Stage 4: Product Ranking
Now we have rich, intelligent data. Time to combine it into one number that captures genuine product love.
The Love Score
The Love Score is our custom metric that identifies products people truly love — not just products with high ratings.
The Philosophy
- High ratings from paid reviewers don't count as much
- Products loved by diverse skin types are more universally good
- Engagement (photos, helpful votes) signals genuine enthusiasm
- Recent activity matters more than old reviews
- Penalize red flags, boost green flags
Try the Calculator
See how the Love Score works with an example product:
Love Score Calculator
What unpaid reviewers actually think
Helpful votes, photos, detailed reviews
Percentage of genuine reviews
Works across skin types/tones
Recent activity and momentum
Final Love Score
69
This product is genuinely loved!
The Five Components
The Love Score is a weighted combination of five factors:
1. Organic Quality (35%)
What it measures: What unpaid reviewers actually think.
How it's calculated:
- 60% weight on average rating from organic reviewers
- 40% weight on recommendation rate from organic reviewers
Why this weight: This is the most important signal. If people who didn't get free product still love it, that's genuine.
2. Engagement Quality (25%)
What it measures: Community validation.
How it's calculated:
- 50% weight on helpfulness ratio (helpful / total votes)
- 30% weight on percentage of reviews with photos
- 20% weight on percentage of substantive reviews (50+ words)
Why it matters: Reviews with lots of helpful votes are crowd-validated. Photo reviews show real usage. Long reviews indicate genuine engagement.
3. Authenticity (15%)
What it measures: What percentage of reviews are genuine.
How it's calculated: Simply the ratio of organic reviews to total reviews.
Why it matters: A product with 90% organic reviews is more trustworthy than one with only 60% organic.
If we can't determine authenticity, we assume 13% organic (the average across all products). Yes, incentivized reviews are that common.
4. Diversity (15%)
What it measures: Does the product work for different people?
How it's calculated:
- 50% weight on skin type diversity
- 50% weight on skin tone diversity
Why it matters: A moisturizer reviewed only by oily-skin people might not work for dry skin. Products loved across all types are more universally good.
5. Trend (10%)
What it measures: Current momentum.
How it's calculated: Reviews in last 180 days / total reviews
Why it matters: A product popular 5 years ago but with no recent reviews might be discontinued or outdated.
The Confidence Multiplier
Raw scores don't mean much without confidence. A product with 3 reviews might have a perfect score, but we can't trust it.
The formula:
confidence = log(organic_reviews + 1) / log(150)What this means:
- At 0 organic reviews: ~0% confidence
- At 10 organic reviews: ~46% confidence
- At 50 organic reviews: ~78% confidence
- At 150+ organic reviews: 100% confidence (saturated)
The final score is multiplied by confidence, so low-review products get penalized even if their metrics look good.
The Adjustments
After the weighted score and confidence, we apply adjustments for red flags and green flags:
Penalties (Red Flags)
| Adjustment | Max Impact | Trigger |
|---|---|---|
| Inflation Penalty | -15% | Paid reviewers rate much higher than organic |
| Staff Penalty | -14% | More than 30% reviews from employees |
| Polarization Penalty | -10% | High variance + many negative reviews |
| ML Quality Penalty | -12% | Average review quality below 0.5 |
Boosts (Green Flags)
| Adjustment | Max Impact | Trigger |
|---|---|---|
| Power User Boost | +8% | Experienced reviewers (21+ reviews) love it |
| Rating Trend | ±10% | Recent ratings better/worse than historical |
Every product shows which adjustments were triggered and why. No black box — you can see exactly how the score was computed.
The Final Formula
raw_score = (0.35 × organic_quality)
+ (0.25 × engagement_quality)
+ (0.15 × authenticity)
+ (0.15 × diversity)
+ (0.10 × trend)
weighted_score = raw_score × confidence
final_score = weighted_score
+ inflation_penalty (up to -0.15)
+ staff_penalty (up to -0.14)
+ polarization_penalty (-0.10)
+ rating_trend (±0.10)
+ power_user_boost (up to +0.08)
+ ml_quality_penalty (up to -0.12)
(capped between 0 and 1)Product Tiering
For products that need detail scraping (price, ingredients), we use a simpler point-based tiering:
| Factor | Points |
|---|---|
| 1,000+ reviews | 40 pts |
| 500-999 reviews | 35 pts |
| 100-499 reviews | 25 pts |
| 4.5+ star average | 30 pts |
| 4.0-4.49 stars | 20 pts |
| 100+ substantive reviews | 20 pts |
| 500+ unique reviewers | 10 pts |
Tiers:
- High Priority (70+ pts): Scrape ASAP
- Medium Priority (40-69 pts): Worth scraping eventually
- Low Priority (15-39 pts): Skip for now
What's Next?
The Love Score is computed. Now see what we output.
Next: Outputs → — The final deliverables.