Every term NextComp uses, defined concretely with the mechanism it points to. Tagged with schema.org DefinedTerm so AI answer engines can cite specific definitions directly.
0–100 trust score on a BaT auction outcome.
A 0–100 score we attach to every BaT auction sale measuring how likely the outcome reflects a real arm's-length transaction. Starts at 100 and subtracts for known noise patterns: Reserve Not Met (-30), retail relist at a premium (-25 to -35), self-dealt outcomes (-40), low bidder diversity (-20), top-bidder concentration (-15), and a few smaller signals. Each contributing factor is named in the tooltip on /bat.
Per-dealer measure of buying from auctions and reselling at retail.
A per-dealer scorecard derived from the relist_linkages table: n_active_sourced (active inventory traced to a recent auction), median relist premium over the auction close, median time-to-relist, and stale-share (sourced listings ≥60d on lot). No black-box rollup — each metric stands alone on the dealer page.
Anomaly flag — repeat bidder against the same seller.
Fires when an auction's top bidder has bid on ≥3 of the seller's auctions historically. Indicates a network conflict-of-interest pattern even when winner ≠ seller. Persisted in auction_anomaly_flags; downweights authenticity by 15 points.
Median ask discounted by expected reduction share × median cut size.
When BaT median sold prices aren't available, the fair-value anchor falls back to the dealer-median ask discounted by the share of inventory we've observed cutting × the median cut percentage. Shown alongside median ask and hedonic-fair in the Pricing Compass.
Days a listing has been live since first observed.
Days between today and the listing's first_seen timestamp in our snapshot history. Used as the x-axis for stale-inventory analysis and as a denominator for cut-velocity metrics (cut $/day).
Spec-aware price gap vs the model cohort.
Per-model OLS regression of log-price against mileage, age, and key spec flags. The residual is each listing's ask minus what the model predicts for its spec. Standardized to a z-score so positive = priced over the spec-aware fair value, negative = priced below.
Anomaly flag — single bidder ≥40% of all bids.
Fires when one bidder placed ≥40% of all bids on an auction with ≥6 total bids. Indicates the auction was thin even if the top-line bid count looks healthy. Persisted in auction_anomaly_flags.
Per-trim Heating / Stable / Softening label.
Per-filter sentiment composite derived from price trend slope, inventory delta, cut share, and sell-through. Each component is scored in [-1, +1] and averaged. Activates only when ≥7 distinct snapshot days have been observed.
Auction → retail relist trail under a listing.
An inline trail badge on active listings indicating the same VIN sold or RNM'd on BaT or Cars & Bids within the last year, with gap-days and dealer markup. Sourced from the relist_linkages table.
VIN-matched auction → retail pair.
A row in the relist_linkages table tying a (auction_source, auction_id) to a (retail_listing_id) when the same VIN appears on both within 0–365 days. Stores gap_days, markup_pct, and matched_via='vin'.
Auction ended without meeting the seller's reserve.
Bring a Trailer and Cars & Bids status indicating the high bid didn't clear the seller's hidden reserve. The price shown is the bid the market refused — NOT a clearing transaction. Authenticity scoring downweights RNM rows by 30 points.
Anomaly flag — winner = seller on the same auction.
Strongest anti-clearing signal: the winning bid was placed by the same username as the seller. Cars don't sell when the seller is the buyer. Authenticity scoring subtracts 40 points.
Per-model buyer recommendation: Buy now / Negotiate / Wait / Patient buyer / Hold off.
Synthesized from market_state, cuts_share, sell_through, ask-vs-BaT, and inventory direction. Verdict is computed independently from confidence so thin-data models still get a recommendation alongside a clearly-labeled confidence chip.
Named behavior signal lowering market_trust.
One of: repeat-relist (0.30 weight), price-anchoring (0.25), stale (0.20), volatile-pricing (0.15), aggressive-cutter (0.10). Market trust = 1 − sum of triggered flag weights. Each flag has a detail string explaining why it fired.
Related: Methodology (formulas + confidence rules) · How it works · Features.