unpopular opinion: november is the worst month to creative test.
cpms climb into black friday, so an ad set funded at 4x CPA learns from an auction that's gone by december.
show me bfcm winners holding CPA in january and I'll change my mind.
everyone says pack more into a digital product so it feels worth $49.
what i keep seeing: a 3-page SOP for one task outsells a 200-page guide to a whole channel.
buyers aren't paying to learn.
they're paying to skip a weekend of trial and error before black friday.
gta 6 capture buttons work opposite ways on ps5 and xbox.
ps5: double-press create to record, long press screenshots. xbox: hold share for your last 30 seconds, tap screenshots.
meaning tapescribe gets a still to caption, not your chase.
Pay for a defined audit or a fixed build of specific flows, not an open-ended retainer. You're buying a sequence that earns while you sleep, not a calendar of sends.
Email revenue lives in four automated flows, not the newsletter: welcome, abandoned cart, browse abandonment, post-purchase. If an agency pitches campaigns before showing you those four, keep walking.
Red flags: a flat monthly fee with no revenue reporting, a template library they call proprietary, or a promised lift before they've looked at your list size and margins. Any of those and I'd pass.
Three things decide this category for a third party seller: pack count, stick material, and whether the box gets reordered. Paper sticks and bulk counts are the usual ways out of a pure price fight. Neither one is a moat on its own.
A 500 count box of cotton swabs is about as commodity as beauty gets. Same cotton, same stick, same count. Once a house brand sits in the same search results, the only axis left is price per swab, and the margin gets decided before anyone reads the listing.
The rename from Amazon Basics to Amazon Essentials matters less than who holds the box. Amazon is the seller, the owner of the private label, and the search engine in this one. Any third party selling swabs is competing against the platform's own shelf.
what kills it. Once every ad shares one budget, Meta sends the spend to whichever creative got lucky in the first few hours, often the one with the cheapest clicks, not the one that converts.
A $50/day CBO with 20 creatives isn't a test. It's the same $50 spread so thin that nothing exits learning, and you get back noise instead of a winner.
The shape is simple. One workflow, one human check, one metric you already track. Claude or GPT drafts the creative, a person picks the winners, and ROAS decides what stays live. Dull, and it compounds every week you keep it running.
The stores I see pulling ahead on AI aren't writing clever prompts. They pick one boring job: drafting hook variants off last month's best sellers, sorting support tickets into a queue, or pulling search terms nobody has time to read.
Cache hits at $0.075/M matter more than the headline input price for catalog work. Same brand guide gets resent on every SKU, so that's where the bill lives.
@blvckledge Same goes for customers. The brands that last treat repeat buyers like the actual business, and the ones chasing the first sale with hype usually pay for it later in refunds and churn.
@arnavsawantt The reply that saves a late box opens with the real tracking status and a straight ETA, not a template apology. People forgive a delay fast once they know exactly where the box is.
solo store, bfcm in 12 days, one build slot.
n8n cart recovery, or a claude agent drafting support replies you approve.
i keep claude. klaviyo defaults already cover carts. nothing covers the ticket spike.
which do you keep, and what breaks when you drop the other?