宠物食品/包装 · 企业分层投放策略 Skill
A reusable playbook + tooling for turning a raw list of pet-food / pet-packaging companies into a prioritized, tagged, channel-ready targeting plan. Built from a real 41,844-company engagement (→ 13,805-company integrated table, 92 web-enriched, full contact backfill).
When to use
- User supplies a company list (xlsx/csv) and wants a 分层投放策略表 with S/A/B/C tiers.
- User asks to redesign a scoring system with explicit priority/weights (e.g. "场景分权重最高").
- User wants a 客户信息源搜索与补全方案 (dimensions, sources, missing-field completion priority).
- User wants per-tier 投放策略/渠道/预算 or a 客户生命周期阶段 playbook.
Core design principle (scene-dominant scoring)
场景分 is the dominant, highest-weight signal. Priority order and weights:
场景分 50 > 规模分 30 > 注册资本分 20 — total 100, no keyword-bonus component by default.
This guarantees: a non-强匹配 company caps at 80 (A), and S (≥85) requires a 强匹配 scene. Verify dominance before shipping any revision.
Workflow (execute in order)
- Screen raw list → candidate pool via two gates:
- Industry relevance: name contains "宠物" OR 行业中类 ∈ {饲料加工,宠物服务,农林牧渔产品批发/零售,食品饮料批发/零售}.
- Scene match (drives 场景分): 强匹配=name has 宠物食品/宠粮/宠物饲料/宠物主食/宠物零食/宠物营养 OR (宠物 + 饲料加工/食品饮料/农林牧渔) → 50; 中匹配=宠物 but not strong → 30; 弱匹配=industry-only, no pet in name → 10. - Enrich missing fields by priority P0→P3 (see
references/scoring_framework.md§补全优先级). Store web results in_websupp.jsonkeyed by exact company name; never fabricate — mark unverifiable as "待核". - Backfill contacts (法定代表人/手机号/邮箱) from the source list into both重点表 and 候选池 sheets.
- Score & tier with
scripts/scoring.py(scene-dominant 100-pt). Output integrated single table; de-dup 重名 keeping the权威 重点表 version. - Tag with business tiers + multi-dim labels: S=高价值战略直客, A=潜力重点拓展, B=长尾主力批量, C=培育待激活; plus 场景/规模/资本/资质/业务/生命周期 labels.
- Strategy per tier (priority S>A>B>C; budget建议 S40-50% / A25-30% / B15-20% / C5-10%) and per lifecycle stage (认知→兴趣→评估→成交→复购→流失预警) — see
references/scoring_framework.md§策略.
Reusable assets
scripts/scoring.py— deterministic scene-dominant scoring + tiering + integrated-table builder. Read its--helpfor args; patch paths per environment.references/scoring_framework.md— full rule tables (screening, scoring, grading thresholds, enrichment priority, tier→strategy map, lifecycle playbook, compliance notes).
> 版权声明：本技能版权归个人所有（© Astra · yuxingxing），非公司共同财产；授权公司及团队在 WorkBuddy 内部使用。详见 LICENSE。
Output conventions
- Deliver an integrated single-sheet xlsx (企业分层投放总表) with a 来源 column (重点表/候选池) and 26 columns covering all dimensions; freeze header; sort by 等级 then 总分 desc.
- Always state that budget/strategy final decisions belong to the user (Astra); provide recommended starting allocations only.
- Compliance: public-source data only; 注册资本/资质 以国家企业信用信息公示系统年报为准; no 极限词; B-end contact use within 个保法/GDPR bounds.
