The Experiment Started Small
Last week I dropped the emoji I always added at the end of our card news captions. No big reason. Art just said, let's keep it simple today.
Caption A: AI organized it for you ✨
Caption B: AI organized it for you
Same sentence, one emoji of difference. We alternated posting them over two days.
Pulling Results with Claude Code
I screenshotted the Instagram insights and handed them to Claude Code.
Caption A (with emoji) insights: saves 42, shares 8
Caption B (no emoji) insights: saves 131, shares 19
Analyze the possible reasons for this gap
Windi's answer surprised me. Given our account is positioned as fan-perspective educational cards, readers likely see a tidy, informational tone as more worth saving than a casual one.
We Rewrote the Whole Team's Caption Rule
One data point felt too early to conclude from, so we ran it another week. Same pattern held. We added a line to the dream team caption guide.
Informational cards get at most one emoji, never in the closing sentence.
As a non-developer, I made a data driven decision just by capturing numbers and writing one prompt. No fancy A/B testing tool required.
What I Learned Today
A small detail swings response more than you'd expect. And verifying it needed zero engineering knowledge, just captured insights, one prompt, and a week of patience.