A/B testing

A/B testing lets you publish two versions of the same post with different captions and see which performs better with your audience.

Creating an experiment

1. Open any draft with a caption
2. Click More → Create A/B variant
3. The app clones the draft as variant B with an AI-rewritten caption - same facts (event, time, venue) but a different hook, structure and tone
4. The original becomes variant A; both share an experiment ID

You're taken straight to the B variant to review and tweak the rewritten caption.

Running the test fairly

For meaningful results:

• Same image - both variants keep the original render so only the caption differs
• Similar timing - schedule them at comparable times (e.g. consecutive match days at the same hour), or one per location if you run multiple venues
• Don't edit mid-test - changing a caption after publishing muddies the comparison

Reading the results

Once both variants are published and the daily insights sync has run (post performance pulls once a day for posts from the last two weeks):

• GET /api/insights/experiments returns each experiment with both variants' reach, likes, comments, saves and shares
• Engagement rate = (likes + comments + saves + shares) ÷ reach
• A winner is declared automatically once both variants have insight data

What to test

Good caption variables: question vs statement openers, emoji density, short vs long, urgency ("today only") vs relaxed, with/without a call to action. Test one variable at a time.

WARNING

Insights require a working Instagram connection (Meta Graph). If the connection is down, posts publish but engagement data won't flow back.