Meet your specialized AI agent for scaling social marketing.
Social Marketing
Joined December 2024
- Tweets1.1K
- Following56
- Followers1.9K
- Likes865
Salt & Stone runs 23 versions of the same image ad at once.
Three of them clear our winner bar. Twenty do not.
We pulled 566 of their 568 active Meta ads and scored every one we could.
If you run an ecom store, the useful part is not their concept list.
It is that the same concept landed on both sides of the line inside their own account.
Lifting a competitor's idea does not transfer their result.
Our playbook, not a finding from this library:
get more ideas in front of the algorithm before you pour budget into any single one.
Superscale runs that loop.
The competitor research surfaces the live ads in your niche and what each one scores.
The Agent then builds variants of your strongest idea, and you steer it by chat: change the hook, swap the character, shorten the second scene.
Superscale retweeted
This is exactly why publishing ads based on data is so important instead of blindly testing and hoping to get more than 1 winning ad every 3 months.
Are you testing your ads long enough?
From 1,069,340 new Meta ads we analyzed, only 1 in 10 ever reached €500 in spend.
That is roughly enough to tell whether an ad performs or not.
And it is still far from a winning ad.
Of everything launched, only 1 in 38 became one.
In this thread:
>what counts as a winner
>why 9 in 10 ads never get judged
...and what that does to your quarter.
Are you testing your ads long enough?
From 1,069,340 new Meta ads we analyzed, only 1 in 10 ever reached €500 in spend.
That is roughly enough to tell whether an ad performs or not.
And it is still far from a winning ad.
Of everything launched, only 1 in 38 became one.
In this thread:
>what counts as a winner
>why 9 in 10 ads never get judged
...and what that does to your quarter.
How many of the ads you launched last quarter ever crossed €500?
That's your real test rate. Everything below it never got a verdict.
Everything we learned from a month of tearing down premium ecom ads, in one rule:
Find the format your positioning owns.
Then scale variants inside it.
Do not scatter across formats.
The evidence:
>Mr Marvis with 20 of 20 winners in one format.
>lululemon with 30 of 30 winners as statics.
>Aesop with 12 of 12 as produced video.
>And at the volume end, Gymshark runs 113 simultaneous variations of a single concept, a top grooming brand 38.
Concentration finds the winner. Variation scales it.
That loop is what Superscale runs for you: test variants on real data, read the Meta performance feedback back into the Agent, kill what loses, generate the next round from what wins.
One format, many variants, compounding data: superscale.ai
We analyzed 167 advertisers who run both static and video ads to see what performs better.
And we found that every "video beats static" benchmark you have read was probably built the wrong way.
Here is how they are usually made: take all the video ads from hundreds of advertisers, take all the static ads, compare the two piles (pooled dataset).
The catch is that different companies sit in each pile.
If the brands that run more video also happen to buy cheaper audiences, video comes out looking cheap.
You measured who runs what, not what the format does.
So we did it the other way around:...
What that means if you are the one buying:
"Switch to video" is not a strategy, it is a 50/50 bet. Video earns its place when the message needs motion, a demo, a transformation, a face. Not because a chart said it converts better.
And if you moved budget into video last year, check one number. Not your lead count. Your cost per lead.
