Alexander Kemperman | $50M+ Creative Strategy

After $50M in spend behind my creatives, After 2 years as an in-house lead creative strategist, 2 years at an agency, and 2 brands of my own, I decided I wanted to share the info that's in my 🧠 kemperman.com

@shaharan_akram You can but it’s not cheap

Can you pay real doctors to give their feedback on your health wellness products and use them for marketing? Is it cheap tho? And how does it work

@shaharan_akram Just to add: the first image was made with a ref image and a prompt but it still failed, even when generating x4. I got the other image in one go where it was perfect. This is a native ad you can run with a good copy if it fits your TA.

Totally valid approach and it works well for single shots. The reason I prefer JSON is that it removes any ambiguity in how the model interprets the structure. With guided prompts you're still relying on the model to infer what you mean, which is fine until you need consistent, repeatable outputs across a pipeline. JSON just gives you that reliability out of the box. And when something breaks, debugging a JSON block is a lot faster than figuring out where your natural language prompt lost the model.

I hate to break it to you but if you're not going to learn json prompting in the next 7-14 days, you will fall behind The gap in realism is very noticeable, right?

@miejscov oh and before you launch, use your gut feeling and ask yourself; will this creative do what I want it do to? If your instinct says yes, you should be ok. Might be a f* weird approach but we're humans and selling to humans so we know best :-).

how fast can I know if an ad creative is ok?

@miejscov Give it approx 3 days to check the first results, don't let your emotions make changes, that'll reset learning process. If stats look good but below breakeven, I always give it 2 more days of spend.

@ChrisLangSocial Can't get it more authentic then this. No AI can replicate this 1:1

yapped founder content while farm mogging all comp no cap fr fr

@johnmeyerscale Amazing man, research is 🔑. Keep us posted!

First Hours Of a New Brand, Ads started at 5AM First Sale after 10 Minutes. This is the first Brand I really made deep research about Pain Customer have. Before that I thought „ChatGPT please make Research about XY“ was enough. Lets see where it goes.

@MrOverpaid Did the last thing of your list 3 days ago, been procrastonating about it for a few years. Wish I started sooner but happy to see where this is going. Agree with the message entirely.

Don't delay anything Whatever you wanna do - DO IT NOW - Travel to that country - ask that girl out - start that business - post that video Just do more and think less

@kamal_razzak Are you able to share the ads library? Can't find 'Spruce', curious to see the founder ads!

@Satishdtc @mrbrandowner Are you sending native ads to your pdp directly? No prelander? That's new.

@mrbrandowner primary text copy doing all the heavy lifting here and filtering the unecessary traffic plus the PDP is perfectly dialed in

40+ women are easiest to sell too native ads be cooking hard 🧑‍🍳 dropped the full sauce: whop.com/konvert-ads/na…

@Satishdtc What is your AOV and profitable first sale?

run more native ads 💰

@IstvanicMarin Your ad is a mirror

Most ads try to sell with features and benefits. They fail. Here's what actually moves product: narrative. People don't buy because of your 5-point bullet list. They buy because they see themselves in the story. The struggle feels familiar. The outcome feels possible. That's why advertorial-style ads consistently outperform product feature dumps. Your customer isn't looking for specs, they're looking for proof that someone like them solved the problem they're facing right now. Mine your reviews for real moments. Build ads around those struggles and resolutions. That's the difference between an ad that informs and an ad that converts.

@VladislavK58016 Thats what I thought

@kmprmn Yeah the realism on the first picture with floating hand is beyond imagination

@fetchmoneyx Same output as tools like Freepik, also higgsfield had huge queues for nano banana 2 and the kling avatar output was shit with weird symbols spawning randomly. I prefer kie.ai with json prompting; cheaper, faster, more reliable and all the models.

@kmprmn or just use higgsfield

I made 30 native creatives like the example yesterday, with json prompting + research, you can steal my prompts here; youtube.com/watch?v=TduRbQ…

@SarahLevinger got some left over? ;)

⭐️New achievement unlocked: homemade beignets

@erik_mayer_ Appreciate this, thank you

@kmprmn love this content brother

Native ad copy can be a living fucking nightmare, I know Long texts, no clue where to begin.. However….😏 it carries 80% of the weight and effect of your ad. Steal my prompts to write 3 ad copies based on your customer research in under 5 minutes, with top-notch storytelling and proven frameworks. I put the link to access them in the comments ↓

@digitaladguide don't lie to us

What if I told you there is more than 1 way to succeed with meta ads 🤯

@oprydai warehouse propaganda always hits

get into debt if you must, but own a warehouse