Rokas Simkus

Actors for auction houses that publish their inventory and results openly, normalised into one shape so datasets from different houses can be joined without an adapter.

What these do differently

• Sold vs unsold is reported, never inferred. On a closed auction every lot shows a bid amount, including lots that never sold because the reserve was not met. On one sampled vehicle auction that was 30 of 48 lots. Each row carries the auction house's own sale term and an explicit outcome.
• Prices are labelled by what they are - hammer, current bid, starting price or estimate - so an average never silently mixes them.
• VAT treatment is captured. Margin scheme versus VAT-reclaimable is roughly 19-21% of real cost on a cross-border purchase, and most listing data drops it.
• Unrecognised values stay null. A blank field is honest; a guessed one quietly corrupts whatever you group by.

On access: robots.txt is re-read and re-checked on every run rather than assumed from when an Actor was written. No logins, no paywalls, no bot-protection bypass. These scrape lots, not people - no seller names, phone numbers or email addresses are collected.