
Hey everyone, I'm David. I'm building Concipe, a tool that turns user feedback (reviews, support tickets, interviews) into evidence-backed product decisions.
To show what it does, I ran 8,000 Duolingo Google Play reviews through it — split into before and after they switched "hearts" to "energy" in July 2025.
What Concipe found:
Before the switch (Jan–Jun 2025):
→ Score: 5.3/10
→ Severity: Painful (2/3)
→ Signal: monetization friction building
→ Users kept saying one thing: "let us practice to earn hearts back"
After the switch (Jul–Dec 2025):
→ Score: 7.2/10
→ Severity: Blocking — churn risk (3/3)
→ Signal: no user validation before shipping
The score jumped from 5.3 to 7.2. The warning escalated from "painful" to "churn risk."
What actually happened:
→ 75% of Q4 2025 reviews mention the energy system
→ Average rating on those: 1.6 stars
→ Stock dropped 46% in 2025
→ CEO admitted the backlash hit growth numbers
→ Users with 1,000+ day streaks started quitting publicly
The quick math:
12.2M paid subscribers. $60–84/year depending on plan. If just 1% churn from this backlash, that's 122,000 users, roughly $9–19M/year walking out the door.
The takeaway:
The reviews had the answer months before the decision. Users didn't want free access. They wanted the hearts system to be more forgiving. Concipe would have flagged this as a growing risk and recommended testing before a global rollout.
I built Concipe for solo founders and PMs using AI coding agents (Cursor, Claude Code, Windsurf). You paste your feedback, it extracts opportunities, scores them by evidence and business impact, and generates specs you can hand to your coding agent.
If you're running a product and drowning in user feedbacks, DM me your source - I'll run it through Concipe for free and show you what it surfaces.
Full decision reports:
→ Before: https://www.concipe.com/shared/decision/ab52a48e-511b-492b-941c-27c8f098cf48
→ After: https://www.concipe.com/shared/decision/cbc42bab-0118-4535-b502-859ae7613943