A price is only useful with its context.

Sources used

Archive captures retain the original publisher as the source. Internet Archive supplies capture provenance; it is not counted as another independent retailer.

Internet Archive Wayback Machine supplies the archived HTML and PDF captures. Source links on each product history open the exact dated capture.

Dates, prices and freshness

Archived observations use the snapshot’s calendar date, not a claimed first-sale date. Current pages use their actual fetch date. Dates more than seven days old carry a stale-price warning. Availability is never inferred from an archived price.

Prices are listed Australian dollars. Mwave’s Australian retailer context and MSY’s Australian GST-inclusive parts lists establish the currency. Shipping, membership restrictions and surcharges are recorded where visible; unknown delivery cost prevents a delivered-price ranking.

Normalisation and provenance

Products are separated by category, brand, exact model, total capacity and known variant. Manufacturer part numbers are preferred; otherwise conservative product names and retailer SKUs prevent ambiguous variants from merging. Different retailer identities may remain separate when equivalence is uncertain. A 2 × 16 GB kit contains 32 GB; missing capacity stays blank.

Each row stores its source URL, original publisher URL, capture timestamp, fetch timestamp, parser version, evidence locator and SHA-256 raw-page fingerprint. Re-running ingestion does not duplicate identical evidence. Unsupported or ambiguous offers are quarantined outside the observations table.

Charts and derived figures

Annual medians summarise observed listed prices. They are sensitive to changes in the product mix, retailer coverage and archive availability. Missing years remain gaps. Dots represent samples rather than a continuous daily price series.

A matched-product index appears only where at least three exact products have observations in every calendar year of the selected span. Each product’s annual median is divided by its base-year median; the geometric mean of those ratios is multiplied by 100. No missing years are filled.

RAM $/GB divides price by total kit GB. Storage $/TB divides by advertised decimal TB. GPU VRAM is never used as a performance-per-dollar ranking.

Price percentile is 100 × (historical prices below the candidate + half the tied prices) / sample count. Lower is cheaper. Verdicts require at least ten comparable observations across three dates with matching condition and shipping/offer basis. This build suppresses verdicts where those comparison conditions are unverified.

Access and limitations

Fetches identify PriceDrift and pricedrift.com.au, obey robots.txt, wait at least one second per host or the longer published crawl delay, and back off on rate limits. Raw fetches are cached below 5 GB. No paid APIs or accounts are used.

PCPartPicker and camelcamelcamel are excluded. JW Computers prohibits scraping, PC Case Gear restricts commercial reuse, and Centre Com’s robots endpoint could not be verified. OzBargain’s consumer-use terms were not treated as permission for this commercial database. These sources contribute no rows.

Coverage is a sample of recoverable evidence, not a census of the Australian market. Comparisons between different years may involve different products, and archived promotions may omit conditions. The database makes those limits visible rather than implying complete market history.