Tag: hypothesis testing

Creating 100 ad creatives from one brief is a task that scares most marketers. But in reality, it’s not chaotic generation, but strict math: five angles of presentation, four first frames, and five formats with calls to action. The result is not just a pile of files, but a verifiable matrix of hypotheses ready for launch. In this article, we’ll break down how to turn one document into a full-fledged ad campaign using AI and clear processes.

Making 100 ad creatives from one brief means breaking the original assignment into a matrix of options: five angles of presentation, four first frames for each angle, and five combinations of format with a call to action. The result is a hundred verifiable materials. Each option undergoes brand verification, receives AI disclosure labels, and separate result tracking.

100 materials are needed not because “more is better.” They are needed when the team doesn’t know which approach will work. In short advertising, the winner is often found not in the first video, but in a combination: the right first frame, the right angle of presentation, the right format, and the right phrase.

Meta describes Andromeda as an ad selection system that uses machine learning at the stage of selecting ad candidates and helps display ads more personally. The practical takeaway for ad teams: platforms need not dozens of identical copies, but truly different materials with clear signals.

The Main Idea: You Don’t Have to Build Your Own Content Factory

To make 100 ad materials from one brief, a brand doesn’t have to build a complex production department inside the company. There’s no need to hire dozens of writers, designers, editors, AI specialists, editors, and quality managers. You can give the brief to VibeVO and get a large set of ready-made options.

For example, a brand can provide:

• a mascot image;
• a character description;
• character behavior rules;
• product information;
• examples of the desired style;
• a list of prohibited topics;
• the required number of options;
• markets and languages;
• material format.

From there, the task is broken down into production assignments. Authors and AI tools start creating options: the character says different phrases, shows different actions, explains different benefits, appears in different scenes, and works in different formats. The brand doesn’t need to manually manage each executor. It receives materials, checks the result, and pays for approved units within an agreed model.

The brand sends an image of the mascot and writes: “This character should explain how the app helps save time. We need 100 short variations: some with a speaking character, some with text overlays, some with interface demonstrations.” VibeVO launches production, creators generate variations, and the brand receives a ready-made set for testing.

The key in this approach is not to build a factory, but to set the task correctly. The clearer the brief, the faster you can get a large volume of materials.

Why Exactly 100 Materials

One material tests one hypothesis. 100 materials test a map of hypotheses.

But it’s important not to overdo it. If the budget is small, 100 variations won’t have time to gather data. In that case, it’s better to narrow the matrix to 20–30. If the budget is sufficient, 100 variations help quickly find winners and eliminate weak solutions.

The Matrix: From One Brief to 100 Variations

The basic formula:

5 angles × 4 first frames × 5 combinations of format and call to action = 100 materials

Level	Quantity	What Changes
Angles	5	Meaning and reason to pay attention
First frames	4	What the user sees at the start
Format + call to action	5	How the message is presented and what to do next
Total	100	Testable grid of variations

Google’s ABCD model highlights four principles of effective video advertising: attention, branding, connection, and direction. For the 100-material matrix, this is a useful framework: each variation should grab attention, show the brand, create a connection, and lead to action.

How VibeVO Turns the Matrix into Production

The matrix itself is just a plan. VibeVO turns it into a workflow: each angle, first frame, and format becomes a separate task for creators and AI tools.

Matrix Part	What the Production Team Receives
Angle	What idea to convey
First frame	How to hook the viewer
Format	In what form to create the material
Call to action	What the user should do
Constraints	What cannot be shown or said
Mascot / character	Who will deliver the message
Style	How the material should look

Thanks to parallel work, you can get not just one perfect material, but a whole set for testing faster. This is especially important for growth teams: the winner is often impossible to predict before launch.

An angle of presentation is not a headline. It is the reason why a person should pay attention.

Stage 1. Five Angles of Presentation

An angle of presentation is not a headline. It is the reason why a person should pay attention.

Example for an app:

Angle	Meaning
Time saving	“Do it in 5 minutes instead of an hour”
Simplicity	“No complex settings”
Money	“Don’t overpay for extra actions”
Control	“Everything visible in one place”
Social proof	“Thousands of users do it this way”

For an online store, the angles may be different: price, quality, delivery, selection, novelty. For a game: emotion, character, victory, competition, reward.

Stage 2. Four First Frames for Each Angle

The first frame determines whether the user will continue watching.

For each angle, we create four openings:

First Frame	Example
Question	“Why does this take so much time?”
Problem	“Did you forget to update the data again?”
Demonstration	Show the product in action
Comparison	“Before / After” or “Old way / New way”

5 angles × 4 first frames = 20 basic ideas.

Stage 3. Five Combinations of Format and Call to Action

For each of the 20 ideas, we create 5 presentations.

Format	Call to Action
Talking head	“Try it”
Screen demonstration	“Download”
Product shot	“View the selection”
Comparison	“Compare for yourself”
Video overlay	“Learn more”

20 ideas × 5 formats = 100 materials.

How AI Accelerates Production

AI is useful at three levels:

1. Quickly create draft scripts.
2. Generate visual options.
3. Prepare voiceovers, language versions, and banner variations.

But AI should not choose the winners on its own before launch. It can suggest options, but the team must verify meaning, rights, style, disclosure, and compliance with the brief.

McKinsey emphasizes that the value of AI is more pronounced where companies redesign workflows and measure results, rather than simply adding AI to old actions.

Brand Review at This Volume

100 materials are difficult to review manually without a system. There are two approaches.

Approach	When to Use
Full review	Regulated categories, new brand, sensitive claims
Sample review + rules	Low-risk materials, simple overlays, repetitive templates

For most brands, it is safer to start with a full review of all materials. Then you can move to a sample review if templates are stable and risks are low.

Mandatory checklist:

• no prohibited words;
• logo is correct;
• no unsubstantiated claims;
• character does not resemble a real person without consent;
• no third-party music or likeness;
• AI disclosure is present if required;
• versions are correctly named;
• tracking tags are in place.

Disclosure Metadata at Scale

With 100 pieces of content, disclosure cannot be done manually “by eye.” Rules are needed.

Each piece of content must have:

• type: image, video, voice, character;
• status: AI-generated / AI-modified / human+AI;
• required disclosure;
• disclosure text;
• tag in the file;
• internal record in the system;
• country of placement;
• person responsible for review.

This is especially important for the EU due to the transparency rules of the AI Act, which from August 2026 require the identifiability of AI content and clear labeling of specific types of synthetic content.

Placement and Tags

100 pieces of content should not be launched chaotically, but according to a clear structure.

Example name:

APP_TIME_Q1_HEAD_CTA_TRY_V01

Where:

• APP — product;
• TIME — angle of approach;
• Q1 — first frame with a question;
• HEAD — talking head;
• CTA_TRY — call to action to try;
• V01 — version.

Without such a structure, it is impossible to understand what won: the angle, the first frame, the format, or the call to action.

How to Read the Results of the 100-Piece Test

Don’t immediately look for the “best seller” among all 100. Read the test in stages.

Stage	What we filter out
First seconds	Weak first frames
View-through	Unclear or boring content
Transition	Weak calls to action
Action	Unprofitable options
Comments	Annoying or untrustworthy content

If each piece of content has little data, compare not individual videos, but groups: angles of approach, first frames, formats. Then create a second wave from the best combinations.

FAQ

Isn’t 100 pieces too much for brand review? For manual review — yes, a lot. For a system with a matrix — it’s normal. The main thing is that the options are built according to rules, not created chaotically. In regulated categories, it’s better to check all 100. In low-risk ones, you can check templates fully, and individual options selectively.

Can ad platforms understand that the content comes from the same brief? Platforms don’t “read the brief” as a document, but their systems can see the similarity of content through visual, textual, and behavioral signals. Therefore, 100 options should differ not only in color, but in meaning, first frame, format, and approach.

How much data is needed to test 100 materials? If the goal is sales, 100 materials require a large budget, otherwise there won’t be enough statistics. For initial screening, you can look at early signals: retention, clicks, engagement, cost per click. For sales decisions, it’s better to shortlist 5–10 winners and test them more deeply.

Conclusion

Creating 100 advertising creatives from a single brief is not magic, but a systematic approach. By using a matrix of angles, first frames, and formats, you turn one document into a full-fledged testing ground. AI accelerates production, while clear labels and checks ensure quality. Start small: take one brief, break it down using our formula, and run a test.