reddit · SideProject · 10-16 16:50 63 分 ← 返回列表

I pointed an AI agent at the Western Cape for-sale market and it built a 52,894-listing price database

Side project that got out of hand: I set an AI agent on the Western Cape for-sale market and it scraped, merged and normalised every listing it could reach into one table. The result: 52,894 properties, 33 columns, one clean sheet (plus SQLite and an 11-sheet Excel workbook). It merges both major portals - 26,484 rows from Property24 and 26,410 from Private Property - and de-duplicates the same house appearing on both. What the data actually says: - median asking price R2,850,000 (range R

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：房产中介做 CMA 比价靠直觉，手工整理可比房源要 3-4 小时，缺少快速、干净、可查询的区域成交/挂牌数据库。

目标用户：南非西开普省房产中介、估价师、房产投资者和小型地产数据团队。

证据：estate agents price on gut feel, and a comparative market analysis takes 3 to 4 hours by hand. This builds the comp set in minutes. Whoever wants the property data, the way in is email dominus@quantanova.net .

变现：按区域/城市订阅数据库访问；按次出售 Excel/SQLite 数据包；给中介提供 CMA 报告生成 SaaS；线索授权给贷款、保险或搬迁服务。

MVP：选 1-2 个城市，先做可搜索的挂牌价与去重可比房源看板，支持按 suburb、房型、卧室数、面积筛选并一键导出 CMA 摘要报告；用邮件/Stripe 收费，人工更新数据。

竞品：Property24、Private Property、Lightstone、PropData、WinDeed

摘要：作者用 AI agent 抓取并合并西开普省两大房产门户，生成 52,894 条挂牌、33 列的干净数据库，其中 Property24 26,484 条、Private Property 26,410 条。痛点在于中介手工做 CMA 要 3-4 小时，这个数据库可分钟级生成可比房源。作者留了邮箱变现入口，但没有公开定价或收入数据。