Listing Matching (SKU Matching)

Listing titles are not standardised across recommerce platforms: the same physical device might appear as "Apple iPhone 13 Pro 128GB Graphite Unlocked" on one marketplace and "iPhone13Pro 128G Grey SIM-Free" on another. Listing matching parses these unstructured titles and attribute fields to determine that two listings describe the same configuration, typically using pattern extraction, fuzzy string matching, or classifiers trained on marketplace catalogue data. It is a distinct step from price normalisation: matching answers whether two listings are the same device, normalisation adjusts for how their condition, currency, or tax treatment differ once that is established.

Matching errors compound silently. A parser that resolves "Graphite" and "Grey" to the same colourway when the platform actually treats them as separate SKUs, or that fails to separate a 128GB listing from a 256GB one due to inconsistent title formatting, will corrupt every downstream price comparison built on that match. Regional variants add a further layer of risk: a carrier-locked US model and a SIM-free EU model may share a marketing name while running different hardware, and pooling their prices without disclosing the distinction misrepresents both the competitive benchmark and any buyback quote built on it.

Because catalogues span thousands of model, storage, colour, and condition combinations across 30 or more platforms, listing matching cannot be maintained manually at any meaningful scale. Automated repricing and buyback pricing tools depend on match tables that are updated continuously as new models, colourways, and carrier variants are released; a match table that lags a launch silently excludes the new SKU from competitive monitoring until it is manually added. Reliable listing matching is the foundation every other pricing intelligence process - normalisation, indexing, alerting - is built on top of.