Back to Writing
February 2, 20266 min read
Share:PostLinkedIn

Bayesian Ranking vs. Recency: Why Chronological Feeds Dilute Marketplace Quality

Why chronological feeds favor spam over quality, and how we built Groveh's feed using weighted Bayesian estimations.

Chronological feeds create a structural bias: they reward posting frequency over quality. Well-reviewed vendor listings get pushed down by low-effort recent posts.

The Fallacy of Simple Averages

Sorting by raw average rating is equally broken: a single 5-star review outranks fifty reviews averaging 4.9 stars. This penalizes established vendors with high review volume.

Weighted Bayesian Estimations

Groveh uses a Bayesian rating model that adjusts average ratings based on review volume relative to marketplace baselines:

W = (R * v + C * m) / (v + m)

Where W is the weighted rating, R is the listing's average rating, v is review count, m is the minimum review threshold, and C is the marketplace average rating. This keeps proven vendors at top ranks while fairly benchmarking new listings.

Geospatial Re-ranking

Cached quality scores are combined with Uber H3 hexagonal indexing. Distance acts as a secondary tie-breaker within high-quality candidate sets rather than an absolute sort key.