Alphabetsoup 🥥 🌴

🏳️‍🌈🎮🤓👬

[Alphabetsoup 🥥 🌴 @azuniga1187]

This is when I was feeling my oats. 🥰

[Paul🦋 @pjohnson360]

An old pic you still like …

[Alphabetsoup 🥥 🌴 @azuniga1187]

@zackpaid Fable

[Alphabetsoup 🥥 🌴 @azuniga1187]

@0xROAS guide

[Shaun Eng @shauneng]

Your winning ad dies. You launch more variations. Try more volume. Nothing works. Move on to the next thing without figuring out why.

Six months later you're still looking for the next winner.

Bro in Evolve went from $1k/day to $130k/day in 30 days.

Not by copying competitors or spamming AI ads. By running 1 ad concept a day with intense feedback loops.

Most of you think you're doing feedback loops. But you still can't crack a new winning ad because your learnings look like this:

"Hook too boring"
"Improve CTA"

These are not learnings. These are shallow observations.

Just because you do more reps doesn't make you better. Only perfect practice makes perfect.

The problem is you still end up making the same ad over and over again.

Every creative you run is you paying Facebook to understand your audience better. If you don't analyze it afterward, that spend was wasted twice. Once on the ad. Once on the learning you never extracted.

When an ad dies, spend 20-30 minutes on it.

Did the hook stop the right person? The first 3 seconds need to call out the exact avatar you researched. Too broad, wrong visual, wrong language, they'll click off. Not "hook good, hook bad." Break down why.

Was the whole ad congruent? If the hook speaks to one person and the body speaks to another, the whole ad breaks regardless of how strong each part is individually.

Was it the idea or the execution? Wrong visuals for the avatar. Listing features instead of telling a story. No emotional response. That's execution, not a reason to kill the concept. Run 2-3 iterations fixing execution before you move on.

Most of you only study winners. Wrong move.

You have more losers than winners. More data there.

Chances are you've been making the same execution mistakes across every losing ad and never caught them.

Log the specific reason after every test. Winner or loser. Was it the hook? The avatar? The story? The visuals? The language?

Build a checklist. What's working. What's not. Review it before you create the next ad.

Before you brief anything: Why am I testing this? Who exactly am I targeting? Why will this resonate with them specifically?

If your answer is "it's working for a competitor" that's not a reason.

Another Evolve member Spencer scaled to $100k/day doing less than 7 ads a week. Just intense feedback loops on every single one. We broke down exactly how on a podcast on our YT.

The brands finding winners consistently might not be running more creatives than you. They just compound every learning into the next brief.

You're a few months into looking for your next winner.
But the answer is in the ads you already ran.

Go back and actually study them with intent instead of half assing it.

[Shaun Eng @shauneng]

42 brands in Evolve hit their first $100k day with the same playbook.

I posted this 1.5 years ago
⠀
Yet 90% of you still aren't doing it.
⠀
You could 2x your revenue in 24 hours. No new ads. No new product.
⠀
One change: Ad-Based CRO.
⠀
Let's say you're running a supplement brand at $20k/day.
⠀
You find a winning ad about energy for busy moms. You scale it.
⠀
Then you send every customer to the same generic page. They land. They don't see what made them click. They bounce.
⠀
You paid for that click.
⠀
That's the belief gap.
⠀
The fix is simple. Build a page for that exact angle. Energy for busy moms. Same avatar, goes deeper.
⠀⠀
If you hit a winning ad, you may be able to double your spend. Hit a winning landing page on top, you could double your spend AGAIN. Now you just 4x’d your spend.

Same product. Same margins.
⠀
1.5 years ago you couldn't just spin up a congruent landing page in minutes. Now you can with Claude code or Manus. No excuse.
⠀
Yet most of you still do this:
⠀
Ad talks about energy for busy moms.
Product page talks about "general energy support."
Cooked.
⠀
The page's job is to kill every doubt before they think it.
⠀
"Will this work for me?"
"Is this actually for my situation?"
⠀
Answer it before they ask. By the time they hit buy, they're already convinced.
⠀
9-figure brands like Ridge, Hexclad, AG1 run many pages for the same product. One for every angle they're scaling.
⠀
Ad about durability? Page about durability.
Ad about lifestyle? Page about lifestyle.
⠀
Why split test button colors when you can build a page for an angle that's already working?
⠀
Ad-Based CRO > random split tests.
⠀
Treat landing pages like creatives. Take your best ad, build different pages, test until you find the winner.
⠀
The content comes straight from your customers. Reviews. Objections. The exact words they keep using. Stop writing what you think they want to hear. Read 50 of your own reviews. Pull the phrases that repeat.
⠀
There's never too long. Only too boring. Right avatar, right page, and they'll read every word.
⠀
Test them IN Facebook. Not in some split testing tool where you wait weeks for data.
⠀
Two ways:
⠀
New creative batch - 3:2:2:2. 3 creatives x 2 body copies x 2 headlines. Duplicate all with a new landing page. 6 ads per ad set. Let Zuck pick the winner.
⠀
Existing winners - pull your top ads, send them to a new page as a separate test inside your main campaign. This is where most of your potential winners are sitting right now.
⠀
Every test compounds.
⠀
Same people who test 3 creatives and quit will test one LP and say "this doesn't work."
⠀
Meanwhile 9-figure brands are running 50+ LPs. Your competition is building pages while you're changing button colors. And you're wondering why CPMs feel impossible to make work. You have to squeeze the most out of each click.
⠀
Few

[DTC Midas @DTCMidas]

Here's how to structure a high converting product page:

1. Announcement Bar: "Back to Work Sale: Buy 1, Get 1 FREE + Free Shipping"

2. Product Media Gallery (Horizontal Funnel):

- HQ Product Image

- Product In Use + Benefits Overlay

- Before & After Transformation

- What's in the Package / Ingredients

- How It Works (3-4 steps)

- Us vs. Them Comparison Chart

- Review Screenshot / Quote

- Lifestyle shots (if suitable)

3. Product Title, Reviews, and Quick Benefits:

- Clear, keyword-rich product title

- 5-star reviews + count (★★★★★ 4.8/5 based on 123 reviews)

- Short description (1-2 lines)

- 3-6 USP bullet points

4. Offer & Purchase Section:

- (Bundle) pricing + strike-through higher price (~30% markup)

- Bold Add to Cart button

- Guarantee + shipping estimation

- Micro-trust signals ("Ships in 24 hours," secure payment icons)

- Optional: Free gift unlock ("Free X when you buy 2+")

5. UGC-Style Product & Testimonial Videos

6. FAQ Section Handle objections: ingredients, usage, sizing, shipping, "will this work for me?"

7. Press Logos / "As Seen In"

8. Desired Outcome Section
- Large lifestyle image
- Outcome-driven headline
- Short benefit text (2-3 lines)

9. Testimonial Section

10. How It Works / Timeline: 3-4 visual steps or realistic results timeline (week 1, week 4, week 8)

11. Us vs. Them

12. Risk Reversal: "X-Day Risk-Free Guarantee" callout

13. Review Section

[Alphabetsoup 🥥 🌴 @azuniga1187]

@JustinWinches16 @Mho_23 this a weird ass comment. lol

[Alphabetsoup 🥥 🌴 @azuniga1187]

@Aaroo06 @Mho_23 claude is not used for Image generation. Each LLM model out their has their strong suites, and claude is for copywriting, coding, and now analysis on images. Gemini(nano banana) is king, but that's my opinion. look at "LMArena" to keep up which model is leading in category.

[DTC Midas @DTCMidas]

This is how to test products from scratch.

Start with coming up with 4 different angles for your product.

Create one generic funnel that works for all angles.

Something like:

ad > pre-sell page (if product needs more education, other wise not needed) > product page > simple cart > checkout.

Set up a CBO with 4 adsets (1 per angle), 3 ads per adset and add. $10-15 min spend per adset. Minimum $100 CBO budget, more if your AOV is higher then $50.

Let it run for at least 3 days, preferably 7 days, then check which angle performed best.

No traction?

Rinse and repeat with 4 new angles.

One angle shows promise?

Double down. Create 4 new adsets with different concepts based on that winning angle.

NOW you rewrite your funnel to match that specific angle.

Most people do it backwards.

They build elaborate funnels for untested angles and wonder why they fail.

Test angles first, optimize around it afterwards.

[Alphabetsoup 🥥 🌴 @azuniga1187]

@davidfigeira Who is your video editor? Or I’d like to learn to make these type of videos 😭

[Alphabetsoup 🥥 🌴 @azuniga1187]

@beyoncegarden The hand movements say a lot “like what was the purpose!?”

[Biru @birufgc]

This is lowkey the video that made me wanna get into fighting games lmao

[Miko @Mho_23]

how to generate AI photos that look 100% real:

nano banana normally has a very grey scale color grade which is why we are going to be using inspiration photos for color grading and creative direction

> go on pinterest and grab any photo you like that has the aesthetic you want

> upload that pinterest photo into chatgpt 5.1 thinking mode and ask for this:

"analyze this photo and give me a very detailed json prompt that can recreate it, it should be very detailed, really make sure to break down the color grading and all the exact colors in the photo"

the reason we're using chatgpt is because of its very strong photo analysis capabilities + its less lazy than gemini when writing json code

we're not using claude because it doesnt do actual thinking photo analysis, which is important for good json with this method

> take the json prompt gemini gives you and put it into nano banana pro in google ai studio

> upload your product image and prompt: "using this json as reference, generate a person holding my product [attach product image]"

> nano banana generates the photo with that person holding your product

you will notice the photo looks more realistic than usual and the color grading will be really good

> save that generated character photo as your reference

> for future generations, attach the character photo each time for facial consistency

now you have a consistent ugc model that can hold different products across multiple photos

you can also interate the base json prompt and chat back anf foruth with chatgpt to create a json prompt based on what you want, then repeat the same process..

i like this because it makes the color grading really good in the photo

the #1 sign a photo is generated with nano banana is the grey colours, but with this method it removes that

this is how you can build a library of high quality ugc content without hiring creators

the json from chatgpt handles the realistic lighting and color grading, nano banana handles the character consistency, and you control the product placement

takes 5 minutes to set up, then you can generate unlimited variations

[Machina @EXM7777]

https://t.co/XXdYeXQwAo

[Alphabetsoup 🥥 🌴 @azuniga1187]

@silenth7lls Sameee. 😂😂😂

[Mesh🇧🇧 @rahsh33m]

Because why do we do this 🤣

[Alphabetsoup 🥥 🌴 @azuniga1187]

@PEACHYKEENCUTZ @ArtOfDialogue_ Ok girl! Go off, I guess! 😂

[Alphabetsoup 🥥 🌴 @azuniga1187]

@tnysl New map is so laggy. Can’t play hitscan characters 😭

[CARDIA 🐦‍⬛ @CCardib2]

Barbz swore they was finna get another album when whole time cocaine barbie made up a fake album date just to take attention off Cardi after AITD dropped 😂😂😂