What “stronger” means in this one report
The public sample lists relevance scores of 81, 79, 77, and 72. For this audit, 81 and 79 are described only as the stronger two within that report, and 77 and 72 as the weaker two. The labels preserve the report order; they do not reveal a formula. A useful comparable still has to be read across artist or attribution, subject and composition, medium and support, size, condition, provenance, venue, date, and price basis.
- ✓4 selected different-artist rows
- ✓Case-specific scores: 81, 79, 77, and 72
- ✓Observed score spread: 9 points
- ✓0 rows are by the same artist as the described subject
Auditability changes the evidence conversation
The Thomas Sidney Cooper row has the highest report score, but this audit did not retrieve its original auction-house page and the sample does not state dimensions. The Hendrikus van de Sande Bakhuyzen and Franz van Severdonck rows have original-host pages that support core metadata and reported amounts. The Matthijs Quispel row remains a report-only transcription. That pattern is why relevance and auditability should not be collapsed into one confidence number.
- ✓2 original-host rows retrieved and checked
- ✓2 report-only transcriptions
- ✓Highest relevance score does not mean strongest source trail
- ✓Missing fields remain missing rather than inferred
Two verified rows, two different price bases
AAG identifies lot 44 as an oil-on-panel work measuring 38.5 × 52.8 cm and labels €2,500 as the Hammer amount. Its current lot page displays June 25, 2017, while the sample report transcribes June 26; the one-day discrepancy remains visible. Bonhams identifies lot 27 as an oil-on-panel work measuring 45 × 66.5 cm and states that it sold for £5,250 including premium. Those figures cannot be treated as the same basis.
- ✓AAG: €2,500 hammer amount
- ✓AAG date: source page June 25; report June 26
- ✓Bonhams: £5,250 including premium
- ✓No cross-currency or raw-price ranking
Method, interpretation, and limitations
The audit transcribed the sample, searched for auction-house records, and accepted an original-host page as primary support when retrieved. It did not substitute commercial discovery databases for original records. Condition, restoration, provenance, attribution strength, and fee basis are not consistently available across all four rows. No FX, inflation, price-per-area, market-growth, or authentication adjustment is calculated, and the audit does not retest the report value conclusion.
- ✓Company-authored research by Appraisily
- ✓Source pages and calculations checked October 27, 2026
- ✓Selected case audit, not a representative cohort
- ✓Not an appraisal, formula, market index, or independent validation
Questions people ask
What makes a comparable stronger or weaker?
For one assignment, relevance can depend on artist or attribution, subject, composition, medium, support, size, condition, provenance, venue, date, and price basis. A strong source trail is a separate question from how closely the object matches.
Do the relevance scores create a valuation formula?
No. The scores of 81, 79, 77, and 72 are case-specific professional judgments preserved from one public sample report. They are not a validated universal scale, a price adjustment, or an authentication probability.
Can the four sale prices be ranked against each other?
No. The rows use GBP, EUR, and USD, and price bases are inconsistent. AAG labels its €2,500 result as hammer, while Bonhams states £5,250 including premium. No FX or inflation conversion is applied.
Which rows were checked on original auction-house pages?
The AAG Auctioneers Bakhuyzen lot and Bonhams Severdonck lot were retrieved and checked. The Cooper and Quispel rows remain report-only transcriptions in this audit.
Why are two dates shown for the AAG row?
The sample report transcribes June 26, 2017, while the current AAG lot page visibly displays June 25. The one-day discrepancy is unresolved and retained instead of silently normalized.
Can I inspect the audit data?
Yes. The page links to a four-row CSV, a JSON file containing the method and limitations, the public sample report, and the two original-host lot pages. Source pages and calculations were checked October 27, 2026.