Original Summary

I've been sitting on this for a few weeks and I need to get it out, because it's the first genuinely big project of my career and honestly I'm still not fully over it. Hands were shaking for half of it. Bear with me. A mid-size US pet food DTC brand was bleeding somewhere between $240k and $400k a year to churn. Around 8.7% of customers walking every month. And here's the part that still gets me: the food was good. People weren't leaving because of the product. They were leaving because they could not physically take the amount of messaging anymore. Let me paint the "before," because it's genuinely one of the wildest setups I've ever walked into. The CEO believed, with her whole heart, that every customer had to be reached on every channel or the abandoned cart flows wouldn't fire. So they were on Discord, email, SMS, RCS, WhatsApp, and Instagram DMs. If an app had a DM box, they were pushing OTPs and promos into it. Picture ordering dog food and then getting your login code on Discord from the pet brand. People muted them first. Then they cancelled. And the engineering behind it was a slow motion car crash: The first "WhatsApp agent" ran with zero protection. The official business number got banned in week 2. The dev's entire fix was to keep buying new numbers. That was the strategy. Buy, burn, repeat. The CEO then spent a full weekend on a call with a vendor's CTO rebuilding it with Claude and Codex. Two smart people, the Albert and Einstein meme in real life, brilliant at starting and hitting the exact same wall every single time. Numbers kept dying. Instagram accounts got banned. Round and round. 129 WhatsApp numbers banned. One hundred and twenty nine. At some point Meta should have just mailed her a plaque. They applied for the green tick on IG, FB and WhatsApp. That verification and sending broadcast ended up costing more than the entire fix I eventually built. * They finally discovered proxies, months and a lot of money later, and went with commercial ones from ranges that were already flagged to hell. So the bans slowed from every 2 weeks to every 3 or 4. They were rotating numbers on a schedule with a Postgres cron pinging every 15 minutes just to see which numbers were still breathing. All of that effort, all that money and stress, moved churn from 8.7% to 8.4%. That was the reward. Three tenths of a percent. I saw her post about the whole disaster on LinkedIn and you could feel the exhaustion coming off the screen. I'd built smaller versions of this kind of thing for a few brands, nothing near this scale, but I knew exactly why it was failing. So I reached out. We messaged a bit, and one random Monday morning she asked me to hop on a "quick 15 minute call" with her, the CTO and the CFO. That call ran almost two hours. I had nothing prepared. Normally I show up with notes, a rough proposal, an objection sheet, the whole thing. She'd told me it was jus


  • 情报分类:工作与职业机会
  • 分类依据:内容涉及招聘、求职或职业发展
  • 信息来源:Reddit · SaaS
  • 发布时间:2026/9/22 03:05:24