Leveling up your creatives
Lesson 8 of 16 in the course “Marketing 2027”. Free, with animations and a quiz.
Why 10 similar videos count as one
In the first lesson we saw that Meta’s delivery system picks who sees each ad on its own. If you upload 10 videos with the same girl in the same room and change only the text, to the system it’s almost the same creative: it reaches the same people. Marketers notice this effect all the time, and Meta’s own Advantage+ advice directly recommends creative diversity. The logic is simple: the creative is now the targeting. Different people on screen, different places and formats help the system find different groups of buyers. Ten different videos are ten chances to find your people.
The diversity matrix
To make videos truly different, use a matrix with four axes. Format: UGC video (shot as if by a regular person on a phone), photo, carousel. Person on screen: a student, a skater, a teen’s mom. Setting: dorm, bus stop, park. Main benefit: warmth, style, price. Three options on each axis give 81 combinations. You don’t need to shoot them all: pick 10 that differ on at least two axes at once. Then each video talks to a different person about a different reason to buy. That’s the kind of diversity the system can use, not 10 shades of one video.
One product, different reasons to buy
The strongest axis of the matrix is the person and their reason. People buy the same hoodie for different reasons. The student doesn’t want to freeze in class. For the skater it matters that the fabric survives falls. The teen’s mom wonders whether it’ll survive a hundred washes. That’s three different videos with three different hooks, even though the product is the same. How do you learn the reasons? Don’t invent them, read them: reviews, DM questions, comments under your ads and your rivals’ ads. People write why they buy. Your job is to hear those words and put them back into your ads.
One winner, eight hooks
The cheapest way to add diversity is to change only the opening. Keep the body of the video that has already proven itself, and reshoot the first 3 seconds eight ways. Question: “Freezing in class?”. Number: “380 grams per square meter, what does that mean?”. Before/after. Mistake: “Don’t buy a hoodie until you check this”. POV: shot from the hero’s point of view. Test: the hoodie in a freezer. Review: a real customer quote. Comparison: next to a thin sweater. That way you test hooks almost for free, because the main edit doesn’t change.
Cadbury case: diversity at scale
Cadbury in India showed how far diversity can go. During the pandemic small shops were losing customers, and in 2021 the agencies Ogilvy and Wavemaker made the #NotJustACadburyAd campaign. Bollywood star Shah Rukh Khan appeared in Diwali ads urging people to shop at local stores and named them one by one. Using machine learning, they made 130,000 versions of the ad for more than 2,000 shops: groceries, clothing, electronics, furniture. Each version was shown to people near that shop. The result: 94 million views on YouTube and Facebook. The lesson: one strong pattern plus many personal versions beats one generic video.
How to grow a winner
Found a video that sells? Don’t duplicate it ten times, grow it step by step. First, new hooks on the same video body. Then the same words but different people on screen. Then a different format: turn the video into a carousel or photo. Change one big axis per step, otherwise you won’t know what worked. And give files clear names right away: format, person, setting, benefit, hook type. A month later your report won’t show “video_final_2_new” but that students in dorms with a question hook sell best. That’s knowledge that grows along with your budget.
Do it yourself: 10 different concepts
Build your own matrix. Write down three people with different reasons to buy, taking the reasons from real reviews and messages. Add three places from their actual lives and three formats. Now put together 10 combinations that differ on at least two axes. For each, sketch a one-line hook. Name the files right away using “format_person_setting_benefit_hook”. That wraps up the second level of the course: you can find other people’s wins, break them down, make an honest remix and grow different versions from it. Next up: making creatives faster with AI and not fooling yourself in tests.
Check yourself
- Why did 10 videos perform like one?
Correct answer: To the system they’re nearly identical and reach the same people. One girl, one room and one format is one creative with different captions. Diversity has to be in the people, places and format. - Which set is more diverse?
Correct answer: Student in a dorm, skater in a park, mom at home: video, photo, carousel. Here the person, setting and format all change. Each version talks to a different buyer about a different reason. - What should she do with the winner she found?
Correct answer: Make versions, changing one big axis per step. That way the winner grows, and Nika sees exactly which axis drives sales: the hook, the person or the format. - What’s the key lesson from Cadbury?
Correct answer: One strong pattern and thousands of personal versions for different places. There was one idea: support local shops. Machine learning just multiplied it into 130,000 personal versions, each speaking to people near a specific shop.