The pipeline

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.

  1. 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.

  2. 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.

  3. 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.

  4. 04

    Rank

    Duty class is a hard gate. Mileage bands (tight → widened → context) are adaptive by class. Only exact-trim, in-band candidates score.

  5. 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.

  6. 06

    Verdict

    Worth a call, Borderline, Skip, or Insufficient Accurate Comps — each with a plain-English reason and confidence rating.

Guardrails

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.