--- title: Iterating on winning ads and the weekly hook-testing cadence slug: iterating-on-winning-ads summary: Dara Denney's 10 iterations on a winner (messaging, edit, messaging-visual, format, creator, green screen over the ad, user-journey, human-desire, high-fi founder, and when to revive a loser) with the 90/10 net-new rule, plus Savannah Sanchez's weekly framework (2 concepts × 2 written × 2 visual hooks) and "same concept, new creator" tests. domain: marketing tags: [ad-creative, iterations, creative-testing, hooks, meta-ads] related: [creative-diversity-levels, verbal-hook-templates, ugc-visual-hook-formats, creative-metrics-hook-hold-ctr, paid-social-creative-testing-framework, motivator-map-why-buy-why-believe, ad-fatigue-and-creative-longevity, multiple-creative-partners-and-creative-culture, applovin-ad-creative] sources: [yt:lWpSmxuwSts, yt:76KH-GZZ5w4, yt:DqFjak694PQ, yt:4wM-i1cxtAM, yt:El5pN9zYZPA, yt:I0SPikVQdYs, yt:qoOBIPWH_qw, yt:aSF0GiHUZxo] updated: 2026-10-24 --- # Iterating on winners and the weekly hook-testing cadence ## TL;DR - Dara Denney's hot take: **90% of tests on net-new creative, 10% on iterations** — otherwise "iteration paralysis" makes everything look alike and fatigue sooner.[^1][^2] - Messaging moves results more than format/visuals — so the strongest iteration keeps the winning **message** and changes the **format/look** drastically.[^3] - Don't iterate on losers 9 times out of 10 — unless you spot a clear editing or messaging mistake.[^4] - Savannah Sanchez (<$100k/month brands): each week **2 new concepts**, each with **2 written-hook** and **2 visual-hook** variants.[^5] - Found a winning concept? Re-shoot it with a **different creator** to scale or to find who resonates.[^6] ## Details **Dara's 10 iterations (beginner → advanced):** 1. **Messaging** — keep the creative, swap the key message. Pull messaging points from ~8 other top performers. One such static iteration spent **$2M at 3 ROAS** and stayed top performer for 4+ months; only worth it on a real winner.[^7][^8] 2. **Edit (hook visuals)** — green screen, split screen, grid.[^9] 3. **Messaging visuals** — TikTok response bubble with a new headline, brand-native visuals, captions (inconclusive in tests).[^10] 4. **Format** — winning video message → a range of images, and vice versa.[^3] 5. **Creator** — same script/brief to new creators; deliberately go much older/younger to open a demographic; keep creators diverse.[^11] 6. **Green screen over the winning ad itself** — a creator reacts to the well-known ad (Oats Overnight does this); turns image winners into video.[^12] 7. **User journey** — push the same idea further up-funnel (more educational, humorous, curious); almost always extends lifespan.[^13] 8. **Human desire** — swap or inject a desire (often family).[^14] 9. **Founder ad → high-fi** with interview questions (see [[founder-ads]]).[^15] 10. **Losing creatives** — only if the miss was execution.[^4] Diagnoses also flag missing iteration: good learnings never moved into other formats or handed to creators.[^16] **Weekly cadence (Savannah Sanchez):**[^5][^17][^18] - Keep a hook database per client (hook tested each week + link to the ad) and a master list of every hook ever tested, plus a short list of repeat winners. - 2 overarching ad inspirations/week; per inspiration 2 written hooks (text on screen in first 3 s) and 2 visual hooks. More budget → 5–10 hook variants. - Test **broad vs persona-specific** openers (see [[verbal-hook-templates]]). - Narrowing to one variable is impossible at paid-social speed — vary the creator while holding format, hook and edit style similar. **Isolate what's driving the winner (Savannah, 2024–2026).** Make at least one iteration of a top performer every week. Options: keep the first ~2 s and change the body; reshoot the exact ad with different talent; same creator in a new concept; same body with a new hook.[^20][^21] Read the metrics first. A high click-to-purchase ad with a weak thumbstop gets a new hook on the same body and CTA. A scroll-stopper that doesn't convert gets a stronger CTA, new talent, or a landing-page check ([[creative-metrics-hook-hold-ctr]]).[^22] **Same story, new face (Barry Hott).** One of his favourite iterations is having someone else tell the same story. It can fail if the new person tells it slower or faster, or without real excitement.[^23] The Perfect Jean independently landed on the same **90% net new / 10% iterative** split after a six-month slump in which its ads all started to look identical ([[multiple-creative-partners-and-creative-culture]]).[^24] **Message anchors travel.** Ridge's "your wallet sucks" stays constant while visuals change (John Daly, tearing, burning wallets).[^19] ## Disagreements / caveats - Existing app note [[paid-social-creative-testing-framework]] tests 4 creatives per themed ad set over 3 days; Savannah's 2×2×2 is a weekly production rule, not an ad-set structure. - Savannah has said both things. In 2025 she said narrowing to one variable is impossible at paid-social speed; in her 2026 AppLovin talk she recommends isolating one variable (creator, script or format) per weekly iteration. The second is for iterating a proven winner, the first for new-concept tests.[^21] - Emily Hickey prefers milking pillar content for years via key-moment remixes ([[motivator-map-why-buy-why-believe]]) — compatible with 90/10 if remixes look visually new. ## See also - [[creative-diversity-levels]] · [[creative-metrics-hook-hold-ctr]] - [[people/dara-denney]] · [[people/savannah-sanchez]] [^1]: yt:lWpSmxuwSts @ 00:00 — Dara Denney [^2]: yt:lWpSmxuwSts @ 05:40 — Dara Denney [^3]: yt:lWpSmxuwSts @ 04:54 — Dara Denney [^4]: yt:lWpSmxuwSts @ 12:17 — Dara Denney [^5]: yt:76KH-GZZ5w4 @ 18:40 — Savannah Sanchez [^6]: yt:76KH-GZZ5w4 @ 17:23 — Savannah Sanchez [^7]: yt:lWpSmxuwSts @ 00:47 — Dara Denney [^8]: yt:lWpSmxuwSts @ 01:34 — Dara Denney [^9]: yt:lWpSmxuwSts @ 02:20 — Dara Denney [^10]: yt:lWpSmxuwSts @ 03:56 — Dara Denney [^11]: yt:lWpSmxuwSts @ 06:30 — Dara Denney [^12]: yt:lWpSmxuwSts @ 07:17 — Dara Denney [^13]: yt:lWpSmxuwSts @ 08:04 — Dara Denney [^14]: yt:lWpSmxuwSts @ 09:01 — Dara Denney [^15]: yt:lWpSmxuwSts @ 10:35 — Dara Denney [^16]: yt:4wM-i1cxtAM @ 03:52 — Dara Denney [^17]: yt:76KH-GZZ5w4 @ 05:02 — Savannah Sanchez [^18]: yt:76KH-GZZ5w4 @ 16:09 — Savannah Sanchez [^19]: yt:DqFjak694PQ @ 20:10 — Evan Lee [^20]: yt:El5pN9zYZPA @ 26:32 — Savannah Sanchez [^21]: yt:I0SPikVQdYs @ 31:28 — Savannah Sanchez [^22]: yt:aSF0GiHUZxo @ 05:15 — Savannah Sanchez [^23]: yt:qoOBIPWH_qw @ 59:09 — Barry Hott [^24]: yt:qoOBIPWH_qw @ 18:06 — Ovadia Lebaton