← ClaudeAtlas

creative-testing-researchlisted

Plans a creative testing cycle for paid social and researches competitor ads using only public ad libraries (Meta Ad Library, TikTok Creative Center, Google Ads Transparency Center). Use when the user asks for "creative testing plan", "competitor ad research", "what ads are competitors running", "creative briefs for Meta or TikTok", "hook testing", "ad library analysis", or "creative angles for a subscription app".
ua5er/user-acquisition-skills · ★ 0 · Testing & QA · score 75
Install: claude install-skill ua5er/user-acquisition-skills
# Creative testing and competitor ad research Two jobs in one loop: learn what the market already says (public data), then test your own angles in a way that produces a decision. ## When to use - Creative output is steady but winners are rare. - A new market, offer or competitor appeared. - You need briefs for a creative team that tie to hypotheses. ## Inputs 1. Product, offer, price points, audience, markets. 2. Competitor list (names, app store links, domains) supplied by the user. 3. Your own creative results: spend, primary metric per creative, hook and hold metrics. 4. Production capacity per week. ## Method ### Step 1. Public ad research Use only public sources a person can open without logging in to someone's account: Meta Ad Library, TikTok Creative Center and the ad library, Google Ads Transparency Center, public app store pages. Do not scrape behind logins, do not use leaked data, and respect each platform's terms. For each competitor record: how many ads are active, formats, how long an ad has run (long runtime is a weak signal of profitability, not proof), hooks used in the first seconds, offers, landing page type, claims. ### Step 2. Cluster into angles Group ads into angles (problem, outcome, social proof, demo, curiosity, comparison, trend format). Count presence per angle, then mark gaps: angles nobody uses, and angles everybody uses. A crowded angle needs a better execution, an empty angle needs a test. ### Step 3. Write hypotheses One line each: