Amazon Search Term Harvest Planner

Amazon Search Term Harvest Planner

Purpose

Turn Amazon Ads search term data into a safe, specific harvesting plan. Identify search terms that deserve exact-match isolation, product-target expansion, bid direction, destination routing, or watchlist treatment. Do not treat every converting query as harvest-ready, and do not add source negatives unless routing or waste evidence supports it. Use orders or conversions as the negative-safety signal; treat sales as revenue, not order count.

Optimize for clean traffic control, profitable growth, and learning. Preserve brand defense, own-ASIN defense, launch/rank-defense traffic, and strategically valuable discovery until the data proves a safer route.

This skill works in standalone mode from static exports and in Rocketcart MCP mode when a live Amazon Ads + product-intelligence layer is available. It may prepare approval-ready rows, but it must not create keywords, negatives, product targets, bids, budgets, or campaigns by default.

Connection Modes

Standalone Mode

Use this mode when the user provides pasted tables, CSVs, screenshots, or summaries.

• State that Rocketcart MCP was not used.
• Build the best possible harvest plan from static search-term, targeting, negative, product, and destination data.
• Lower confidence when live entity IDs, current negatives, current bids/states, product context, snapshots, or recent changes are unavailable.
• Do not present rows as executable without live preflight.

Rocketcart MCP Mode

Use this mode when Rocketcart MCP capabilities are available or the user asks for live Rocketcart review, preflight, execution, readback, or monitoring.

Initial Rocketcart review is read-only. Do not use write capabilities until the user explicitly approves exact rows after live preflight.

• If profile is missing, use profile discovery first. If exactly one profile fits, state the assumption; if multiple profiles plausibly match, ask the user to choose.
• Read current campaign, ad group, product-ad, keyword/target, negative, budget, state, bid, and destination context where available.
• Read search-term, targeting, recent-change, snapshot/changelog, and entity-history context where available before trusting stale exports.
• Read product intelligence where available: ASIN/SKU mapping, inventory or availability, Featured Offer / Buy Box, price, reviews/rating, category rank/BSR movement, estimated demand, competitor signals, margin/readiness, and mixed-ASIN risk.
• Use live reads to resolve exact IDs, detect duplicate exacts/targets, detect current negative conflicts, verify destination delivery feasibility, and detect stale rows.
• Produce approval packets for exact candidate rows; execution remains separate.
• After any explicitly approved execution, read back affected entities and define 3/7/14-day monitoring.
• Read [Rocketcart search-term harvest mode](references/rocketcart-search-term-harvest-mode.md) when you need the Rocketcart-specific read, preflight, approval, execution, and readback sequence.

Rocketcart Review Modes

Choose one primary mode when Rocketcart is involved:

| Mode | Use When | Output | |---|---|---| | Live Harvest Review | The user wants a current search-term harvest plan using live account state. | Live-resolved classifications and proposed rows, no writes. | | Preflight / Approval Readiness Review | Candidate harvest, negative, target, bid, or delivery-fix rows already exist. | Row-by-row readiness, stale-row blockers, exact approval text. | | Execute Approved Rows | The user explicitly approves exact rows for execution. | Execute only the approved subset after live preflight, then read back. | | Post-Change Readback / Monitoring Review | Changes were executed and need verification or outcome review. | Readback status and 3/7/14-day monitoring verdicts. |

Required Inputs

Gather, derive, or mark unavailable:

• Marketplace, profile/account, currency, timezone, ad type, and date windows.
• Search term report with campaign, ad group, search term, match type/source targeting where available, impressions, clicks, spend, orders, sales, ACoS/ROAS, CPC, CTR, and CVR.
• Targeting or keyword report with existing keywords/targets, match type, bids, states, and campaign/ad group destinations.
• Campaign and ad group structure: targeting type, naming, budgets, states, portfolio, advertised ASINs, and strategic role.
• Product context: advertised ASIN, purchased ASIN where available, parent/child relationship, category, price, margin or target ACoS/CPA, inventory, Featured Offer / Buy Box, reviews, rating, delivery promise, and listing readiness.
• Traffic segmentation: branded, own-brand generic, category generic, competitor brand, competitor ASIN, own-ASIN defense, auto close/loose/substitute/complement, launch/rank-defense, and exploratory discovery where data supports it.
• Existing exact keywords, negatives, product targets, and campaign routing rules when available.
• Recent orders or conversions by normalized search term, keyword, target, ASIN-like term, and phrase family for negative decisions. Default to L30 when available; state any different window.
• Destination delivery feasibility: destination campaign/ad group state, budget status, negative conflicts, advertised ASIN fit, and whether the destination can receive traffic.
• In Rocketcart MCP mode: exact profile, live entity IDs, current live values, recent drift, snapshots/changelogs, entity history, product-ad ASIN/SKU mapping, and product-intelligence freshness where available.

Missing Data Handling

• Missing margin or target ACoS/CPA: avoid firm profitability claims; use ACoS/ROAS/CPA only as proxies.
• Missing search term source mapping: do not recommend source negatives or precise routing.
• Missing destination campaign structure: propose a destination pattern, but mark the action Needs Destination.
• Missing existing exact keywords/targets: flag duplicate-risk before launching new exact terms.
• Missing current negative map: do not mark harvest or destination rows APPROVAL_READY.
• Missing exact profile, campaign, ad group, keyword, target, or destination IDs: keep write rows below APPROVAL_READY.
• Missing purchased-product data: avoid ASIN leakage conclusions.
• Missing recent order/conversion data or unmapped normalized traffic for a proposed negative: mark the negative row NEEDS_DATA, not APPROVAL_READY.
• Missing inventory or Featured Offer / Buy Box: do not recommend aggressive bid or budget scale.
• Missing comparison period: classify by current signal strength and confidence, not trend.
• Missing Rocketcart profile in live mode: list or request profile selection before live reads or execution.
• Missing live resolution or live preflight in Rocketcart mode: keep rows below APPROVAL_READY.
• Live state differs from a candidate row's current value: mark the row stale, block execution, and request refreshed approval.

Write-Readiness Statuses

Every action row must use exactly one status:

• PLANNING_ONLY: strategic recommendation, destination pattern, or analysis that is not a live write candidate.
• NEEDS_DATA: required report, exact ID, destination, product context, economics, duplicate check, negative check, or strategic role is missing.
• BLOCKED: unsafe or non-executable due to duplicate conflict, negative conflict, broad negative blast radius, brand/own-ASIN/launch/rank protection, retail-readiness blocker, budget-starved destination, mixed ad-type contamination, or failed gate.
• APPROVAL_REQUIRED: row is specific enough to review, but still needs explicit human approval and live preflight before any write.
• APPROVAL_READY: human can approve the exact row after verifying live preflight. This status is rare and requires all readiness fields below.

No row may be APPROVAL_READY unless it includes exact profile/account, marketplace, ad type, source campaign ID, source ad group ID, destination campaign ID, destination ad group ID, keyword/target or negative ID where applicable, normalized search term, match type, current state, current value, proposed value, proposed action, duplicate checks, current negative checks, destination feasibility, approval text, preflight checks, readback checks, and 3/7/14-day monitoring criteria.

In Rocketcart MCP mode, APPROVAL_READY means the row is ready for explicit human approval after live preflight. It does not mean "execute now." Execution requires a separate user approval message naming the exact row or row IDs.

Hard Blockers

Classify the row as BLOCKED or NEEDS_DATA, not APPROVAL_READY, when any of these apply:

• Existing exact keyword or product target already covers the normalized term, unless the action is delivery repair for an existing entity.
• Existing exact exists but is paused, archived, budget-starved, blocked by a negative, or in the wrong destination; classify as Scale Existing Exact / Delivery Fix, not duplicate harvest.
• Existing keyword/target report is missing; classify harvest rows as NEEDS_DATA.
• Current negative map is missing or destination has a negative that would block delivery.
• Destination campaign/ad group is missing, paused, budget-starved, wrong ASIN, wrong traffic type, or missing exact IDs.
• Search term is a brand-defense, own-ASIN-defense, launch/rank, or strategic discovery term and the proposed source negative would cut protected traffic.
• Competitor conquesting term has high ACoS but the source campaign's strategic role is conquesting or share defense; prefer Bid Down / Keep Learning, Watchlist, or controlled test.
• Proposed negative keyword, search-term negative, product-target negative, or ASIN/product-target negative has >= 2 recent orders or conversions for the same normalized term, target, ASIN-like value, or covered phrase family. Default to L30 when available, aggregate across relevant campaigns/ad groups in live mode, and do not use sales revenue as an order count.
• One order, one large order, or thin samples create attractive ACoS; use Controlled Test or Watchlist, not Harvest Ready.
• Negative phrase would block relevant query families, own-brand variants, size/color variants, profitable discovery, protected strategic traffic, or any covered query/variant/target/ASIN-like value with >= 2 recent orders or conversions. Use narrower negative exact or block the negative.
• Sponsored Products, Sponsored Brands, and Sponsored Display data are blended and cannot be separated for the action.
• Product readiness, inventory, Featured Offer / Buy Box, margin, or listing relevance could change the decision and is unavailable.
• Rocketcart live state shows a current value, entity state, negative, destination, product-readiness gate, or recent drift that differs from the candidate row.

Evidence Thresholds

Use user-provided thresholds when available. Otherwise use these defaults, adjusted for product price, category, lifecycle stage, and account volume:

• Harvest Ready: at least 2-3 orders, relevant intent, ACoS/CPA within target economics, acceptable CVR, no retail-readiness blocker, duplicate checks passed, current negative checks passed, destination is feasible, and a clear exact destination exists.
• Controlled Test: 1-2 orders or promising CVR/CTR but not enough volume for a confident harvest; use lower bid, limited budget, routing review, or watchlist. Do not default a one-order term to a negative.
• Scale Existing Exact: the term already exists as exact and has enough profitable volume; recommend delivery, bid, budget, state, negative-conflict, or placement review instead of duplicate harvesting.
• Product Target Candidate: search term or purchased ASIN indicates an ASIN target should be tested, with relevance and economics checked.
• Bid Down / Keep Learning: relevant term with orders but above target economics; reduce bid only when current evidence and strategic role support it.
• Negative Candidate: irrelevant or structurally mismatched term, or spend exceeds 1.5-2.0x target CPA with zero orders and no strategic defense, launch, ranking, discovery, or recent-order guard reason.
• Watchlist: below threshold, missing data, unclear relevance, or strategic role not yet resolved.

Do not overfit tiny samples. If evidence is thin, use Watchlist, Controlled Test, or Needs Data.

Lifecycle adjustments:

• Launch/rank window: tolerate higher ACoS only when the rank objective, budget cap, stop-loss, inventory, and monitoring rules are explicit.
• Brand defense: prioritize coverage and defensive routing over short-window ACoS, but do not add source negatives without approved defense coverage.
• Mature profit campaign: require economics, duplicate checks, and stable CVR before harvesting or bid increases.
• Exploration/discovery: prefer controlled tests and watchlists over negatives until waste is isolated.

Traffic-Type Routing Matrix

| Traffic Type | Preferred Destination | Source Negative Default | Key Gate | |---|---|---|---| | Brand defense | Brand exact or defense ad group | Blocked unless defense coverage is approved and verified | Preserve branded and own-brand demand. | | Category generic | Nonbrand exact by ASIN/product group | Negative exact only after destination delivery is verified | Check relevance, margin, inventory, and duplicate exacts. | | Competitor brand | Competitor/conquest exact campaign | Usually blocked; bid down or controlled test first | Confirm strategic role and expected ACoS tolerance. | | Own-ASIN defense | Own-ASIN product-target or defense campaign | Blocked unless own-ASIN defense is intentionally rerouted | Requires ASIN map and purchased-product context. | | Substitute/complement ASIN | Product-target test | Blocked until ASIN relationship is known | Avoid own-ASIN or irrelevant target mistakes. | | Launch/rank | Rank-support exact or launch campaign | Blocked unless launch/rank coverage is preserved | Requires objective, stop-loss, inventory, and monitoring. | | Discovery/exploratory | Controlled exact test or watchlist | Blocked unless waste is isolated | Do not cut learning from thin samples. | | Irrelevant query family | No destination, negative review | Negative exact or narrow phrase only | Phrase negatives need blast-radius proof. |

Harvest Safety Gates

Before recommending an exact harvest, check:

• Normalize the search term for case, punctuation, plural/singular variants, spacing, ASIN casing, and simple token order differences before duplicate checks.
• The search term is not already isolated as an exact keyword or product target unless the action is to fix routing, state, delivery, budget, or negative conflicts.
• The destination campaign/ad group is known or the output clearly proposes one.
• The destination matches traffic type and business goal: branded, category generic, competitor, own-ASIN defense, rank-growth, profit, discovery, or product targeting.
• Destination delivery is feasible: enabled state, budget headroom or budget plan, no current negative conflict, correct advertised ASIN/product group, and exact destination IDs when claiming approval readiness.
• The advertised ASIN is retail-ready enough for the traffic.
• The term is relevant to the destination ASIN or product group.
• The proposed match type, bid direction, budget stance, and negative strategy are explicit.

Before recommending a source negative, check:

• The term has been safely captured elsewhere or the source traffic is clearly wasteful.
• Recent order/conversion data is available for the normalized term, target, ASIN-like value, and covered phrase family. Default to L30 when available; state any different window.
• Orders/conversions are aggregated across relevant campaigns/ad groups when live data is available, not judged only from the source row.
• The proposed negative is blocked when the same normalized traffic has >= 2 recent orders or conversions. If it has exactly 1 order/conversion, default to

…

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: nospicyplease
• Source: nospicyplease/amazon-ppc-advanced-skills
• License: MIT
• Homepage: https://rocketcart.io

Install and usage instructions live in the source repository linked above.