Six stages. No black boxes.
Every LotSignal verdict comes out of the same six-stage pipeline. Nothing is hardcoded per vehicle; nothing is hidden from the buyer defending the offer.
- 01
Parse
Extension parses the private-party listing (or a pasted description) into a structured vehicle identity — year, make, model, trim, mileage, drivetrain, fuel, title status.
- 02
Identify
The identity engine normalizes to canonical make/model/trim families. HD trucks are separated from LD trucks; hybrids from gas variants; cab configs and drivetrain are preserved.
- 03
Retrieve
MarketCheck is queried with a ladder of targeted filters. Zero-result classifiers explain why filters miss when they do, without silently over-relaxing.
- 04
Rank
Duty class is a hard gate. Mileage bands (tight → widened → context) are adaptive by class. Only exact-trim, in-band candidates score.
- 05
Price
Retail low, median, and high are computed from the effective pricing set. Recon is applied by category. Suggested offer, max buy, and expected gross fall out of the math.
- 06
Verdict
Worth a call, Borderline, Skip, or Insufficient Accurate Comps — each with a plain-English reason and confidence rating.
What LotSignal will not do
- —Invent comps. If exact-trim comps aren't available, it says so and refuses to score.
- —Relax standards to produce a verdict. Insufficient comps stays insufficient.
- —Overwrite a saved appraisal. Every re-check adds a new immutable snapshot.