vsc-pricing / data_dictionary.md
anas-muhammad's picture
Publish schema v3.6.7
fbc9eec verified

Data Dictionary — Schema v3.6.7

This document describes every column in every table of the OptimalCover VSC Pricing Dataset (v3.6.7). All values are non-PII, methodology-derived reference data.


vsc_rates

Canonical reference prices for Vehicle Service Contracts at the authoritative $100 deductible. One row per (vehicle_class, mileage_band, term, band) combination.

Column Type Description Allowed / example values
coverage_level string Coverage tier the rate applies to EXCLUSIONARY (current release covers exclusionary only)
vehicle_class string Risk class assigned to the vehicle A, B, C, D (A = highest reliability / lowest price, D = lowest reliability / highest price)
mileage_band string Label of the mileage band at purchase e.g. "0 - 48,000", "48,001 - 60,000" — see mileage_bands table
term string Term length label e.g. "24 / 24,000", "48 / 48,000" — months / miles
deductible integer Per-claim deductible, in USD Always 100 in this dataset. Other deductibles ($0, $250) are derived at runtime per pricing_config.deductible_handling
band string Which end of the reference range this row represents REFERENCE_LOW or REFERENCE_HIGH
price integer Reference price, in USD, rounded to nearest $50 e.g. 1450
term_months integer Term length in months e.g. 24, 36, 48, 60, 72
term_miles integer Term length in miles e.g. 24000, 48000, 72000

How to use: pair REFERENCE_LOW and REFERENCE_HIGH rows sharing the same (vehicle_class, mileage_band, term_months, deductible) to form the reference price range for that scenario.


vehicle_eligibility

Make/model → vehicle class (A–D) mapping plus eligibility status.

Column Type Description Allowed / example values
vehicle string Display name of the make e.g. "Chevrolet", "BMW"
normalized_vehicle string Uppercase normalized name used for lookups e.g. "CHEVROLET"
vehicle_class string | null Risk class assigned to the make A, B, C, D, or null if ineligible
eligibility string Whether any model under this make can be priced ELIGIBLE, INELIGIBLE
models JSON-encoded array of strings Model names covered by this row e.g. ["1500","2500","3500"]. In CSV/Parquet this is a JSON string; in the JSON file it is a native array
note string | null Optional note (model exceptions, special handling, etc.) free-text

What is NOT eligible (per methodology): vehicles used commercially, older than 15 years, over 84,000 miles, 10/12-cylinder engines, EVs, and hybrids. These exclusions are enforced by the pricing engine, not encoded row-by-row here.


mileage_bands

Definitions of the mileage bands referenced by vsc_rates.mileage_band.

Column Type Description
label string Human-readable band label (matches vsc_rates.mileage_band)
min_odometer integer Inclusive lower bound of the band, in miles
max_odometer integer Upper bound of the band, in miles
max_inclusive boolean true if the upper bound is inclusive (i.e. <= max_odometer)

Current bands (v3.6.7): 0 - 48,000, 48,001 - 60,000, 60,001 - 72,000, 72,001 - 84,000.


pricing_config

Methodology constants encoded as JSONB key/value pairs. One row per configuration key.

Column Type Description
config_key string Name of the configuration block
config_value JSON object (in .json) / JSON-encoded string (in .csv / .parquet) The configuration payload — structure varies by config_key

Configured keys

config_key Purpose
schema_version The canonical schema version this release corresponds to (e.g. 3.6.7)
deductible_handling How non-canonical deductibles ($0, $250) are derived from the canonical $100 prices — multipliers, rounding, caps
constraints Class/term caps (e.g. Class D can only be priced up to 60 months under certain mileage bands) and extrapolation policy
eligibility_constraints Scope rules — e.g. max_vehicle_age_years = 15
rounding_policy Rules for rounding derived prices to $50 increments

Integrity

Every file in this release has a SHA-256 checksum in checksums.sha256. Verify with:

sha256sum -c checksums.sha256

Source of truth

This data is exported from the OptimalCover production pricing database (Supabase, v3.6.7 shadow tables). The export pipeline lives in the backend repo (scripts/exportOpenData.ts) and runs against the canonical *_v367 tables.

Questions or discrepancies? Open an issue at github.com/Optimal-Cover/vsc-pricing-data.