Methodology

A raw price is not yet information.

TrueCostData separates persistent market structure from transient fluctuations to produce point-in-time market references that can be documented, reproduced and challenged. This page sets out the principles. The full methodology is shared with institutional partners under NDA.

Raw observations→ Admissibility→ Statistical structure→ Signal→ Economic decision

The Problem

Physical-asset markets are noisy by construction.

Observed prices for vehicles, electronics and other physical assets mix genuine market movements with noise. A single listing, or a simple average, cannot tell the two apart.

Noise

Speculative listings

Unrealistic asking prices set by sellers under no pressure to transact.

Noise

Forced sales

One-off liquidations that temporarily understate the value of an asset.

Noise

Collection anomalies

Input errors, duplicates, conditional offers and bundles that do not reflect a comparable price.

Our objective is not to publish more prices. It is to estimate the latent market level, quantify the uncertainty around it, and detect when a change is persistent rather than incidental.

Core Principles

Eight commitments behind every published value.

Point-in-time, no look-ahead

A T-1 reference only uses observations that were available at the reference date. Later data never revises a historical reference.

Multi-source consensus

Prices are observed daily across multiple independent retailers, marketplaces and portals, then confronted with each other rather than taken in isolation.

Strict normalization

Electronics are compared at model, capacity and condition level. Vehicles are compared by cohort: model, build year, engine, mileage band and market.

Logged admissibility

Every observation is either retained or excluded by an explicit rule. Exclusions — outliers, stale data, conditional offers — are recorded with their reason.

Signal estimation, not smoothing

Temporal models distinguish a genuine market pivot from a statistical fluctuation, weighting sources by their observed reliability.

Uncertainty is published

Each reference comes with its admissible range and the number of contributing sources, so its robustness can be judged — not assumed.

Reproducibility

The same request for the same date returns the same value. A reference can be recomputed with the methodology version it was produced under.

Versioned methodology

Methodology changes are versioned and documented. Every published value carries the version identifier used to compute it.

T-1 Reconstruction

From an event date to a defensible reference.

For a historical claim or valuation, the reference is rebuilt through a fixed sequence of steps. Each step leaves a trace.

Event date

The case provides the business date at which value must be assessed.

T-1 resolution

The engine resolves the historical snapshot that applies under the documented T-1 rule.

Asset identification

Model, variant or cohort, and market are normalized to a stable identifier.

Source population

Observations from the snapshot are loaded with their provenance and collection timestamp.

Admissibility

Admissibility rules are executed; every exclusion is logged with its reason.

Aggregation

The reference and its admissible range are computed on the retained population only.

Evidence record

The result is linked to an Evidence ID so the report, the API response and the underlying record can be matched.

What a Reference Carries

Never a bare number.

A value is only as useful as the context that allows it to be verified. Every reference is delivered with its metadata.

reference_dateThe snapshot date the value belongs to (T-1).

reference_priceThe computed market reference.

observed_rangeAdmissible dispersion around the reference.

sources_observedInitial population of observations.

sources_admissibleObservations actually used in the calculation.

method_versionMethodology version applied.

evidence_idLink to the traceable evidence record.

calculated_atTimestamp of the computation.

Response structure Illustrative values

{
  "market": "DE",
  "asset_id": "[normalized_asset_id]",
  "event_date": "2026-08-18",
  "reference_date": "2026-08-17",
  "reference_type": "T-1",
  "currency": "EUR",
  "reference_price": 1234.56,
  "observed_range": { "min": 1189.00, "max": 1299.00 },
  "sources_observed": 27,
  "sources_admissible": 23,
  "method_version": "METHOD-vX.Y",
  "evidence_id": "EV-20260818-XXXXXXXX",
  "calculated_at": "2026-08-18T10:42:17+02:00"
}

Market Movement Signals

Value is a level, a direction, a speed — and an exposure.

Beyond the reference itself, TrueCostData measures how a comparable market is moving. Signals are computed on cohorts, not on individual listings.

Direction & speed

Index movement

Index change over 7, 30 and 90 days, compared with the category's own historical pace.

Supply

Comparable offer

Change in the volume of comparable listings. A rapid build-up can indicate selling pressure.

Seller behaviour

Price-cut frequency

Share of listings with price reductions — an early sign that sellers are adjusting expectations.

Consistency

Dispersion

Spread between sources. A widening spread signals a market that is not yet aligned.

Liquidity

Time on market

How long comparable offers stay listed. Lengthening exposure can suggest slowing liquidity.

Materiality

Exposure × movement

Links a market movement to the volume of assets exposed, to show when a small move becomes financially material.

A signal is not a recommendation. TrueCostData does not say "sell this vehicle" or "reimburse this amount". It indicates that a category is moving faster than its historical pace, so that a review can be triggered earlier. Exposure figures are indicative orders of magnitude, not realized losses or guaranteed savings.

Disclosure

What we publish, and what we share under NDA.

Principles are public so that every reference can be understood. Implementation details are shared with partners who need to validate them.

Public

This page

Point-in-time, no-look-ahead rule
Multi-source daily observation
Normalization and cohort definitions
Exclusion categories
Published uncertainty and source counts
Reproducibility and versioning
Evidence ID and reconstruction chain
Known limitations

Under NDA

Full methodology document

Source list and coverage by market
Estimation models and their parameters
Source reliability weighting
Exact outlier and freshness thresholds
Score and index computation formulas
Integrity mechanism and fields covered
Data retention and audit procedures

Known Limitations

A reference is only credible if its limits are stated.

Source availability Robustness depends on the observations available for a given market, asset and date. Source counts are always shown.

Low-volume assets Rare models or thin cohorts can show higher statistical dispersion and wider ranges.

Very short promotions Flash offers lasting only a few hours may not be fully captured by daily observation.

Ancillary costs Delivery, insurance and similar costs are not systematically included in observed prices.

Individual asset condition Cohort references reflect comparable market value, not the specific condition, options or history of one asset.

Backward-looking by design Indices and signals describe observed market behaviour. They are not price forecasts.

TrueCostData provides market references and technical traceability. It does not provide a certification, a legal or regulatory attestation, or a substitute for the client's own internal controls and compliance analysis. The client remains responsible for business decisions made using this data.

Full Methodology

Need to validate the method before relying on it?

The complete methodology document — models, parameters, thresholds and integrity mechanism — is available to institutional partners under a mutual NDA.

Request full methodology under NDA → See the method applied to real data Or test it on 20 of your own cases