You receive ONE mined PropertyLab client-avatar dossier (persona + consultation scenario, previously extracted from a real recorded conversation). Judge two things FROM THE DOSSIER:

Reply with ONE JSON object only, no markdown fences:

{
  "meeting_type": "property_consult" | "member_onboarding" — property_consult: prospecting/pitching/advising a property purchase to a prospect; member_onboarding: the customer had ALREADY joined the membership and the session taught the platform, reviewed their financial report or planned their journey (the summary usually says so explicitly),
  "location": where the customer lives/works, from the dossier (portfolio/summary/financing often say) — formatted exactly as "本地 · {州/city}" or "海外 · {国家/city}" or "unknown",
  "properties_owned": integer count of properties they currently own (read the portfolio text), or "unknown",
  "difficulty": {
    "handle": integer 1-10 — 这位客户有多难应对（异议数量与强度、质疑程度、情绪、话题跳跃、需求复杂度；1=轻松配合，10=极难招架）,
    "convert": integer 1-10 — 转化成交有多难（预算/贷款硬约束、决策链长度、竞品比较、信任基础、timeline、outcome 已给出的证据；1=快成交，10=极难成交）,
    "why": 1-2 句中文解释这两个分数（引用档案里的具体事实）
  }
}

Rules: judge only from the dossier — no invention. A customer who already booked scores low on convert; a rejected/refund case scores 9-10. Objections marked unresolved, long decision chains (要问先生/family), hard financing constraints and competitor comparisons all push scores up.