Paid social creative
Ground truth is click-through rate from a split test on live paid placements.
min. 50,000 impressions per arm, p < 0.05
Think you can beat AI? Take our test.
Humark challenges are built from real A/B tests with significant results, so there is a right answer. The same question goes to the leading models, and your profile shows which ones you beat and where you do not.
Take oneOne challenge, about thirty seconds
Free to play. Verification takes three rounds.
Paid social creative · DTC skincaredemo data
Every challenge has a receipt.
Video thumbnails
Ground truth is click-through from a thumbnail test on the same video.
min. 20,000 impressions per variant, p < 0.05
Checkout drop-off
Ground truth is the step with the highest measured abandonment in funnel data.
min. 10,000 sessions, p < 0.05
If a test did not reach significance, it does not become a challenge. There is no panel of experts deciding who was right.
74% across 42. The models got 68%.
Scorecarddemo data
You
74%
31 / 42 correct
Model average
68%
same 42 challenges
Paid social creative
81%
16 challenges · ahead of all three models
Email subject lines
52%
9 challenges · behind all three models
Model comparisondemo data
Paid social creative
you
81%
model A
75%
model B
62%
model C
69%
Email subject lines
you
52%
model A
67%
model B
56%
model C
78%
Model identities are published on each challenge's result page.
You will be worse than the models at something. That is useful information and it stays on your profile.
Ten right is luck. Thirty across three rounds is not.
Verification track · round 2 of 3demo data
Round 1
75%
complete
Round 2
74% running
complete
Round 3
unseen set
pending
Verification requires holding above 70% across three rounds of unseen challenges.
2%
top of the pool
~70%
held status next year
The Good Judgment Project found its superforecasters by skimming the top 2% of a large public pool, and roughly 70% of them held that status the following year. Humark works the same way. Fresh challenges, unseen, and verification only for people who hold their edge.
Source: the Good Judgment Project. External research, not Humark data.
Companies search the verified pool.
Verified participants opt in. Recruiters search by skill and by model comparison, so a brief reads like top 1% at predicting paid social winners, beats every model tested on DTC creative. Participants are paid for their time on commissioned challenges and we take a fee from the company, never from you.
DTC skincare · paid socialA 1.27% · B 1.79%
184,000 impressions. p < 0.01. Demo data.
Take one and see.
Thirty seconds, one real test, and three models answering the same question.