--- title: When to iterate on a winning Meta ad vs make a net-new concept slug: meta-iteration-vs-net-new-creative summary: Diversity (net-new concepts) explores new personas; iteration exploits a proven persona × angle combination — how the 2025–2026 practitioners split effort, what counts as each, and the cases where a small tweak is enough. domain: marketing tags: [meta-ads, creative-strategy, creative-testing, iteration, andromeda] related: [meta-creative-similarity-entity-ids, meta-persona-angle-creative-dimensions, meta-creative-testing-by-budget, meta-ad-diagnostic-cheatsheet, meta-landing-page-congruency] sources: [yt:ZqM6jPUAaeQ, yt:aNCUFwQw524, yt:aCkulCPAYzw, yt:7knQyPYLmfo, yt:cnWpC1g5qek] updated: 2026-10-24 --- # Iterate vs net-new ## TL;DR - **Net-new (diversity) = explore**: a different persona, angle, emotional territory, visual style or context. **Iteration = exploit**: new hook, talent, body, audio, CTA or a mash-up of winners on a proven combination. You need both.[^1][^2] - Under Andromeda, every iteration must still be **visually different** from the original, or it gets clustered with it.[^3] - Triage rule (Josh Brampton): test e.g. 10 angles on a persona → kill the ~5 that fail, tweak the ~2 that are OK, make several *visually different* versions of the ~3 that work.[^4] - Zach Stuck (Mars Men, Hollow) now spends little time on iterations — net-new concepts produce more winners; after a winner he tests a **new landing page** rather than iterating the ad.[^5] - Small tweak is right when engagement is strong and one metric lags (e.g. great hold rate, weak CTR → change only the end/CTA).[^6] ## Details **Definitions.** Manchi (Ready Set): diversity means genuinely distinct assets for different personas, messaging angles, emotional territories, visual styles and contexts; iteration means incremental changes to an existing ad based on its signals.[^7][^1] When a winner emerges, exploit it into *other formats* — UGC video, animation, partnership ads — while keeping the dimensions that won.[^8] **Explore → Exploit → Amplify (Ready Set).** Every concept sits in one of three phases. The phases run in sequence for each concept and in parallel across the account.[^14] - *Explore*: validate a persona (CTP), an angle and the trigger that makes people act. - *Exploit*: keep the validated CTP × angle and vary the testing dimensions: talent, visual hooks, formats, styles.[^15] - *Amplify*: take a fully exploited angle to a new persona, which restarts Explore. Manchi says this is where compounding happens, and the angle can also move to other channels.[^16] When a fresh ad shows promise after a few days, Ready Set's **fast iteration engine** (run in Motion) ships iterations within **1–3 days**, without reshooting footage, to ride the momentum.[^17] **Fatigue response.** When a former winner's CPM and CPA rise, the fix is iterate/refresh/rotate from what already worked.[^9] **How big the swings should be depends on stage.** Dara Denney: at the growth stage ($30–50K/month) variations inside a test should take big messaging swings and test different hooks and video lengths; at $1M+/month brands already know what works and test slighter variations — but also bigger production swings like celebrities or a high-production version of a lo-fi founder ad.[^10][^11] **What net-new looks like** (Zach): a different person, a completely different style and format — a "yapper" talking-head vs a claymation AI video — not a native static vs a designed static.[^5] **Persona × angle grid.** Brampton: after Andromeda many teams declared iterations dead, but the best brands run hard with combinations that work — if "eliminate guesswork" is the winning angle for one persona, make many visually diverse versions of it.[^12][^3] ## Disagreements / caveats - Zach (net-new first) vs Brampton/Manchi (balanced explore/exploit) is partly about budget: Mars Men runs 5 agencies plus internal pods, so it can afford constant net-new; small teams may get more from iterating proven combos. - Smart Marketer's 2023 tactic — refresh by swapping only text overlay — squeezed ~$100K extra spend then, but predates Andromeda's similarity clustering.[^13] ## See also - [[meta-creative-similarity-entity-ids]] - [[meta-persona-angle-creative-dimensions]] - [[meta-ad-diagnostic-cheatsheet]] — which metric tells you what to tweak - People: [[people/ignacio-manchi]], [[people/josh-brampton]], [[people/zach-stuck]] [^1]: yt:ZqM6jPUAaeQ @ 15:14 — Ignacio Manchi [^2]: yt:ZqM6jPUAaeQ @ 16:47 — Ignacio Manchi [^3]: yt:aNCUFwQw524 @ 17:14 — Josh Brampton [^4]: yt:aNCUFwQw524 @ 39:18 — Josh Brampton [^5]: yt:aCkulCPAYzw @ 51:00 — Zach Stuck [^6]: yt:aNCUFwQw524 @ 28:14 — Josh Brampton [^7]: yt:ZqM6jPUAaeQ @ 13:36 — Ignacio Manchi [^8]: yt:ZqM6jPUAaeQ @ 18:17 — Ignacio Manchi [^9]: yt:ZqM6jPUAaeQ @ 20:06 — Ignacio Manchi [^10]: yt:7knQyPYLmfo @ 09:16 — Dara Denney [^11]: yt:7knQyPYLmfo @ 13:53 — Dara Denney [^12]: yt:aNCUFwQw524 @ 14:07 — Josh Brampton [^13]: yt:cnWpC1g5qek @ 39:24 — Molly Pittman [^14]: yt:ZqM6jPUAaeQ @ 38:02 — Ignacio Manchi [^15]: yt:ZqM6jPUAaeQ @ 39:36 — Ignacio Manchi [^16]: yt:ZqM6jPUAaeQ @ 40:40 — Ignacio Manchi [^17]: yt:ZqM6jPUAaeQ @ 45:58 — Ignacio Manchi