mbfinotti/advertising-skills / skills/ad-negative-keywords
ad-negative-keywords
VerifiedBuild and maintain negative keyword lists from search term reports to cut wasted ad spend in paid search - n-gram query mining to surface irrelevant search terms, the right negative match types and exclusion levels (account, shared list, campaign, ad group), and a review cadence that avoids overblocking queries that convert. Use whenever the user mentions a search term or search query report, irrelevant clicks, junk traffic, wasted ad spend, or wants to exclude or block keywords - even if they never say 'negative keywords'. Covers B2B lead gen and B2C/e-commerce; needs the account's query data. Do NOT use for account-wide underperformance with no query-level evidence - use mbfinotti/advertising-skills@ad-account-diagnostic instead.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 16.2 KB |
| evals/evals.json | config | 20.0 KB |
| references/platform-notes.md | markdown | 4.4 KB |
| references/worked-example.md | markdown | 6.5 KB |
references/platform-notes.md
View on GitHub# Platform notes (optional)
Vendor-specific mechanics for the main workflow. Load this file only when the user names their platform; the SKILL.md workflow stands on its own without it. Limits and defaults drift - verify in-account before building at scale.
## Google Ads
- Negatives never match close variants: excluding broad `flowers` blocks "red flowers" but NOT "red flower". Add plurals, synonyms, and stems manually.
- Casing and misspellings are covered automatically.
- When adding a negative from the search terms report UI, the pre-selected match type is exact - the trap flagged in the skill's failure modes. Switch to phrase before saving.
- Levels and limits:
- Up to 10,000 negatives per Search or Performance Max campaign (PMax raised from 100 in March 2025).
- 20 shared lists per account, 5,000 keywords per list.
- Account-level list applies to all eligible Search and Shopping inventory including PMax.
- Display/Video considers at most 1,000 account-level negatives and treats them as thematic exclusions, not literal blocks.
- Performance Max and standard Shopping have no positive keywords - negatives (plus brand exclusions in PMax) are the primary query control. PMax negatives became self-serve in December 2024. Shared-list support arrived August 2025.
- Search term visibility is partial since September 2020. Practitioner measurements put hidden terms around 40% of spend (ranging 20-80%). This is what makes the n-gram pass necessary rather than optional.
- Formatting rules:
- Max 80 characters and 10 words per negative.
- A query longer than 16 words can bypass a negative whose match falls after the 16th word.
- Usable symbols are `&`, accents, and `*` - accented and unaccented count as different negatives.
- Most other punctuation errors out.
- Conflict tooling: the platform's "conflicting negative keywords" recommendation misses shared-list conflicts - pair it with a script or third-party conflict check rather than trusting it alone. Never auto-apply it.
- Data pull via API (GAQL): `keyword_view` segmented by date returns one row per keyword per day per match type - deduplicate on (ad group, keyword text, match type) and aggregate metrics, or drop the date segment, before computing thresholds.
## Microsoft Advertising
- Default negative match type is phrase (opposite of Google's exact default when adding from a report).
- No negative broad match at all - phrase and exact only.
- Negative keyword lists attach to campaigns only, never ad groups: one list per campaign, 20 lists per account, 5,000 keywords per list.
- Close variants are not filtered by negatives here either: plurals, synonyms, and misspellings of a negative still match - add variants manually, same as Google.
- A dedicated negative keyword conflict report exists - use it for the quarterly conflict sweep.
## Amazon Ads
- Only negative phrase and negative exact exist - no negative broad.
- Opposite of Google on variants: Amazon negatives DO block close variations of the phrase. Fewer manual variants needed.
- Negative product/ASIN targeting blocks specific product detail pages - a lever Google has no equivalent for. Use it against poorly converting placements and competitor pages.
- Three distinct reports:
- The advertising Search Term Report - actual shopper queries, the input for this skill's workflow.
- Search Query Performance in Brand Analytics - organic search behavior, brand-registered sellers only.
- Search Term Impression Share - your share per query, 30-day lookback.
- Common use of negative exact: stop auto and manual campaigns from cannibalizing each other on the same query.
- Cadence practitioners use: weekly above ~$10k/month ad spend, bi-weekly below.
- The 15-20-click zero-order gate (10-12 for CPCs above ~$3) originated in this ecosystem and transfers directly.
## Cross-platform reminders
- Never assume one platform's match-type defaults or variant behavior on another - the three above disagree on both.
- The skill's match-type ranking (`phrase > exact > broad`) survives a platform with no negative broad match: broad already sits last, so Microsoft and Amazon simply lose the bottom rung. The level ranking does move - where shared lists attach to campaigns only, the shared-list rung loses the ad-group reach the default order assumes, and a recurring theme inside one campaign drops to the campaign rung.
- Keep one master exclusion taxonomy per business, but maintain per-platform lists: syntax, limits, and levels don't port automatically.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Unverified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |