Been testing different ways to phrase our NPS survey and getting wildly different scores.
Same users, same timing, but changing “likely to recommend” vs “would you recommend” shifts results by 15-20 points.
Anyone else notice this or am I overthinking the wording?
Wording impacts scores for sure. We noticed similar shifts with our survey questions too.
Definitely relates to how people interpret questions
Testing wording variations is key. I found similar results with app store review prompts. The word “likely” leads to overthinking, while “would you recommend” gets a more instinctive response.
Using simpler questions is best for mobile apps since users are often distracted. Different scores can offer a clearer picture of actual feelings.
Wording changes everything with NPS.
“Likely to recommend” makes people think about probability and future scenarios. “Would you recommend” feels more immediate and concrete.
The 15-20 point swing you’re seeing is normal. Users respond differently when they’re imagining a hypothetical vs making a direct statement.
Pick one version and stick with it. Consistency matters more than perfect wording for tracking trends over time.
Ran into this exact issue with a dating app two years ago. Same massive swings in scores.
What we found was “likely” triggered more analytical thinking while “would you” got gut reactions. The gut reactions ended up predicting actual referral behavior way better.
We A/B tested the wording against actual recommendation rates over 3 months. The “would you” version had stronger correlation with users who actually shared the app.
Your 15-20 point difference isn’t noise - it’s measuring different things. Go with whichever version gives you data that matches real user actions.