Code what is visibly or audibly present, then write a testable hypothesis. Do not label an element “good,” infer an audience from appearance, or claim an algorithm read a particular meaning. All examples below are synthetic illustrations, not client results or performance benchmarks.
The Hook
The first meaningful visual, line, or action—not a universal three-second formula.
Can the intended viewer identify the situation, product, or benefit quickly enough to decide whether to continue? Test comprehension and qualified attention; do not assume the hook selects an audience.
Visual Style & Composition
Subject, context, lighting, framing, legibility, and representation.
Does the visual make the product and claim understandable while matching the actual customer experience? “Premium,” “authentic,” and “high intent” are interpretations to validate—not qualities an image automatically creates.
| Dimension | Record objectively | Review risk |
|---|---|---|
| Context | Product-only, in-use, environment, people shown | Staged use, unrepresentative outcome, unsafe use |
| Composition | Focal point, scale, crop, safe-zone placement | Hidden qualifiers, illegible labels, cropped captions |
| Representation | Who appears and how they are portrayed | Stereotypes, tokenism, inferred protected traits |
| Asset provenance | Creator, license, consent, allowed channels and term | Unlicensed stock, creator likeness, trademark misuse |
Keep the source file, release, license terms, expiry, territory, paid-media permission, edit permission, and AI-generation disclosure decision with the asset. A model producing an image does not establish that every depicted logo, person, style reference, or training-derived element is safe to publish.
Copy Structure
Promise, support, qualification, disclosure, and next step.
Can a reasonable customer understand what is offered, to whom, on what terms, and with what evidence? Clear copy is more valuable than a “formula.”
| Angle | Synthetic illustration | Required review |
|---|---|---|
| Problem → feature | “Separate wet gear with the lined compartment.” | Verify the feature and show representative use |
| Benefit | “Carry a 13-inch laptop in the padded sleeve.” | Test the stated size and note exceptions |
| Social proof | “Rated 4.6/5 from verified buyers as of 20 Aug 2026.” | Substantiate denominator, date, method, and permission |
| Availability | “Offer ends 31 Aug at 23:59 EET.” | Use the real deadline and honor it consistently |
| Story | “Designed after our sample testers requested a wider strap.” | Confirm the account and avoid fabricated testimonials |
Emotional Frame
The feeling invited by the message—not a lever for manipulating vulnerable people.
Format Selection
Choose a format for the communication job and placement constraints, then test it.
| Format | Candidate use | Accessibility / quality check |
|---|---|---|
| Static | One product or proposition | Readable contrast, concise alt-text handoff, visible qualifiers |
| Carousel | Sequence, variants, or steps | Each card makes sense; order and disclosures survive swiping |
| Short video | Demonstration or one idea | Captions, audio description where needed, safe zones |
| Longer video | Comparison, walkthrough, or story | Transcript, chapters where supported, claim context retained |
| Stories / vertical | Full-screen placement treatment | Controls do not cover text; offer terms remain readable |
Pacing & Rhythm
Shot duration, information density, transitions, and time allowed to read.
Can viewers follow the demonstration and read every material qualifier? Faster cuts are not inherently better. Test pace by placement and audience, and include comprehension or quality guardrails alongside watch metrics.
Sound & Music
Voice, music, effects, silence, captions, and the rights attached to each.
Does audio add clarity or emotion, and does the asset still communicate when muted? A popular sound is not a deterministic reach mechanism.
- License: confirm commercial paid-media use, territory, duration, edit rights, and platform-specific restrictions.
- Captions: provide accurate speech captions and identify meaningful non-speech audio.
- Mix: keep speech intelligible and avoid startling volume changes.
- Experiment: compare licensed music, voiceover, and silent/captioned treatments under the same decision plan.
CTA Design
A clear, truthful next step aligned with the destination.
- Match the verb to the actual action: “View sizes,” “Compare plans,” or “Request a quote.”
- Repeat only when it improves clarity; do not obscure content with persistent prompts.
- State price, recurring terms, eligibility, expiry, or stock conditions where material.
- Never use a fake countdown, reset a timer per visitor, or state inventory that the system cannot verify.
- Confirm keyboard, focus, contrast, tap-target, and screen-reader behavior on the destination.
Trust, Locale & Fulfilment
Evidence and operational facts customers can verify—not decorative badges.
From Checklist to Experiment
The checklist becomes useful when it produces a controlled, reversible decision.
- Change one interpretable treatment or use a design that can estimate interactions.
- Predeclare assignment or holdout, primary metric, contribution assumptions, guardrails, minimum sample, and conversion/return lag.
- Do not stop at the first favorable slice; document exclusions and multiple comparisons.
- Require a named human to approve expansion. Keep the prior creative available for rollback.
- State incrementality limits whenever the result is platform-attributed rather than causal.
Built by @itsmazinzaki — AVAMARTECH