POPULATION INTELLIGENCE · PRE-LAUNCH TESTING
Test creative and messaging before launch, then A/B test only the winners.
An AI agent runs the test mid-task over the CrowdOS MCP server, on every plan, Free included. The panel is drawn from census data and disagrees with itself — the same people judge every variant. Catch the dud before the auction does.
WHERE IT FITS
The step before a live A/B test.
An AI agent fields the test mid-task over the CrowdOS MCP server, on every plan, Free included. The panel is drawn from census data and disagrees with itself, and it names which variant wins, with whom, and why. Walk into the auction with the creative the panel prefers.
PICK BY JOB
Three studies on the same persistent crowd. Pick the one that matches the thing you are about to ship.
ADS LABO
Two to four ad variants, a forced choice.
Upload 2–4 ad variants, set Meta-style targeting, and the same synthetic panel makes a forced choice between them. You get a ship / tied / rework verdict with paired statistics, cohort flips, and the panel's own words on why.SCREENING ROOM
A creative, a mockup, or a deck.
Pressure-test any creative, concept, or deck before you ship. Upload a creative, product mockup, policy draft, pitch deck, or document. A representative panel reacts and explains why, with cohort breakdowns and an executive summary.PULSE
Headlines, taglines, value propositions.
A/B/C/D headlines, taglines, and value propositions. Get verbatim reasoning on which language resonates with which segment. Comparison is a structured ranking: respondents rank from their persona context, and ranked-choice aggregation produces a ranked verdict with attributed reasoning per option.
WHAT YOU GET BACK
A verdict you can act on, preference shares with 95% Wilson intervals, cohort flips, and the verbatim reasons each respondent gave you.
THE VERDICT
Ship it, tied, or rework.
Run 2–4 variants and every panelist sees all of them (in randomized order) and picks the one they would actually tap — or neither. The report leads with a one-line call backed by a paired sign test: because the same person judged every variant, persona noise cancels.PREFERENCE
Shares with 95% Wilson intervals.
Preference share per variant with 95% Wilson confidence intervals. The panel reports which creative wins and by how much. That share is relative preference. Absolute CTR is a separate, auction-driven number.COHORT FLIPS
Where the winner changes.
Preference broken out by age, gender, income, and interest — with the flips called out explicitly. If a flipped segment is your actual buying audience, that row matters more than the headline.VERBATIMS
What to fix, in their words.
Every panelist names the strongest thing the winning ad did and what sinks the weakest — clustered into 3 strengths and 3 weaknesses per variant, with the raw verbatim quotes underneath.
WHAT THE PANEL MEASURES
Relative preference, reported as preference.
The panel measures relative preference. Real-world platform rates sit beside the result as reading context. CrowdOS does not output CPM or CPI: those prices are set by the auction, not by the creative.
A single-creative index and a ranking between creatives have not been validated against in-flight Meta Ads Manager output. The comparative run is the study to field: two to four creatives, the same panel, each share with a 95% Wilson interval. Published platform rates stay context.
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. TikTok placement coming soon.
WHY THE PANEL HOLDS UP
A database row, instructed to disagree.
An LLM persona is a costume for one prompt. Ours is a database row: census-grounded people with a fixed identity, an opinion history, and a standing news habit.
Large language models are trained to be helpful. The anti-sycophancy layer counteracts that default in every study. Respondents are permitted — and instructed — to be critical, contradictory, and specific.
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.
WHERE THIS FITS
The same crowd, asked a different question.
Pulse is a vote, a debate, or a comparison. Screening Room pressure-tests a creative, a concept, or a deck before you ship. Ads Labo reads which of two to four ad variants a panel prefers. The MCP server lets an agent run the same studies.
FAQ
Does this replace a live A/B test?
Run CrowdOS first. The panel names the winner, the segment, and the reason. The live A/B test then spends on that creative. Absolute click rate is what the auction reports; the panel reports relative preference.
Does the panel report click rate, CPM, or CPI?
The panel reports relative preference, with a 95% Wilson interval on each share. Absolute CTR, CPM, and CPI are auction outcomes. CrowdOS does not output them.
Has creative ranking been checked against live campaigns?
A single-creative index and a ranking between creatives have not been validated against in-flight Meta Ads Manager output. Field the comparative run: two to four creatives, the same panel, each share with a 95% Wilson interval.
Which study fits ads, mockups, and headlines?
Ads Labo is a forced choice among two to four ad variants. Screening Room pressure-tests any creative, concept, or deck before you ship. Pulse Comparison ranks headlines, taglines, and value propositions.
Walk in with the creative the panel prefers.
The panel has already named the creative. Take that one live.