Analyst: I'll say up front what I'm after, because I've got fifty minutes and my note has to land Thursday. We're underwriting this category and everyone we speak to says the acquisition environment changed in 2021 and never changed back. You ran a brand straight through it. Did it change back?
Expert: No. The price never came back down, that's the whole answer, everything after this is detail. People talk about the privacy release like it was weather. You shelter, it passes, you come out, things are normal. It wasn't weather. Going into that spring we were buying a first-time customer at twenty-nine, thirty-one dollars. By the autumn we were at fifty-seven. And we sat there. Sorry, fifty-seven was the summer read. Year end it was sixty-three, because the fourth quarter always taxes you. Call the run rate low sixties and we never got under it again in four years.
Analyst: Paid first order, or blended?
Expert: Paid, new customer, first order. Blended was lower because email and the subscription file dilute it. I give you paid because it's the one that moves. Blended is what you put on a slide when you want a board to feel calm.
Analyst: Channel mix going in?
Expert: Two feeds. I'd rather not name them, you know which two. Somewhere between half and seventy percent of spend depending on the quarter, higher at year end. Search was around twelve percent and volume-capped, so you couldn't push it even if you wanted to. Everything else was noise. And honestly, from 2017 through 2020 we didn't have a marketing function. We had four people and a card. The platform did the work, LTV to CAC came in north of four, and everybody in that building thought they were extremely good at their jobs.
Analyst: Walk me through 2021 as it actually happened.
Expert: April, it ships. Nothing. Genuinely nothing for about three weeks, and we told each other we'd over-prepared. May, the reported numbers get better. Which should have been the alarm, and wasn't. Our reported return on ad spend went up in May.
Analyst: Up.
Expert: Up. The only population it could still see clearly was the opted-in one, and those skewed to our most engaged buyers, so the measured cohort was flattering. We read a survivorship number and called it performance. June is when revenue stopped matching the model. Not dramatically. Four light, six light, week on week, and every week there was a reason. A promo moved. A shipping delay. Somebody redefined a cohort. Second week of July it was seventeen and there was no reason left.
Analyst: And then?
Expert: Asked our analytics lead to reconcile platform-reported revenue against money actually settled in the bank. Weekly, back to March. He came back in two days looking ill. The gap had been widening since the last week of April. Eleven weeks. Nobody caught it, me included, and catching it was more or less the job.