coffeetrove started with a simple question: if you are in a new city and want good coffee, where do you go? the answer turned out to require building a scoring system from scratch. not because existing review platforms are bad, but because they optimize for engagement, not signal.
a five-star rating tells you almost nothing when 80 percent of rated places sit between 3.8 and 4.5. the range is too compressed. the inputs are too subjective. and most cafes in our database have no user ratings at all. i needed a scoring system that works with what i actually have: structured data about the cafe itself.
the golden drop score
the golden drop is a composite score built from data completeness and business signals. it does not measure taste. it measures how much we know about a place and what that knowledge implies about quality. a cafe with a full address, listed hours, a website, an active social presence, and specialty coffee indicators scores higher than one with just a name and a pin on the map.
data points per cafe: up to 55
score range: 0 + 100
tiers: legendary + exceptional + notable + common
independent bonus: +10 points
the scoring breaks down into two parts. the first is data completeness, worth up to 55 points. having operating hours adds points. having a website adds points. having photos, a phone number, wheelchair accessibility information + each piece of structured data contributes. this is not arbitrary. a cafe that publishes its hours and maintains an online presence is signaling something about how seriously it takes its business.
the chain problem
one decision shaped the entire system: how to handle chains. starbucks has complete data for every location. by pure completeness metrics, a starbucks in an airport would outscore a beloved local roaster that does not bother with google business profiles.
the solution was a three-tier chain classification. global chains (starbucks, costa, tim hortons) get identified and tagged. local chains (blue bottle, intelligentsia) get their own tier. independents + the single-location cafes + get a ten-point scoring bonus. this is an opinionated choice. i believe independent cafes are the backbone of coffee culture, and the scoring system should reflect that. you can explore this on the coffeetrove world map, where chain badges appear alongside each cafe's score.
working with missing data
the hardest design constraint was incompleteness. most scoring systems assume you have all the inputs. i do not. a cafe in tokyo might have perfect data from google. a cafe in lisbon might only exist in openstreetmap with a name and coordinates. the score needs to be fair to both.
the approach is additive, not subtractive. you start at zero and earn points for what is present. there is no penalty for missing fields. this means a cafe with limited data gets a low score, not a negative one. and as we add data sources over time, scores naturally improve without requiring manual recalibration.
what i learned
building this taught me that a scoring system is a set of values encoded as math. every weight is a belief. the independent bonus says "we value small businesses." the data completeness approach says "we trust transparency." the absence of user ratings says "we do not trust crowds with small sample sizes."
the golden drop score is now live across all 440,000 cafes on coffeetrove.com. cafes can embed their score as an svg badge on their own website. the score updates automatically as data improves. it is not perfect + no scoring system is. but it is honest about what it measures, and that matters more than precision.
the knowledge section documents the full methodology, including the brewing guides and bean origins that feed into our understanding of specialty coffee culture. the scoring system is one piece of a larger effort to make coffee discovery genuinely useful.