Method
The starting point: what you already have
A provenance isn’t written, it’s reconstituted from scattered pieces you often own without knowing: invoices, correspondence, studio notebooks, exhibition catalogues, labels on the backs of prints, emails, bank transfers.
Start with an inventory of sources, before any data entry. Nothing is more discouraging than building a chain and then finding an invoice that contradicts it.
Ranking sources
They aren’t worth the same, and that ranking should be explicit in your practice:
- The document contemporary with the event: invoice, contract, deposit slip, notarised deed. The strongest evidence.
- Period publication: exhibition catalogue, article, mention in a reference work.
- Physical inscription: label, stamp, handwritten note on the back of a print.
- Testimony: a statement from a rights holder, a gallerist, a collector.
- Inference: what you deduce from a set of clues.
The last two levels are admissible provided they are flagged as such. An inference presented as a fact is what destroys a catalogue’s credibility.
When two sources contradict each other
This is the common case, and the wrong answer is to choose silently. An invoice dated 1971 and an exhibition catalogue placing the work with another owner in 1970: one of them is wrong, or there were two transfers.
Good practice is to keep both claims, state which you retain and why. The reader can then reconstruct your reasoning — exactly what scholarly work requires. Overwriting the discarded source turns a documented judgement into an unverifiable opinion.
Documenting a gap
A hole in the chain isn’t a weakness to hide, it’s information to publish. “Location unknown between 1938 and 1952” is a useful claim: it bounds future research and honestly signals the risk.
A gap papered over with vague phrasing — “private collection” with no date — is far more damaging: it gives the appearance of continuity without guaranteeing it, and turns against you at the first serious check.
The confidentiality reflex
At every entry, one question: is this row publishable? A purchase price, a living buyer’s name, an address are commercial and personal data. The default must be “no”, lifted by an explicit decision after review.
That discipline costs little at entry and prevents the one irreversible incident in this field: a disclosure no correction can undo.
In practice
- Inventory your sources before entering data.
- Attach to each link the source attesting it and its reliability level.
- Keep contradictory claims, stating your judgement.
- Publish gaps with their bounds rather than filling them.
- Normalise counterparties into reusable entities, not free text.
- Treat confidentiality as default deny, row by row.
Obscura Flow models provenance as dated events with normalised counterparties and attached sources, and marks unreviewed rows as confidential — publishing a row is a decision, not a configuration oversight.