Radar Room Scoring
This document explains how feedback submitted to the Radar Room is evaluated and how the prize pool is distributed. Our goal is to ensure you pay for quality feedback, not spam or low-effort submissions.
Why We Score Feedback
Beta testing bounties attract both genuine testers and bad actors. Without scoring, you could end up paying for:
- AI-generated feedback — Users paste your bounty description into AI tools and submit the output
- Generic praise — “Great UI, love the design” from someone who never opened your product
- Volume spam — Dozens of low-effort messages to game the prize pool
- Duplicate ideas — Paraphrased copies of other users’ feedback
Our scoring system filters these out and rewards testers who actually use your product and give you actionable insights.
How Feedback Is Processed
Every submission goes through several checks before it can earn a share of the prize pool.
1. Pre-Filters
We apply basic rules before any AI processing:
- Minimum length — Very short messages (“nice”, “looks good”) are rejected
- Rate limits — Users can’t flood submissions
- Cooldown — Short wait between submissions to prevent bot-like behavior
- Duplicate check — Exact or near-exact copies of prior feedback are rejected
2. AI & Spam Detection
We detect feedback written by AI tools and reject it.
We also filter spam, including:
- Promotional content
- Shilling
- External links
- Referral links
- Self-promotion
3. Classification
Each feedback is assigned one category. Only three categories qualify for rewards.
Eligible
| Qualifies | Category | Meaning |
|---|---|---|
| ✅ | Relevant and actionable | Detailed feedback with suggestions or bug reports |
| ✅ | Feedback | Human feedback with some specifics |
| ✅ | Question | Genuine question about your product |
Rejected
| Rejected | Category | Meaning |
|---|---|---|
| ❌ | AI generated | Written by AI |
| ❌ | Spam | Promotional or ads |
| ❌ | Irrelevant | Not about your project |
| ❌ | Copy others feedback | Duplicate idea |
| ❌ | Low effort | Too vague to be useful |
| ❌ | Normal message | “thanks”, emoji, etc |
4. Quality Scoring
Qualifying feedback is scored on four dimensions (0–10 each):
| Dimension | What We Look For |
|---|---|
| Relevance | References your specific features or flows |
| Actionability | Clear suggestions or reproducible bugs |
| Specificity | Concrete details, UI elements, steps |
| Evidence | Screenshots, recordings, logs (no media = 0) |
Feedback that could have been written without opening your product receives low or zero scores.
5. Quality Gate
To earn rewards, feedback must:
- Be in an eligible category
- Meet minimum thresholds for relevance, specificity, and actionability
- Pass an overall quality floor
If any check fails → Submission earns $0
Examples of Good vs Bad Feedback
✅ Good Feedback (Qualifies)
Actionable Bug Report
“On the checkout screen, when I enter a quantity and switch currencies, the input resets to 0. Happens every time on Chrome (Mac). Screen recording attached.”
Why it qualifies:
- Mentions exact feature and flow
- Reproducible steps
- Environment specified
- Includes evidence
Specific Improvement Suggestion
“The onboarding tooltip covers the ‘Next’ button on mobile (iPhone 13 Safari). Maybe reposition it above or delay showing it until after the first tap.”
Why it qualifies:
- References real UI interaction
- Suggests a fix
- Shows actual product use
Insightful Product Feedback
“It wasn’t clear that projects auto-save. I refreshed and thought my data was lost. Consider adding a ‘Saved’ indicator or confirmation message.”
Why it qualifies:
- Identifies usability confusion
- Explains expectation vs reality
- Provides solution idea
❌ Bad Feedback (Rejected)
Generic Praise
“Great app, love the design.”
Why rejected:
- No details
- Could be written without opening product
Low Effort
“Doesn’t work.”
Why rejected:
- No context
- No steps
- Not actionable
AI-Style / Template Feedback
“Your platform demonstrates a highly intuitive interface with strong usability principles and seamless navigation.”
Why rejected:
- Vague language
- No real interaction details
- Detectable AI pattern
Spam
“Check out my project too: mylink.xyz”
Why rejected:
- Self-promotion
- Not feedback
Duplicate Idea
“Add dark mode.”
(When multiple users already submitted the same idea earlier)
Why rejected:
- Not original
- Adds no new value
What You Can Expect
- Higher-quality feedback — Testers are incentivized to be specific and provide proof
- Less noise — Spam and AI submissions are filtered out
- Fair rewards — Distribution favors users who add real value