Hey everyone!
There are tons of myths around the "correct" Meta Ads structure. In my experience, a completely unique structure can work every single time.
Right now my team is testing the following pipeline: one campaign, one ad set, 50 creatives.
There are examples of this approach working well in other products – now I want to replicate it at scale.
But before launching the first campaign, it's important to do the groundwork.
Step 1. Technical Check
First – verify the technical setup:
Is the MMP integrated correctly? Are events flowing from the MMP to the ad accounts?
Is attribution set up? Are the attribution windows configured?
Is revenue being passed correctly?
Are subscriptions tracked? Are trials tracked – and if so, all of them or only those that lasted longer than an hour?
In 90% of cases when something doesn't work, the reason is the same – a misconfigured integration. So check the technical side before launch, not after.
Step 2. Competitor Research
I don't like launching ad campaigns in a niche with zero successful case studies.
It makes finding your first ad concepts harder – and there's a chance paid traffic simply doesn't work in that niche at all. That happens too.
The fitness niche is the opposite situation: there are tons of competitors.
Before launch, you need a fairly large number of creatives. I start with at least 20–30, ideally 50, to fill the first ad set and give Meta Ads the chance to test everything and pick the winners.
Where do you get ideas?
Meta Ads Library
Look at what's working for your competitors:
Pick the top 10–20 highest-grossing apps in your niche.
Find them in the Meta Ads Library and go through all their creatives.
Write down the concepts that could work for you.
It's important to understand: which angles your competitors use, which visual elements and/or animations they apply, whether they localize their creatives, whether they focus on English-speaking or Spanish-speaking markets.
All of this becomes the foundation for your own creative work.
TikTok Organic
Here you can borrow interesting concepts from influencers. Look at what's happening in your niche:
what's getting views
which hooks are being used
what pain points the audience has
what solutions exist
On TikTok I rarely find ready-made videos that can be used 1:1 for Meta Ads. In most cases, organic content is a poor fit for paid ads.
However, TikTok videos help you create your own creatives that stand out from your competitors'.
Step 3. Creative Production
Here I use several approaches.
In the past, I would take competitors' creatives and simply copy them, swapping in my product. This method still works today – it significantly speeds up and simplifies the launch.
But a lot of this can now be done with AI. Especially if your competitors have a large volume of creatives – chances are they're using AI to generate them too. You can replicate that process.
For my case, I combined both approaches: I used competitors' creatives as the base and added AI generation (image-to-video) to produce my own.
I rarely start my first launches with my own ideas.
As a rule, first ideas are fairly shallow and obvious.
That's exactly why they don't work.
Early on, it's far more effective to borrow ideas from competitors.
Step 4. Testing
Then the actual tests begin.
The goal at this stage is to cover the maximum number of concepts, different in both idea and visual execution, launch the first campaign, and see what sticks in Meta Ads.
How the Tests Work
Creative testing used to be a two-step process.
First – CPI testing to gather initial stats and decide whether the creatives were worth sending to the main campaigns.
Then – a test with purchase optimization.
Today CPI testing is no longer relevant.
A creative that performed great in a CPI campaign can perform terribly in an event-optimized campaign – and vice versa.
And if you can't trust that data, why keep this stage at all?
The old playbook also said you had to spend $100–300 on every creative (depending on your event cost) just to confirm it doesn't work.
Now I believe a different approach should work: if a creative doesn't get spend from the algorithm – it's a bad creative. The absence of spend is itself a test result.
That's exactly why I load a large number of creatives into one ad set, one campaign, and don't worry that many of them won't get any spend.
Some portion will get spend anyway. And if a creative doesn't resonate and Meta doesn't serve it to the audience – it's not a creative worth developing further.
During testing, I look at the entire funnel:
CPM – which audience we're reaching.
CTR – I want to see at least 1%.
Click-to-install conversion – for Meta, that's 25–35%.
Hook rate – the percentage of people who watch the first three seconds. I want to see 30–40%.
Install-to-target-action conversion – heavily depends on the niche, but aim for 10–15% to trial, 30–40% from trial to paid, or 4–8% straight to paid if you have a paywall without a trial.
Almost always, if you're underperforming on one of these metrics, the problem is a bad creative:
Low hook rate – the first 3 seconds don't grab attention. The user needs to instantly understand what the video is about and want to keep watching.
Low CTR – unclear call to action, unclear what's being advertised.
Poor click-to-install conversion – you're simply reaching the wrong audience, or your video is misleading.
Your job is to learn which concepts, visuals, and creative structures either don't work at all or need refinement – and then start working on each specific part of the creative, trying to improve it.
That's all for today.
In the next issue, I'll show you how I built a creative generation pipeline that now lets me produce dozens of creatives per day.
