过去几天我一直在 X 上进行冷启动。以下是我的发现:

1作者: timczm10 个月前
类似 HN 的数据,我这里有一些: 扫描账号数:710 已发送私信:126 (17.7%) 无法发送私信:584 (82.3%) 在这些私信中: 积极回复:4 (3.2%) 拒绝:1 (0.8%) 未回复:121 (96%) 流程如下: 首先,我确定了我的产品(Snapdemo.io)的目标客户是谁。然后,我找到了该领域的一些关键意见领袖(KOL),并使用 API 获取了他们所有的粉丝。这为我提供了 37000 个潜在客户。 接下来,我使用 AI 扫描个人简介,并猜测某人是否可能是我的理想客户。对于每个账号,我让 AI 告诉我: - 他们是理想客户吗?(是,否,不确定) - 为什么? - 他们的标签是什么?(例如,AI 开发者,SaaS 高管) 我通过 OpenAI 的批量 API 运行了所有这些,以节省成本和开发时间。 一旦我有了几千个理想客户,我就跳过了“不确定”的那些,因为已确认的客户已经足够了。 我制作了一个简单的仪表盘来跟踪所有内容,并使用一些模板自动生成私信。我还记录了哪个模板发给了哪个用户,这样我以后就可以衡量效果了。 如果没有一个可靠的自动化工具,而且大多数用户都不愿意接受私信,整个过程花费了大量时间。如果你做过类似的事情,我很乐意听取你的建议。 如果你也是一位创始人,欢迎私信我(X: @timchengb)交流想法。
查看原文
HN like data, here&#x27;s what I&#x27;ve got:<p>Accounts scanned: 710 DMs sent: 126 (17.7%) Can’t DM: 584 (82.3%) From those DMs:<p>Positive replies: 4 (3.2%) Rejected: 1 (0.8%) No response: 121 (96%)<p>Here’s the process:<p>. I started by figuring out who my product(which is Snapdemo.io) is for. Then I found some KOLs in that niche and used an API to grab all their followers. That gave me 37k potential leads.<p>. Next, I used AI to scan bios and guess whether someone might be my Ideal Customer. For each account, I asked AI to tell me:<p>- Are they ICP? (Yes, No, Not sure) - Why? - What’s their label? (e.g. AI builder, SaaS exec)<p>. I ran all this through OpenAI’s batch API to save on cost and dev time. Once I had a few thousand ICPs, I skipped the “Not sure” ones since the confirmed ones were enough.<p>. I made a simple dashboard to track everything, and used a few templates to auto-generate DMs. I also logged which template went to which user, so I could measure performance later.<p>Without a solid automation tool, and with most users not open to DMs, the whole process took a lot of time.If you’ve done something similar, I’d love to hear your suggestions. And if you’re also a founder, feel free to DM me (X: @timchengb) to exchange ideas.