Methodology

How BikeCluster compares products

BikeCluster collects public product and offer data from specialist cycling retailers, then maps store listings to one canonical product.

Product matching

Strong identifiers such as GTIN, EAN, MPN, and manufacturer SKU are considered first. Normalized brand, model, category, and product attributes provide supporting evidence. Ambiguous decisions are held for review instead of being merged automatically.

Prices and availability

Offers show the latest successfully observed price and stock state. Runs that fail health checks use a conservative ingestion policy so incomplete scrapes do not erase previously good data.

Price intelligence

Price-history signals are calculated from append-only daily observations, separately for each product, canonical variant, and supported delivery market. A signal is shown only when the current offer is fresh and the history has sufficient day coverage, store continuity, variant consistency, and market evidence. When local evidence is insufficient, any displayed fallback is explicitly identified as global history.

Typical prices use the median of daily observed lows rather than a retailer reference price. This prevents an unverified recommended retail price or a brief outlier from being presented as a deal.

Editorial content

Store descriptions may be used as source material for an original summary. AI-assisted content remains a draft until an administrator approves it.

Affiliate disclosure

Some outbound store links may be affiliate links. This does not change the price paid by the customer or the order in which offers are compared.