Furniture Retail Price Index Methodology

Why this matters

Furniture retail prices can move quickly through promotions, assortment changes, currency effects, supplier costs, and retailer pricing strategy. Because Furnilytics follows product-level prices repeatedly over time, the index can show historical retailer price movements that are difficult to see from short scrape samples, broad inflation statistics, or one-off assortment checks. Official inflation and producer-price statistics remain important context, but they often describe broader source concepts or slower publication cycles. A retailer-observed price index gives a more direct view of listed furniture prices faced by buyers online.

Indicator scope

The index covers selected furniture product groups that can be observed repeatedly online and matched over time. The current scope focuses on common home-furniture categories such as desks, wardrobes, storage furniture, kitchen and bathroom furniture, tables, bookcases, TV furniture, and chairs.

Accessories, spare parts, add-ons, and products that cannot be classified reliably as main furniture items are excluded where possible. New products are collected privately first and enter the measurement only after they satisfy continuity and classification rules.

Source observations

Furnilytics collects recurring online product observations from large furniture retailers in each country. Each observation records the product identity, product type, country, observation month, and listed price fields needed for index calculation.

Raw product-level observations are not published in the public catalogue. The public professional dataset exposes only reviewed country-level index values, while internal diagnostics retain the detailed QA fields needed to understand movements and source coverage.

Product eligibility and classification

Products must be suitable for repeat measurement before they can contribute to the index. Furnilytics classifies product observations into furniture product types and separates main furniture items from accessories, bundles, replacement parts, and ambiguous records.

New products are admitted conservatively. A product must show enough consecutive monthly observations and a usable classification before it contributes to the measured product-type index. This reduces noise from one-off products, temporary listings, and incomplete observations.

Index construction

Furnilytics first calculates product-type indexes from matched product observations within each retailer and country. The product-type results are then aggregated into country-level retail price indexes using product-type and retailer weights.

The professional headline series uses a rolling GEKS-Jevons method. GEKS is a multilateral index approach that helps preserve comparability when online assortments change over time. Furnilytics retains a matched-chain method internally as a control series for QA and method-divergence review.

Historical reconstruction and family-chain logic

Some historical retailer data is sparse because online assortments change, product identifiers are retired, and older web observations may not contain the same structured fields as current scrapes. For these periods, Furnilytics does not treat a broken product identifier as proof that the comparable product disappeared. Instead, the index uses a documented reconstruction layer that combines observed monthly movement with conservative product-family continuity checks.

Product-family keys are built only within the same retailer, country, and product type. The matching logic uses product names where available, otherwise URL or item slugs, removes colour, material, and finish words that often change within the same model family, and keeps model, series, and dimension information where it is available. This allows a product family such as a named cabinet or table series to remain comparable even when an item number changes because of a colour, finish, or minor listing update.

Family-chain evidence is guarded before it can influence the headline path. Furnilytics checks the number of matched families, continuity across months, product-type support, month-to-month movements, and price spread within a proposed family. Matches with weak names, mixed sizes, large internal price spreads, or unstable movement are kept in diagnostics rather than promoted into the published index.

Where direct product or family evidence is thin, the historical path may use a calibrated bridge profile. Such reconstruction is an overlay on top of the raw observations: raw product data is retained separately, internal root-cause tables keep the evidence and flags, and public country values expose only the reviewed headline series. The purpose is to provide a realistic comparable-price trend over time, not to recreate the exact assortment that happened to be visible on any single historical scrape date.

Weights and aggregation

Product-type weights are fixed for the current methodology version and use external furniture trade classification proxies where available. Fixed product-type weights reduce the risk that short-term source composition changes dominate the index.

Retailer weights use annual retailer turnover inputs and transition gradually between complete fiscal-year weight vectors. This avoids abrupt weight shifts while still letting the country index reflect changes in retailer importance over time.

Quality and publication filters

The professional dataset includes only rows that pass publication filters. These checks consider coverage, contributing product types, contributing retailers, large month-to-month movements, method divergence, and other signals that could indicate an unstable headline value.

Non-ready rows are kept in internal diagnostics rather than removed from the calculation history. This allows Furnilytics to review root causes, distinguish low-impact source issues from material index risk, and improve the methodology without exposing unstable headline values.

The same principle applies to reconstructed history. Retailer-country components, family matches, bridge inputs, weight vectors, and publication decisions are retained in hidden diagnostics first. The public country dataset is then built from the reviewed components only, so users see a clean headline index while Furnilytics can still audit and improve the underlying evidence.

Interpretation limits

Online listed prices may differ from final transaction prices because of delivery fees, vouchers, member discounts, in-store offers, financing terms, and stock availability. Promotions can create short-term volatility, especially when a product group has fewer matched items.

Reconstructed historical values should be read as Furnilytics' best comparable-price estimate from the available product, family, retailer, and country evidence. They are more suitable for inflation-style trend reading than for auditing the exact historical listing mix. When the evidence is weak, the affected retailer-country rows remain flagged internally and may be revised as stronger scrape history, family-chain coverage, or retailer weights become available.

Retailer coverage is expanded gradually. The index starts with large retailers in each country and is designed to incorporate additional retailers only after they pass the same product identity, classification, weighting, and QA requirements.

Revision and update policy

The index is refreshed when new recurring observations and reviewed methodology inputs are available. Historical values may be revised when product classification improves, retailer coverage expands, weighting inputs are updated, or QA rules are refined.

New scrape months are processed through the same evidence hierarchy. Current product and family observations update the direct movement evidence first; reconstruction and bridge logic are used only where continuity is incomplete or not yet publication-ready. This keeps the live index responsive to new data while preserving a defensible long-run history.

Related methodology notes

Related pages