we just wrapped up a killer 9 days: $255k in sales on $27.7k ad spend - 9.21 blended ROAS. that's Meta and Google combined. the last week and a half was a big step up, and it came down to a few key decisions.

The biggest change - cutting a money-losing collection

We had this one low-ticket collection that was bleeding cash. After looking at the last 4 months of P&L, we'd lost about $200k on it between ad spend, content creator fees, free product - the whole circus. So we pulled all advertising from that collection. now it still gets sales organically, and those sales are actually profitable because we're not dumping money into it. Meanwhile, our most profitable collection did $304k in sales on $43.6k ad spend (6.98 ROAS). by cutting the dead weight, we freed up resources to focus on the three collections that actually work. That alone doubled our team's efficiency.

One unique ad per concept, not three variations

We used to launch three unique ads per test batch. but after analyzing over 2,000 ads from the last 3 months, we noticed something: 79% of the time, the first original ad in the batch was the winner. so now we test one unique ad per concept, with one backup that we don't launch until the original starts fading. That shift gives us 33% more time to create genuinely new concepts. when you're pushing out hundreds of ads a month, that extra capacity means more meaningful tests. variations are only useful when a concept is declining - not from day one.

Videos over images - 90/10 split

For my brand (high-ticket products), video ads drive 90% of performance. Static images are just 10%. That makes sense because high-ticket purchases aren't impulse buys - people need more info. for lower AOV clients we work with, static still works at a 60/40 split. i don't see us going 100% video, probably stay around 80/20. the split fluctuates, but video is the workhorse here.

Content creation - the compound effect is real

We post 20+ pieces of content daily across stories, posts, and reels. competitors who've been around 7 years have a fraction of our social presence - we're only 2.5 years old but 5x their size. our sales team sends a daily report card, and customers regularly mention they follow our organic content and love the customer stories we share. social builds trust, and trust shortens the purchase decision window. it's a deciding factor when people compare us to competitors.

Custom events per collection - underrated

If you sell products at vastly different price points or solving different problems, you need custom purchase events. we set up separate events per collection so the pixel learns which customers buy which type. Most orders come from lower AOV, but most profit comes from higher AOV. Custom events let Meta target the right ICP for each collection instead of chasing the easiest conversion (usually the cheapest). a huge chunk of ad spend gets wasted showing ads to people who were never going to buy. Custom events cut that waste.

Super specific ads - the real ROAS driver

Ads have to be hyper-targeted. If your ad is generic and broad, no campaign structure or targeting setting can save you. meta's algorithm increasingly relies on the ad itself as the targeting signal. we obsess over message quality. every person on the marketing team listens to 20-30 sales calls a week - some 5 minutes, some 20. Those calls teach us exactly what customers want. listening to sales calls has been the single biggest unlock for our team this year. Don't sleep on getting to know your customers. call them, email them. Your ad performance will improve when your ads actually talk to them.

That's the update. cutting the losing collection was the hardest decision but the most impactful. everything else just amplified the gains