A Field Guide to Cataloging Art

Find your page

Most people pick up a catalog because something forced the issue. Find yours, and start there.

Contents

The problem in one sentence

A painting is a physical object in a room; a catalog is a row in a database; and almost everything that goes wrong in this category is a failure of the link between the two.

Start with the shoebox

Imagine a family photo collection. There are three ways it can exist.

Software for cataloging art is the third one. Everything in this guide follows from that: the identity of each work doesn't depend on where it sits.

The four things a catalog holds

Cheap software does the first well and the fourth not at all. Expensive software does all four. That's most of what separates the prices, but not quite all of it. The next page explains why.

Two kinds of record

The category looks like one market with a price ladder. It's really two different kinds of record with similar-looking screens. Choosing between them decides which questions your catalog can answer in thirty years.

Why it's not just "more fields"

Paying more for a commercial record buys more fields. It doesn't buy the second model. Three questions can't be answered from a commercial record at any price:

Free, open-source software built by museums exists and has deeper models than anything sold to artists. What it costs instead is setup and a technical person for as long as the catalog exists. That's a good trade for an institution with staff, and a hard one for a single person running an estate.

Words you'll meet

Most of these come from museums, where people have been keeping these records for a century. Some pairs look like synonyms and aren't, and those are the ones that matter.

Provenance

The chain of custody: who held the work, from when, and to whom it passed. From the French for origin. Documented provenance back to the studio is what makes a work probably genuine, and it raises the value accordingly. p. 8

Accession

A record of one transfer into a collection: what arrived, from whom, when, on what terms, in what condition. Provenance is the whole story; an accession is the receipt for the part you witnessed. p. 8

Accession number

The identifier given at that moment, usually shaped like 2026.14: the fourteenth thing to arrive in 2026. Never reused, even after the object leaves.

Deaccession

Formally removing an object from a collection. Museums treat it as a governed act with a policy behind it, not a deletion.

Current holder

Whoever has the work now. Not the same as the buyer. A former student may hold a painting nobody ever sold them. Recording them as "sold to" invents a sale.

Basis

What a claim rests on: a document, a photograph, an inspection, or someone's word. A provenance line without its basis can't be defended later.

Provenance gap

A stretch of a work's history with no documentation. Ordinary in older works, and serious for European works between 1933 and 1945, where a gap raises a question about looting that a buyer has to resolve.

Catalogue raisonné

The complete documented body of work of one artist, compiled by scholars: every known work, with everything known about each. It's the reference that decides whether a newly surfaced painting is accepted. p. 16

Constituent

Museum word for any person or organization connected to an object: maker, donor, lender, conservator, photographer. The point of the word is that they're all the same kind of record, linked to objects by role.

Registrar

The person responsible for the records: arrivals, loans, locations, condition. In a small museum it's part of somebody's job. In an estate it's nobody's job, and that's the source of most of the difficulty.

Edition

A work made in numbered copies: prints, casts, photographs. One work, many instances, each individually identifiable. Software that doesn't understand editions stores fifty records that don't know they're related.

Consignment

A dealer holds and sells someone else's work for a share. The terms (who, since when, what split) are a relationship, not a status. A work on consignment hasn't been sold.

Value, care and pictures

Fair market value

What the work would sell for between willing parties. The number for estate tax, donation and sale. p. 7

Insurance (replacement) value

What it would cost to replace. The number for the insurer. Usually higher than fair market value, sometimes considerably.

Asking price

What the studio or gallery is asking today. A third number, and it tracks neither of the other two.

Appraisal

A professional's opinion of value, for a stated purpose and as of a date, made under professional standards. A document about the work, written by someone outside the catalog. Software stores appraisals; it shouldn't produce them.

Insurance schedule

A list of works with values that an insurer uses to set coverage. It has to be complete, current, and reproducible whenever the collection changes. Software can produce this one.

Condition report

A dated description of a work's physical state, made before and after it travels, so any damage can be traced to the journey.

Verso

The back of the work. Labels, stamps, gallery stickers and inscriptions live here: provenance, physically written down. p. 9

Credit line

Exactly how a photograph must be credited when published. It belongs to the photographer, not the painter. p. 10

IPTC metadata

A standard set of fields carried inside image files: photographer, copyright, credit line, usage terms. Less well known, it also has a set describing an artwork shown in the photograph, kept separate from the photograph's own details.

Authority control

Linking a field to a published, identified term instead of typed text, so the record points at the concept "etching" rather than holding the word. p. 12

Getty vocabularies

The best-known published term lists in this field, free to use: AAT for materials, techniques and styles; ULAN for artists' names; TGN for places.

Interchange standard

A published format that another system already knows how to read, so a catalog can move without a translation project. The museum world uses several: LIDO, Dublin Core, MARC, plus the SPECTRUM and CCO practice standards. p. 13

Take-out (export)

What comes with you when you leave the software. The property that matters most, and the one least often advertised.

What a work is worth, and when

Two documents for two audiences, drawn from the same records. An insurance schedule is a list of works with values, for an insurer. An appraisal is a professional opinion, written by someone outside the catalog. Software makes the first and stores the second.

The three-values problem

Any work has at least three values at once, and mixing them up is the most common error in this kind of data.

A catalog with a single "price" field holds whichever number was typed in last. An estate that reports it as fair market value on a tax filing has made a consequential error, and the cause is the shape of the record.

An appraisal is an opinion as of a date. Keeping every value with its date, instead of overwriting the old one, means a work's value is a history, and a collection's value over time becomes a chart rather than an exercise.

What an estate needs that most catalogs don't give

• A defensible date on every value. Storing the current value carefully but the date casually produces documents nobody can rely on later.
• Values on works you don't hold. A work known from a former student's photograph has a fair market value that matters to the body of work, and no insurance value at all, because nobody on your side insures it.
• Every line traceable to its opinion. An insurance schedule listing 436 values should be able to say, line by line, which appraisal each one came from.

Museums treat insurance as a subject of its own, with policies, coverage periods and values per policy. That handles a work covered under a different policy while it's out on loan, or a lapse somebody needs to date. For most private owners, an inventory report with a value column is enough.

Provenance, and the receipt nobody keeps

Provenance is the story. An accession is the receipt for the chapter you witnessed. Nearly all software offers the first and almost none outside museums offers the second, and the second is what makes the first defensible.

What provenance usually looks like

In most catalogs, provenance is a list attached to the work: prior owners, in order. Each owner is typed text. The line has no document of its own and no record of who said so, or when. That's enough to print for an appraiser. It isn't enough to defend a claim, because the record doesn't carry its own basis.

In a museum system, each line points to a person record, and each arrival is an event with a date and an agreement. Ask "on what basis do we say this passed to X in 1974?" and there's a place for the answer.

The case that breaks it: a work turns up

It isn't exotic. It's the ordinary way an estate grows. A former student, a sitter, a neighbor or a relative has a painting nobody knew about. They offer a photograph, or the work itself. That record has three properties most catalogs can't express:

The smallest useful receipt

For each work that arrives from outside the original inventory, write down:

A spreadsheet is enough for this, and it's roughly what much of the museum world keeps underneath its software.

What ties a record to the painting

Every catalog stores a record of an artwork. The record is only useful while something links it to one physical object in the world. That link is rarely discussed, and it's the first thing to fail.

Three ways it's done, and how each one fails

They fail independently, which is the argument for using more than one. Any catalog worth building does the first two. The third is worth watching and not yet worth depending on.

An estate inherits a numbering scheme with no key, and the number of works whose labels survived is the number of works that can still be identified. The works most likely to look alike (studies, class demonstrations) are also the ones most likely to be undocumented, and most likely to surface later.

Photograph the verso of every work at the same resolution as the front, and keep it on the record. The back carries the labels, stamps, gallery stickers, the inscription in the artist's hand, stretcher marks and repair history. It's where provenance is physically written, and it's unique to the individual object in a way the front often isn't.

Nearly every workflow throws it away, because it doesn't photograph well and nobody wants to look at it. Any catalog can store a second picture. Almost none of them tell you to.

The photographer problem

A photograph of an artwork has two copyrights in it. The painting belongs to the artist or the estate. The photograph belongs to whoever took it. A catalog that records only one will eventually publish an image it has no right to, or fail to credit someone who's owed it.

So: is the photographer a field, or a note? In most software sold to artists and galleries, a note. Pictures are attached to works, and the person who made them isn't recorded. In museum systems, the photographer is a person linked by role, and the rights position is tracked for each object.

The part that's already solved, inside the file

The standard for exactly this problem exists and travels inside every image file. IPTC photo metadata has a set of fields for an artwork shown in the photograph, kept deliberately separate from the photograph's own fields:

Two creators, two copyrights, one file, and the standard says which is which. It even expects the image to point back at the catalog record. Cataloging software generally throws all of this away on upload and keeps only the pixels.

Why it bites

An estate gets photographs from several sources: a professional shoot in 1994, the artist's own slides, a student's phone picture in 2026. That's three different rights positions and one picture field. The difference survives only in whatever someone remembered to type into a notes box.

The record that survives

For each picture: who took it (a person, not text) · when · rights (we own it · licensed · permission only · unknown) · credit line, exactly as it must appear · and may publish: yes or no, defaulting to no. That default is the part that matters. An image of unknown rights that's publishable by default is a rights problem waiting for a website redesign to set it off.

What the world sees, and what it doesn't

A catalog holds two kinds of fact about the same work. Title, year, medium, size, a picture and exhibition history can go on a public page. What it sold for, who bought it, what it's insured for, where it's stored and the owner's phone number must never appear there.

Three ways software draws the line

Where the line leaks, whatever the software

The pattern that holds

Two catalogs, one of them authored. The private catalog is the record of everything. Publishing is a step a person runs, sending an explicit subset they can review before it goes out. It isn't a checkbox.

Who decides what a thing is called

Two people describe the same painting. One writes "oil on canvas", the other "oil paint on canvas support". One writes "Gary Faigin", the other "Faigin, Gary". One says "Seattle", the other "Seattle, WA". Each pair is the same fact. As typed text, they're four different strings, and no search will ever join them.

Authority control means linking a field to a published, identified term instead of typed text. The record doesn't hold the word "etching". It points at a concept that has an identifier, a preferred name, other names, a definition, a place in a hierarchy, and someone who maintains it.

It's the least visible gap between the two kinds of record (p. 4), because a controlled field and a text field look identical on screen. Nobody notices until two catalogs have to be combined and there's nothing to join them on.

The Getty vocabularies: free, and the reference

Published by the Getty Research Institute under an open license, as linked open data with a public search service. Museum systems use them. Software sold to artists and galleries doesn't: there, medium, style, technique and place are free text everywhere.

Why it matters more than it looks

• It's the difference between a catalog and a list. "Every work in egg tempera" returns whatever spellings one person happened to use.
• It makes a catalog joinable to the world. Two catalogs whose artist fields carry the same ULAN identifier can be matched without a name-matching exercise.
• It survives language changing. A record pointing at a concept follows the preferred name as it's revised.
• It's cheap at entry and expensive afterward. Choosing the term while typing costs nothing. Doing it to an existing catalog is a project.

For a working artist it's over-engineering, and free text is the right answer. For anything meant to be read by a second institution, cited by a scholar, or merged with another catalog (an estate, a catalogue raisonné, a museum), it decides whether the data is usable outside the system holding it.

What comes out, and what survives

A catalog is meant to outlive the events that make catalogs necessary. Over fifty years it will outlive at least one vendor, one file format, one hosting arrangement and one person who knew how it was organized. So what matters most isn't what the software does. It's what's left when the software is gone.

Among the products researched, ten of fifteen state no export path at all on their public pages. Most probably have a spreadsheet button. The point is that none of them treats it as a selling point, which tells you how well it's likely to be maintained.

Three grades of take-out

Most software sold to artists, galleries and collectors exports at grade 1. Only part of the museum world reaches grade 3, because museums move data between institutions routinely.

Four things that almost never come out

• Pictures at original resolution. The export may hold the display copy, not the file you uploaded.
• Attached documents. Appraisals, certificates, letters, condition reports, exported as file names if at all.
• Relationships. Which works were in which show, who bought what, which pictures belong to which record.
• The history of changes. Who changed what, and when: the only record of how the catalog came to say what it says.

Who needs what

Six kinds of people keep these records, and they don't rank. The catalogue raisonné project has the most exacting needs and the smallest budget; the gallery has the biggest budget and the shallowest need for provenance; the working artist needs speed above everything and would be harmed by depth. Nobody is only one of these.

What the six have in common

Six habits, whatever you use

They work in any software, including a spreadsheet. Nobody sells them, and each one is worth more than the choice between two similar products.

What I want my catalog to answer

Write down the questions from this guide that matter to you, and any it missed.