The core model
Creative is the targeting lever. Meta, TikTok, and Google increasingly decide who sees an ad based on the signal inside the creative itself: who it speaks to, what problem it names, what proof it shows, and what behavior it attracts. So your job is not to make one perfect ad. Your job is to test messages clearly enough that the platform and your data can tell you what works.It’s a volume game
Because the creative does the targeting, growth is capped by how many fresh, distinct ideas you put in front of the platform. Most lose, a few win, and you can’t reliably predict which; even experienced buyers are wrong often. So don’t fall in love with a single ad. Aim for a steady stream of fair attempts and let the data pick the winners.Concept vs. variation
Keep two terms straight:- A concept is the idea or angle, like “your first 5k, without the dread.”
- A variation is one execution of it: a testimonial, a day-1-to-week-8 montage, a different hook image.
Build a testing matrix
Start with a small grid.
Pick one or two variables to change. Hold the rest steady.
If every ad has a different audience, format, product shot, hook, and offer, you will not know why one wins.
Where concepts come from
You won’t run dry if you keep three renewable sources flowing:Your customers & data
Talk to users, read reviews and support tickets, and look at who actually buys. You’ll uncover ICPs and value props you never planned for.
What's already working
Study ads that have run a long time, since longevity is a strong signal an ad is profitable. Reverse-engineer the angle and value proposition behind them, not the visuals.
Organic social
Ideas that spread organically on TikTok, Reels, and Shorts often transfer to paid. Watching how creators frame a problem is free fuel.
Angle archetypes to adapt
Concepts tend to take a few recognizable shapes. Treat them as prompts, then fill them with the audience and value-proposition work above. The shape is reusable; the content has to be yours.Problem → solution
Problem → solution
Open on the pain, resolve with the product.
Identity: 'this is for you'
Identity: 'this is for you'
Name the audience in the first line so the right person stops scrolling.
Founder / origin story
Founder / origin story
Why the product exists. Works when the story is genuine and specific.
Cost or effort comparison
Cost or effort comparison
The hard, expensive old way versus the easy new way.
Pattern interrupt
Pattern interrupt
An unexpected opening that stops the scroll, then bridges back to the product. Use sparingly, and always connect it.
A simple testing workflow
Pick the learning objective
Decide what you are trying to learn: which ICP, which pain, which value prop, which proof point, or which format.
Generate controlled variants
Ask Superscale for a batch where only the test variable changes.
Launch with enough budget
Give each concept enough spend or conversions to make a decision. Tiny tests create noisy conclusions.
Promote winners
Move winners into the scaling lane, then ask Superscale to generate close variations around the same learning.
Save the learning
Turn the result into context: what worked, for whom, and why.
Useful prompts
- “Build a persona x value proposition matrix for this product. Give me three ad concepts per cell.”
- “Hold the format fixed. Create five hooks testing different pains for the same audience.”
- “Analyze these winners and tell me which variable they have in common.”
- “Create follow-up variants that preserve the winning structure but test a new proof point.”
When there is no account data yet
Start with structured exploration. Use competitor research, customer language, and a small testing matrix. Launch a testing campaign that forces spend across concepts, then move winners into a scaling campaign once you see signal.Campaign structure
Turn the creative testing system into a clean campaign setup.
Media buying on Meta
Budgets, the learning phase, scaling, and reading results.