Best Practices

The distilled "how we make creative" rules for Mucho paid creative, assembled from every recorded creative-strategy source (Notion metastudies + retros, Fireflies meetings, Looms, Slack) as of 2026-10-03.

How to use this note: this is the rules layer. The evidence behind each rule — verbatim quotes, per-source extracts, dates — lives in [Creative strategy feedback highlights](../specs/2026-10-03-creative-strategy-feedback-highlights.md). Where a rule depends on a specific audience or test period, that is called out. Companion notes: [Creative Operating System](2026-10-03-creative-operating-system.md) (the planning/testing system) and [Script power language bank](../specs/2026-10-03-script-power-language.md) (the vocabulary).

The one-line thesis

Pick a specific audience and a testable value proposition; make the visual, text and audio work together to earn attention; then give the viewer a clear reason to stay, believe and act. Measure past the hook, and treat the landing page and offer as part of the creative system.

When to run this

Any time a brief, script, static or video is being written, reviewed or graded.

The rules

1. Measure the whole funnel, not the hook

• A better hook metric is not automatically a better ad. The July 2026 retro raised hook rate 18.7% → 26.9% and ROAS fell 1.08 → 0.95. "We optimised the half of the funnel that was already working."
• Track attention, hold, click, landing-page fit, and conversion together. Do not promote CTR or hook rate to the headline success measure without checking downstream.
• Relatability earns the stop but gives no reason to act. In the July data, If-Then and Contrast converted ~2.4× better than Relatability.
• Statics are the conversion engine (Lifestyle Image 5.58%, Headline 6.30%, Feature-Benefit 6.08% CVR in July). Scale them; don't treat them as filler.
• BOFU converted ~2.4× better than TOFU on far less budget. Build a real BOFU pool instead of concentrating spend at the top.

2. Audience first — pick one, then fit the idea

• Choose the avatar/sub-avatar and angle first; design the hook and script around them.
• Three umbrella avatars: Fitness, Busy, On-the-Go (~11–12 sub-avatars mapped). "Avatar and angle is highly, highly correlated with performance."
• Current direction (from Sep 28 → Oct 2, 2026): concentrate ~80% of concepts on Fitness Curious — interested in fitness, not yet a committed hard-gainer. Do not default to skinny-guy / "I forget to eat" language; the problem is eating enough of the right things and staying consistent.
• Bulking drove performance but is "very exclusionary… we can't break past a certain level." GLP-1 and Blue Girl were abandoned (performance + Meta-restriction risk).
• Make "busy" cascade through the whole script, not sit as a side element.
• Research the audience as a person: what they want, how they speak, what formats they watch, what groups they identify with. Use comments (TikTok/Reddit) as research material.

3. Make the value proposition distinct and honest

• "Convenient" is not enough — ask "how's ours uniquely convenient?" Name a specific mechanism.
• Known tension to resolve by audience, not by rule: the older bulking metastudy says this buyer wants Confidence, not Convenience, and that "High Protein" is a mechanism rather than an angle; the newer (Sep 2026) direction leads with convenience and taste/texture for a broader audience. Choose the value for the audience and funnel stage, then test it.
• Taste is under-used: "you're not going to be consistent if it tastes bad."
• Be precise about claims. Prefer protein count over vague "clean ingredients" (which implies whole-food); drop claims that aren't literally true (e.g. "No Seed Oils" while Maltitol is present). Verify nutrition/ingredient claims before use.

4. Hooks: design all three attributes

• A hook is text + visual + audio.
  • Visual = the thumbnail; it usually does the most to stop the scroll and target the viewer.
  • Text = must pass the YouTube-title test: the viewer understands why to watch.
  • Audio = create emotion, tension or stakes — not decoration.
• Build an open loop (a specific unresolved question), not just a louder opening.
• Raise magnitude, but make stakes feel real — a loud sound effect is not a stake.
• Keep text, visual and audio aligned to the same avatar and idea.
• Three hook variants per concept is the working standard. Diversify across genuinely different emotional states / viewer types, not just reworded premises.
• Steal from proven organic (high-view videos) rather than inventing from scratch.
• Check valence: the tier-list/ranking format creates intrigue, but reframe negative frames ("mistakes you're making") toward a positive result/discovery.

5. Angles, concepts and mechanisms

• Keep them distinct: avatar (who) → angle (why they care, the value) → concept (the specific idea) → mechanism (how the product delivers it).
• One sub-avatar, one angle/value, one main idea per video. Multi-angle scripts make a win or loss undiagnosable: "if this doesn't perform, you wouldn't know which one."
• To test alternatives, build a sub-batch: keep the script stable and make separate versions with different angles.
• Angles are concise — often 3–5-word value propositions — grounded in a real audience insight.
• Universal ideas still work if refreshed with a new mechanism, format or psychological framing. Everything must ladder to one unique mechanism.

6. Copy: sound like a person

• Natural spoken language; read it aloud. If a real person wouldn't say it, revise.
• Short sentences, active voice, simple language (~second-grade level). No run-ons.
• Slang only when it genuinely belongs to that audience/format — not generic "brainrot."
• Cut filler; don't explain what the viewer can infer. The hook→body bridge is where drop-off is worst — keep it tight.
• Introduce the product early in a natural segue.
• Loop the selling points through the script; don't word-vomit them at the end.
• Match the copy structure to the format.
• Stay on one angle throughout — the rest of the script should reflect the opening.
• Recurring proven vocabulary (use as source, not mandatory): "elite macros," "cheat code," "shortcut," "lifesaver," "hack," "real meal," "not a snack, a meal."

7. Formats, creators and production

• Don't chase novelty for its own sake. Use repeatable formats to get enough concepts to learn, chosen to fit avatar + stage + idea.
• Current hypothesis: combine organic-looking hooks with longer VSL-style copy (organic hooks stop but don't convert; long copy explains but has weak hook/hold).
• Transformation ads: clear, dramatic before/after; match the reveal to the beat.
• Every static tests a hypothesis — an image that tests nothing has little purpose.
• "Show, don't tell" — use a visual when a visual makes the point better than text.
• Creator/physique is a high-impact variable. Match the person to the avatar, story and credibility the script needs. Five Fitness-Curious archetypes were proposed (everyday transformation, approachable man, edutainment trainer, comedy/skit, realistic student/9-5).
• Authenticity check on AI: overly polished or visibly AI-generated faces draw negative comments in fitness. Give AI characters distinctive features and make the product itself look accurate. The rule is not "never use AI" — it's use it with human direction and an authenticity check.
• Brief the why behind each shot, not just the shot. Editing: tight intentional cuts, contrast in rhythm, music/SFX that support mood, legible text in safe zones, match visuals to lines, no reused clips, appetizing/accurate product imagery, native/lo-fi execution where the concept calls for it (not as an excuse for carelessness).
• Mucho's creative direction is "controlled chaos" — interpret through the specific brief and brand system, not as random noise.

8. Testing and learning

1. State the hypothesis — what we believe, why, what evidence supports it.
2. Choose the one variable the test is meant to learn.
3. Hold everything else stable — changing avatar, angle, hook and format together makes the result unreadable.
4. Run enough concepts — roughly 3–5 for a meaningful read on an important variable.
5. Review the whole funnel — attention metrics diagnose; conversion and LP fit decide.
6. Iterate forward — build on what the test taught; no sideways/backwards edits.

• Use ICE (Impact, Confidence, Ease) to prioritize tests; buckets are Moonshot / Net New / Iteration; outputs are Winners and Unicorns. Spend thresholds vary by date (winner ~$30–50k, unicorn ~$300k) — treat as historical definitions, not a fixed number.
• Don't compare ads for different avatars as interchangeable variants. Match the landing page to the audience and promise — strong creative fails when it sends people to a mismatched page.

9. Process and team

• Pitch → script → produce, each with a gate. Pitch in one or two sentences; if the concept can't be summarized clearly it's probably trying to do too much.
• Review the batch hypothesis and concept before scripting; gate the script before production.
• Close the loop with reporting + a retro, and record the learning in the metastudy system.
• AI is for QC, reporting, alerts and iteration; keep concept development and final judgment human-led. A one-shot AI draft is not strategy — supply the system, evidence, audience fit and human review.

What "done" looks like

A brief/script/asset that: names one avatar and one angle; has a distinct, honest value proposition and one unique mechanism; designs all three hook attributes around an open loop; reads like a person; loops its selling points; fits its format; and carries a stated hypothesis with the other variables held stable.

Failure modes

Caveats

• Several 2025 weekly-review transcripts are empty; they support no conclusions. The richest internal strategy material is Sep–Oct 2026.
• Historical numbers vary (winner/unicorn thresholds, format results). Check the date and context before quoting a number as current.
• Some rules are audience-dependent and can appear to conflict (e.g. Confidence vs Convenience). Test by audience rather than treating either as universal.
• Loom/meeting transcripts contain ASR errors; verify quotes against source before external use.
• Partner-agency frameworks in the records are not automatically Mucho doctrine.

Related

• [Creative strategy feedback highlights](../specs/2026-10-03-creative-strategy-feedback-highlights.md) — the evidence behind every rule here.
• [Creative Operating System](2026-10-03-creative-operating-system.md) · [Product Messaging Guide](../specs/2026-10-03-product-messaging-guide.md) · [Script power language bank](../specs/2026-10-03-script-power-language.md)
• Source: Notion → Growth → Creative Strategy Handbook (Metastudies, Retros, Agendas); Fireflies; Loom; Slack.