AD CREATIVE TESTING · PRE-FLIGHT
Pre-flight your ad creative, before the auction sees it.
Your AI agent pre-flights the whole ad mid-task over the CrowdOS MCP server, on every plan, Free included: headlines, hooks and calls to action, then the image and video cuts through Ads Labo and Screening Room. A census-grounded panel that disagrees with itself — not one model's opinion — judges every variant and returns ship, tied or rework.
TWO HALVES OF ONE AD
The words and the visuals, from one agent.
For the words, the agent calls compare_options with two to ten text options — the headline, the primary text, the hook line, the call to action — and gets a Borda leaderboard and a winner. preview_cost prices the run first, free to call.
For the visuals, Ads Labo takes 2–4 image or video variants (images up to 10 MB, videos up to 100 MB), Meta-style targeting and a placement: Meta Feed, Reels or Stories. Every panelist sees every variant in randomized order and picks the one they would tap, or neither. MCP carries the same functions as the web app on your plan, so the agent sends the cuts itself.
WHAT ADS LABO RETURNS
A verdict, honest error bars, and the reasons.
01
Ship it, tied, or rework.
A one-line call backed by a paired sign test. The same person judged every variant, so persona noise cancels.02
Preference with 95% Wilson intervals.
Preference share per variant with 95% Wilson confidence intervals, plus a 0–100 Pull Index from each panelist's 0–10 tap-likelihood scores. Relative preference, reported as preference.03
Cohort flips.
Preference broken out by age, gender, income and interest, with the flips called out: the segments where the overall winner loses.04
What to fix, in their words.
Three strengths and three weaknesses per variant, clustered from verbatim reasons. When the verdict is rework, the why-neither answers are the brief.05
Panel sizes.
Quick Test: 100 respondents, 75 credits. Standard Lab: 300 respondents, 225 credits. Deep Test: 500 respondents, 375 credits. Up to four variants on the same panel cost the same as one. Single-creative mode gives a Pull Index and cohorts. Panel sizes follow your plan, the same in the web app and over MCP.
THE OPINION BENCHMARK
The published opinion benchmark.
On the Pew opinion benchmark, the crowd's opinion distribution matches published polling at 93.3% parity, with a mean absolute error of 3.36 percentage points and a Pearson r of 0.984. That figure measures opinion. Creative ranking against a live campaign is a separate measurement.
MARKETS AND PLACEMENTS
The same variants, in the placement they will run in.
Test the same variants across 24 market-level audiences. Meta Feed, Reels, and Stories — the panel judges your creative in placement context.
WHERE THIS FITS
The same crowd, asked a different question.
Pre-launch testing covers messaging, mockups and decks as well as ads. This page is the ad itself — copy, image and video — pre-flighted by your agent over MCP or in the web app.
FAQ
How do I test ad creative before launch?
Let your AI agent pre-flight it mid-task over the CrowdOS MCP server, on every plan, Free included. compare_options ranks the copy. Ads Labo takes the image and video cuts: two to four variants, one census-grounded panel, a ship / tied / rework verdict with 95% Wilson intervals.
Can my agent test image or video ads over MCP?
Yes. MCP carries the same functions as the web app on your plan: whatever your plan runs in Ads Labo and Screening Room — image and video variants included — your agent runs over MCP, alongside the headlines, primary text, hooks and calls to action it ranks with compare_options. The REST API is the one surface reserved for Enterprise.
Does Ads Labo predict CTR, CPM or CPI?
No. The panel measures relative preference: which creative wins, with whom, and by how much. CPM and CPI are set by the auction. Walk into the auction with the creative the panel picked, then let a small live budget confirm the absolute numbers.
How is ad creative testing different from pre-launch testing?
Pre-launch testing is the whole step before a live A/B test: messaging, mockups, decks and ads. Ad creative testing is the ad itself — the copy, image and video variants your agent pre-flights over MCP, judged in placement context.
How accurate is the panel?
On the Pew opinion benchmark, the crowd's opinion distribution matches published polling at 93.3% parity. That figure measures opinion. Creative ranking against a live campaign is a separate measurement, so the Ads Labo verdict is a directional pre-flight call.
Catch the dud before the auction does.
Rank the copy and pre-flight the cuts — from your agent or in Ads Labo.