Your creative library is an asset you've already paid for.
You just can't sell it twice. Yet.
Or with zero setup: Signal — the free teardown tool, built on Mixpeek's Creative DNA.
You've already paid for your next winning ad. It's in your archive.
We build a live intelligence dossier on your market from public ad libraries — your in-market ads, your three closest competitors', and what's trending on TikTok (from TikTok's own Creative Center). It's a page you can search and interrogate, not a deck.
One brand's 77 in-market ads turned out to be a single hero video running as 63 separate placements — 92% of their creative volume, one idea in variant clothing. Nobody inside that brand had that number.
A live dossier
A real brand's ad archive, decomposed into scenes, recurring faces, and formats — searchable, with grounded Q&A. Nothing gated.
Open the Hims dossierThe open recipe
The exact pipeline behind every dossier, published as a versioned manifest your technical team can read before anyone gets on a call.
Read the manifestToday's creative search is vibes.
Strategists carry winning patterns in their heads. Tools can find ads that look like a reference, but nothing more. Performance data sits in a separate warehouse. Every new client brief starts from zero, even when the answer already exists somewhere in the library.
Your team has produced thousands of ads. Some crushed. Some didn't. The ones that crushed have hooks, pacing, compositions, and audience-fit signals that are worth money. But today that signal is locked inside files nobody can search.
An ad isn't one thing.
It's at least four.
What it looks like
Visual composition, color, framing
How it moves
Pacing, motion energy, edit rhythm
What it says
Hook, script, CTA, narrative arc
How it performs
ROAS shape, audience fit, decay curve
Each is its own kind of similarity. None of them alone is enough. The tools you have today collapse them into one — usually the first — and lose the signal.
Performance isn't a filter. It's a fourth axis of similarity.
"Hooks like our Q3 winner, cold fintech audience, 3x+ ROAS"
Want to understand the clustering and taxonomy discovery behind this?
Learn how we index your ad library without manual tagging.
One question. Four axes. Every ad you've ever run.
Mixpeek decomposes every ad into all four axes, indexes each independently, and joins them at query time. A single retriever can ask:
"Find ads with hooks like this one, pacing like that one, that performed like the top of last quarter, in cold fintech."
Nothing else can ask that question, because nothing else holds all four dimensions in the same place.
Revenue from creative you've already paid for, against demand you already have.
Three ways your library becomes a revenue line.
Brief-to-asset matching
New brief lands. Query the library across all four axes. Your team starts from proven components, not a blank page.
Cross-portfolio licensing
A hook that crushed for one advertiser works for another in an adjacent vertical. Mixpeek surfaces those connections automatically.
Performance-conditioned pitches
Walk into a pitch with demonstrable inventory advantage and predicted ROAS from your existing library.
Cold fintech audience, 25-34M. Need a fast-paced hook that opens with a question. Budget for 15s vertical. Target 3x+ ROAS week one.
Your library is the moat.
Mixpeek makes it one.
Every query teaches the retriever. Every new ad enriches the library. Every performance update sharpens the predictions. Six months from now, the model you have is one only your team has — and it's structurally impossible for competitors without a comparable library to catch up.
Built for the people who make ads work.
Creative Strategists
You already think in hooks, pacing, and composition. This is the first tool that respects that. Stop rebuilding briefs from scratch when the winning pattern already exists in your library.
Agency Leadership
Every dollar of revenue requires a proportional dollar of creative production. Mixpeek breaks that ratio. Expand accounts without expanding headcount.
Data & RevOps
Controlled A/B tests. Attributable lift. Methodology you can audit. We don't claim what we can't measure, and we don't charge for what doesn't work.
Assemble from what wins.
From footage you already own.
We index your raw archive, your competitors, the market, and your own best-performing content — then reverse-engineer, from footage you already own, which creative components recur in winners. Instead of producing net-new every cycle, you assemble creative that is structurally like what wins, from footage you've already paid for.
Competitor data comes from the public Meta Ad Library — we cluster their longest-running, highest-signal creative, not claimed ROAS. The performance axis — components empirically tied to your results — is something we build with you in a pilot, on your numbers, not an off-the-shelf claim.
Signal — the free teardown tool, built on Mixpeek's Creative DNA — runs the competitor half on any brand's public creative, nothing gated.
Decomposed into primitives
Finished ads, raw footage, and competitor creative broken into hooks, pacing, composition, palettes — queryable, model-ready components instead of opaque files.
The components that recur in winners
Across your own best-performing creative: which hooks, structures, and openings keep showing up — so briefs start from proven parts, not a blank page.
Market trends, matched to your archive
What the market is converging on right now, matched against footage you already own — so you know what you can re-cut today versus what genuinely needs a shoot.
See your brand's creative DNA — free.
Signal — the free teardown tool, built on Mixpeek's Creative DNA — indexes your brand's public Meta creative and publishes a live page: the themes carrying your spend, recurring talent, searchable scenes, and grounded Q&A over every ad. No signup, nothing gated.
Free layer: themes · recurring talent · scene search · Q&A — computed from public Ad Library creative only.
The free page reads your public ads. The real analysis reads your archive: give us the competitors and market trends you care about, plus your own ad archive — we mine all of it to find which creative elements you can repurpose before you brief a single new shoot.
Request it from the intake form on your brand's Signal page. Every request is reviewed by a strategist before any analysis runs — this is a scoped engagement, not a button.
Or start with a working sessionA working session with your data team.
Your ads. Your models. Live.
Not a demo — a 45-minute build session. We decompose a slice of your own Meta ad library on screen, cluster it, name the themes, and leave your team a running notebook and a seeded namespace they keep. For data science, ML, and creative-analytics teams evaluating whether to build this or buy it.
Your corpus, decomposed live
We ingest a sample of your actual ad library and run the full pipeline — embeddings, clustering, LLM-named themes — in the room, on your creative.
A notebook they keep
Your team leaves with a runnable Jupyter notebook: raw corpus → the four DNA axes → the creative-vs-performance scatter, against a namespace that stays live.
The build-vs-buy answer
Honest scoping of what it takes to run this in-house vs on the API — where the multi-vector, multimodal retrieval is the hard part you'd rather not own.
A scoped pilot on your data
We leave with a concrete next step: a bounded POC on your library with a measurable success metric — not a slide, a plan.
Pay for the lift,
not the platform.
We propose a rev share on measurable, attributable ROAS lift over a controlled baseline. If the platform doesn't generate lift, you don't pay.
We make money when you make more money. No upfront commitment. No paid pilots.
See it on your library
20 minutes. Your ads. Real decomposition. No pitch deck.
Or email [email protected]