@maxwellfinn $100 to $101 resets nothing, that's the part people miss. Raise in 20% steps and log CPA at each one. The step where CPA breaks is that ad's real ceiling, so you know when to stop pushing it.
@oliverkenyon The tan sandal nobody knew existed. Once swatches are on the card, make tapping one swap the card image too, then compare clicks by colour. The colour that jumps is demand you've been hiding.
@herrmanndigital Meta-touched share down to 15% is the number to watch. Pull 30-day repeat rate for AppLovin first-click buyers vs Meta's. If it holds, the 40% cheaper CPM is real, not just cheaper buyers.
@CadeMallett 30 creators on day one, six weeks, nothing learned. Brief the first 5 on one angle with different openers. The winning creator tells you who to cast; the other 25 then test angles, not faces.
@samdtcreative $500k+ on one "competitor as villain" script. Replay it as text-only first since it's cheapest to make. If CTR holds without a face or voice, the script is the asset and every new format is just distribution.
@mitkazakov 1 / margin after shipping and fees, yes. Do it with the discount baked in: 60% margin at 20% off drops to 50%, so break-even goes 1.67 to 2.0. Most set their target before the discount.
@AmyRuthUGC First 3 seconds beating a full reshoot. Cut 3 hooks onto the same body, equal spend, then compare 3-sec view rate AND hold to 50%. A hook that wins the first and loses the second is just clickbait.
@nolantoken 30 of their top 56 oregano ads on one persona is the real tell. Pull a brand's top ads, tag each by persona, then count. Whichever persona owns over half is where their margin actually lives.
@lukasenECOM Day 194, $109 in, $0 back. Before the new product, pull the dead one's CTR and add-to-carts. Clicks but no carts = a price/page problem that follows you to the next product. Low CTR = it was the hook, not the product.
@leov@triplewhale Your story is the real lesson: a 2025 trial still had access in 2026. And uninstalling from Shopify isn't the whole job. The Meta and Google ad account connections an app made get revoked separately, in each platform's business settings.
@jaketheadnerd@triplewhale@ecommcowboy Thank you for splitting confirmed from rumor, rare this week. One add: rotate tokens even if you got no notice. "No action required" covers your Triple Whale account, not the Shopify and Meta keys it held.
Triple Whale confirmed a breach: unauthorized access to internal systems, Oct 31 to Nov 1.
Still not public: what data was taken.
If it's connected to your store, do this before BFCM:
β rotate Shopify + Meta API tokens
β remove apps you don't use
β warn customers about phishing
Hundreds of AI ads won't save Black Friday if none get spend.
Example: $300/day, $30 target CPA.
3 ads β $100 each, room for 3 sales
30 ads β $10 each, can't buy one
β cap ads per ad set at budget Γ· target CPA
β test concepts, not cuts
Spend per ad is the test.
@Lioranpi "Returns land in Q1" is the sleeper. Tag every BFCM order by discount code, then compare return rate by code in January. If the deepest code returns the most, it bought returns, not customers.
@Divance_Ecom For gummies, the buyer's biggest doubt is whether they actually work. Statics that show a timeline, like week 1 versus week 4, usually beat pretty pack shots. If you're looking for more DTC clients, SellTheClick has a free job board with 0% fees and job alerts.
@OlaitanRodiah Sleep masks are a good category to practise on, because the buyer already knows the pain. The statics that tend to win lead with the bad night, not the product. If you want brand work, SellTheClick has a free DTC job board with 0% fees and job alerts for ad designers.
@danoedmund Or six ads that look alike but say six different things. The column can't tell, because it only reads images and thumbnails, not the message.
Claude and ChatGPT can now run your Meta ads via MCP.
Before you hand over the keys:
β official Meta connector only, not random GitHub servers
β read-only first: reports before edits
β no dozens-of-edits-an-hour automations
β label AI-written ads
Fast is great. Banned isn't.