In 2005, Salesforce was losing 8% of its customers every month.

Compound that for a year and two-thirds of the customer base has left you. And yet the company was celebrating record new sales! They were pouring water into a bucket with a hole in it, and everyone was just watching the faucet.

I'd love to say I read that story and shook my head. The truth is I've been in those rooms and spent years staring at new customer numbers. I couldn't have told you our churn rate precisely.

It's an asymmetry so ordinary we look straight past it. When you win a customer, something happens - a notification fires, someone posts it in Slack.

When you lose one, almost nothing happens. She decides, mid-scroll, that she's done with your brand. She doesn't complain. She doesn't unsubscribe. She just never comes back.

One event gets applause. The other doesn't even get noticed.

If you're small enough, you still feel it - you notice when a customer goes quiet. But that instinct doesn't scale. Somewhere between 500 and 5,000 customers, Sarah stops being Sarah and becomes row 11,304 in a spreadsheet that quietly stopped updating.

And she's still there. She gave you money once, and she's been sitting in your database for eight months while your ad account pays premium prices to reach total strangers - most brands now lose money on a first order.

She's not alone, either. On a typical 20,000-customer list, roughly 15,000 people bought once and went quiet. Fifteen thousand Sarahs. Win back just 3% of them at a $70 order and you've found $33,000 - from an email automation that takes an afternoon to build.

One warning, because your email platform will lie to you about this. It counts every returning customer as a win, including the ones who were coming back anyway. So keep 10% of the list out of the campaign and watch them. If they buy at the same rate as everyone else, the sequence isn't working - your brand is.

Pull the customers who bought once, 60+ days ago, and email them this week. It's the cheapest revenue you haven't collected.