A/B test UGC hooks and captions

What to test

• Captions: the first line decides whether people read on. Test a question against a claim against a number.
• Hooks: different first seconds of the same edit. Upload each cut as its own file.
• Formats: the same story as a video and as a slideshow.
• Change one thing at a time, or you won't know what worked.

Rotate captions across accounts

Give a post caption_variants. UGC API rotates them across the accounts (caption 1, 2, 3, 1, 2, 3…) and labels each result caption_1, caption_2… so reports can group by it.

{
  "targets": { "tags": ["fitness-creators"] },
  "spread": false,
  "posts": [{
    "format": "video",
    "media": ["https://cdn.example.com/protein-bar.mp4"],
    "caption": "I tried this for 30 days",
    "caption_variants": ["Why is nobody talking about this?", "3 reasons I switched"]
  }]
}

Label hooks yourself

For different edits, post each file with a variant label (hook-a, hook-b) and set one clip_id for both so the report treats them as one test. Then call the report with group_by=variant.

curl "https://ugcapi.dev/v1/campaigns/cmp_xyz/report?group_by=variant" -H "Authorization: Bearer ugc_live_..."

Reading the result

• Compare avg_views and median_views per variant. If they disagree, one viral post is skewing the average.
• Check engagement_rate too: a caption can win views and lose comments.
• Wait at least 48 hours and use views_at.1d to compare at the same age.
• Then send the winner to more creators: by hand, or with a repost rule that does it when a clip breaks out.