How does Jev classify a UGC video?

Mixpeek turns each field of a response schema into a typed question for Jev. A field with a fixed list of values becomes a pick-one question, so the format field asks Jev to choose one of nine formats from the video's transcript and scene description. Jev returns the chosen label with its probability, and Mixpeek stores that probability next to the answer.

Which UGC format got the most views?

Across 92 tumbler-wars videos, the 11 how-to and hack videos had a median of 44,922 views, 6.5x the campaign median of 6,927. The 27 talking head reviews, the most common creator format, had a median of 3,990, about 0.58x. Skits had the highest median, but with only two videos the sample is too small to call.

I Had AI Label 92 UGC Videos. How-To Won by 6.5x

Name: I Had AI Label 92 UGC Videos. How-To Won by 6.5x
Uploaded: 2026-10-10T23:32:00Z
Duration: 79 s

Multimodal Search in Action

Ethan Steininger

October 10, 2026

Summary

Jev, a decision model from TypeSafe, labeled 92 UGC videos from the Stanley and Owala tumbler wars. It tagged each video with one of nine formats and each spoken sentence with its job, like a hook or a call to action, and it attaches a confidence score to every label so a person can check the unsure ones. With all 92 labeled, how-to videos pulled 6.5x the campaign's median views, and talking head reviews, the most common creator format in the set, landed at about half. Labeling all 92 cost 67 cents. Explore: ugc.mxp.co

jevugcclassificationvideo-analysiscreator-marketingtypesafe

I Had AI Label 92 UGC Videos. How-To Won by 6.5x

Summary

About this video

Frequently Asked Questions