Why a public formula?
Most "best of" review sites hide their scoring. They publish a list with stars and an "editor's choice" badge but never show you the math. That's because - usually - there is no math, just vibes.
We took the opposite approach. The CurioRank is a deterministic, defensible heuristic derived only from product fields you can see on the page: the rating, the volume of reviews, the price tier, the documented pros and cons, the specs, and where available a derived sentiment score. Nothing fabricated. Same inputs, same score, every time.
The five dimensions of every CurioRank
Every product gets scored on between four and six dimensions, where the specific dimensions depend on the category. Two dimensions are shared across all categories:
- Owner Satisfaction - blends Amazon star rating, review count (logarithmic confidence weighting), and sentiment score where available. Weighted at 25% of the composite.
- Value - a function of price tier (budget / mid-range / premium) modulated by the product's rating, calibrated so a 4.5-star $20 item can outscore a 4.7-star $200 item on this dimension specifically. Weighted at 20%.
The remaining ~55% of the score is split evenly across category-specific dimensions, each driven by keyword heuristics that scan the product's pros, cons, verdict, and specs for positive and negative signals.
Category-specific rubrics
A 1,000-piece jigsaw and a UCS Millennium Falcon should never be scored on the same axes. Neither should a kids’ STEM kit and a hobby-heavy strategy game. So they aren’t. Here’s the per-category breakdown:
Adult LEGO Collectibles
- Display Value - does this earn shelf space for a decade?
- Build Experience - varied techniques, satisfying pacing, novel parts
- Piece Value - cost per piece, rare parts, minifigs included
- Theme Execution - accuracy, iconic-ness, level of detail
Board Games (Family)
- Fun Factor - does the table laugh, lean in, ask for one more round
- Replay Value - different every game, expansion ecosystem
- Learning Curve - teach time, intuitive rules, kid-friendly
- Component Quality - premium materials, inserts, art
Board Games (Strategy & Teen)
- Strategic Depth - many paths to victory, asymmetric, decision space
- Replay Value - variable setup, factions, expansion impact
- Rules Overhead - learn-as-you-play, clear rulebook, reference cards
- Component Quality - miniatures, art quality, insert
Card & Party Games
- Fun Factor - table laughs, instant fun, memorable moments
- Scaling - works at 4 and at 12, no dead time
- Teach Time - under 90 seconds is the gold standard
- Replay Value - variable cards, expansion fresh
Screen-Free Audio
- Content Library - size, growth rate, subscription-free
- Durability - drop-proof, kid-tough materials
- Ease of Use - kid operates independently, one-button intuitive
- Parental Controls - schedule, volume cap, app curation
Building Toys (Toddlers)
- Engagement - open-ended creativity, favorite-toy energy
- Developmental Fit - fine motor, spatial reasoning, STEM-friendly
- Durability - strong magnets, won't break, decade-long
- Compatibility - interoperates with Magna-Tiles / Picasso Tiles / LEGO
STEM Kits (Ages 6-10)
- Educational Value - real engineering vs. craft-kit cosplay
- Reusability - many projects vs. single-use
- Age Appropriateness - scaffolded difficulty, doesn't frustrate
- Parent Effort - minimal supervision, kid-runs-it
Plush & Collectibles
- Quality Materials - fabric softness, weighting, washability
- Design Longevity - timeless vs. trend-driven
- Collectibility - active community, secondary market
- Gift Appeal - universal vs. polarizing
Ride-On Toys & Balance Bikes
- Safety - low center of gravity, certifications, tire stability
- Growth Range - adjustable seat, age 1.5 to 5+, convertibility
- Build Quality - frame material, warranty, replaceable parts
- Resale & Hand-Down - 60-70% resale on Strider vs. zero on no-name
Premium Puzzles
- Piece-Fit Consistency - German precision vs. drugstore tier
- Art Quality - original commissioned vs. licensed stock
- Material Quality - board thickness, finish, sustainability
- Packaging - reference poster, sorting trays, premium box
The composite formula
Once we score each dimension on a 0-100 scale, we composite them:
CurioRank =
0.25 × Owner Satisfaction
+ 0.20 × Value
+ 0.55 × (average of category-specific dimensions)The result is rounded to an integer between 0 and 100. We also publish the sub-scores on every product page so you can interrogate the result. If a product scores 92 overall but a 65 on Build Quality and a 99 on Owner Satisfaction, you can see exactly where the gap is.
Try the calculator yourself
Move the sliders. Every product on PawBench resolves to one composite score using these exact weights. No magic numbers.
Composite score
Owner Satisfaction is weighted 25%, Value 20%, and the three category-specific dimensions share 55% equally. Real product scores compute exactly this way - every breakdown row on a product card is one of these dimensions.
What we don't claim
We don't claim to have hands-on tested every product on this site. We're transparent about when we have evaluated a product first-hand and when we've researched it from public sources (reviews, manufacturer specs, community discussion, expert reviews from sites like BoardGameGeek, Brickset, Common Sense Media, and Wirecutter). Where we have hands-on time, we say so explicitly.
We don't accept paid placements. Affiliate commissions on Amazon and other retailers fund this site, but the recommendations are the same recommendations we'd give a friend. We've disclosed multiple products as "skip this" or "save your money" within this site precisely because the recommendations don't bend toward higher-priced products for commission.
How the score gets updated
The CurioRank is regenerated whenever a product's underlying data changes - rating, review count, pros/cons, specs, tier, or derived sentiment. This is automated. So the score on a product page reflects the most recent data we have. We publish a "last updated" timestamp on every product page.
The code is open to inspection
The exact formula lives in lib/scoring.ts in our public-facing source. The dimension keyword lists and weights are in the same file. If you want to audit the math, the math is yours to audit.