands, file reads, or instructions to contact external endpoints or access account credentials. Because it references automation conceptually, be cautious if you later pair this skill with connectors that can execute changes.
Install Mechanism
No install spec or code files are present (instruction-only). That minimizes disk-write and code-execution risk.
Credentials
The skill requires no environment variables, credentials, or config paths. It discusses Google Ads/Bing conceptually but does not request API keys or tokens.
Persistence & Privilege
No elevated persistence requested (always:false). The skill does not request to modify other skills or system settings.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install sem
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /sem 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release
元数据
Slug sem
版本 1.0.0
许可证
累计安装 4
当前安装数 4
历史版本数 1
常见问题

SEM 是什么?

Search engine marketing strategy, keyword bidding, quality score optimization, and paid search campaign management. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 1533 次。

如何安装 SEM?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install sem」即可一键安装,无需额外配置。

SEM 是免费的吗?

是的,SEM 完全免费(开源免费),可自由下载、安装和使用。

SEM 支持哪些平台?

SEM 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 SEM?

由 Iván(@ivangdavila)开发并维护,当前版本 v1.0.0。

💬 留言讨论