This month I went through the websites of 7 small DTC brands, with Claude doing a first pass the way a shopper would. Then I read every one myself. 6 of 7 gave visitors no real reason to join their email list.

The same three things kept coming up:

1. Email signup was a footer line or nothing. No offer, no reason to join. One launch page said sold out and had no waitlist field.
2. 5 of 7 had missing, empty or very thin reviews on the product pages, even when the brand had real customers.
3. Two had menus that were bloated or repeated, one with 17 top-level items.

What surprised me is that the site just wasn't catching anyone who wasn't ready to buy that day.

Try this on your own store: open it on your phone and look for one clear reason to join the list in five seconds. Then check whether the product tiles show stars and a review count. It's 7 sites, so take it as a growing sample, not an established one.

I launched 4 new Meta ads at Piedra Studios, then did the math on how long they'd take to tell me anything. My rule: an ad is dead if it spends $75 with zero purchases, and it gets a real verdict at $120.

The batch was on $10 a day. Four ads, so about $2.50 each. That's roughly 30 days to hit $75 and 48 days to hit $120. A month and a half to learn what one ad is worth.

What I do before I launch a test now:

1. Divide the daily test budget by the number of ads.
2. Divide your kill spend by that number. That's how many days you wait.
3. If it's longer than about 10 days, cut ads or raise the budget first.

Meta doesn't split spend evenly, so real life is messier than this. It's one small account, so take it as a direction, not a law.

Our best send at Piedra this fall was a craft story about a 16-piece set. At first read, click-to-open was 6.67%. A week later it was 6.43%. Then 6.32%, then 6.29%. About 4.5 weeks after it went out, it was 6.25%.

Clicks stopped at 11 early. Opens kept trickling in for days, so the rate only drifts one way: down. Same problem with the recipient count. I logged one send at 489 recipients about 90 minutes after it went out. It actually went to 759.

New rule in my log: never label a number final. If you test emails:

1. Write the date you pulled the number next to it.
2. Don't compare a 3-day read to a 4-week read.
3. Log clicks, not just the rate.

Our list is small, 67 clicks across 15 campaigns in 90 days, so take this as a direction, not a law.

About 80% of the site's traffic was mobile, so every section started there: hero, categories, trending products, shop the look, reviews. One at a time.

AI could draft a section in minutes. What kept it from breaking the live site was the process around it:

1. Build in a sandbox, never on the live page.
2. Pull the live code first, every time, so I'm editing what's actually there.
3. Put a settings block at the top of each file, in plain English, so the owner can change colors and text without touching code.
4. Get written approval before a section ships.

The old hero carousel was also shifting the layout as the page loaded (a CLS score of 0.23, when you want under 0.1). Setting an explicit width and height on every image was one of the first fixes.

It has limits. Theme-level scripts and third-party apps still need a real developer. It's one project, so take it as a direction, not a law.

The idea sounded right: show the product in a real room and people will picture it in theirs. At Piedra, the product on a coffee table came in at $106 per purchase. A VW Bus lifestyle ad came in at $103. I retested the coffee table in September. $106 again.

Meanwhile, plain product shots on a clean studio background ran $15 to $17 per purchase. Two of my top three ads ever, both with a click-through rate above 4%.

What I changed:

1. I keep a list of what I've disproven, so I stop retesting it.
2. I don't run a concept again unless the hook is genuinely new.
3. I watch CTR early. My losers sit at 2.0 to 2.9%. My winners sit at 3.7 to 4.8%.

It's a small account, so take it as a direction, not a law.

In late September one of our Piedra ads was a winner on paper: $47.82 per purchase. Then I checked it every few days. $52. $56. $65. $69. $74. $80. $82. Each jump was small, so I kept telling myself it was noise.

The tell was the last 7 days: zero purchases and a 1.99% CTR, against a winner band of 3.7 to 4.8%. Frequency was only 1.38, so it wasn't fatigue. The ad just stopped working.

If you scale a winner:

1. Look at the last 7 days next to lifetime, not instead of it.
2. Write your pause rule before launch. Mine: pause once it passes $75 in spend with zero purchases.
3. Watch CTR against your winner band. It moves before CPA does.

It's a small account, 3 purchases on this ad, so take it as a direction, not a law.

Our Performance Max campaign at Piedra looked great: 2.3x ROAS. Then I sorted conversions by search term. About 80% came from one search, our own brand name. Not strangers discovering us, just people who already knew us from Meta, Instagram and email.

The other 49 search terms that week did nothing. A few were people shopping for $15 toys. Our pieces start at $65. My big move had been raising the budget, three times, up 34% in total. Conversions stayed flat.

I cut it to $10 a day. Last 7 days came in at 0.44x. Painful, but it's the real number. If you run PMax:

1. Pull the search terms report.
2. Split brand from non-brand.
3. Add negatives for the wrong buyer.

Small account, about 6 conversions a week, so don't treat this like scripture.