Entry cost determines how complete metadata will be

What it is

Whether fields fill up is decided first by how much work filling them takes, not by how many fields the architecture diagram drew. Entry cost determines completeness: each extra required field, each trip into a controlled vocabulary, each moment of recalling whether this date is published or revised, knocks completeness down. Completeness is not a moral property. It is an equilibrium of effort and payoff at the cataloguing desk. Demand ten fields for one field’s worth of retrieval benefit, and the extra nine will be blank or garbage.

Garbage is worse than blank, because it looks complete. When cost is too high, the completeness dashboard can still be green.

Why it happens

Cataloguing is extra labour, usually not done by the people who benefit from lookup. Producers pay; searchers collect. Incentives are misaligned by default. Cost includes time to find the right controlled term, time to understand the field definition, the anxiety of being forced to choose under uncertainty, and the expected future maintenance. Past a threshold the strategies become: skip, pick the first list item, copy the previous record, or stamp everything with the same popular tag.

Automatic extraction and inheritance (from a parent, a filename, an uploader identity) can drop cost from “create” to “confirm,” and completeness can then rise. Hoping that “people will fill it in carefully” fights that incentive structure.

Studying it

Measure real cataloguing work and the data quality that follows, not survey willingness to fill forms.

• Paradigms: time and error on cataloguing tasks (wrong controlled term, wrong date semantics); experiments that add or drop fields; autofill-plus-confirm versus a blank form. Audit completeness and garbage rate afterwards; do not use non-null rate alone.
• Independent variables: number of required fields, size of the vocabulary, presence of prefills, whether “unknown” is legal.
• Dependent variables: cataloguing time per object, non-null rate, audited accuracy, concentration on junk popular tags.
• Methodological note: cataloguing a few items in the lab underestimates cost; production is tens or hundreds a day. Non-null rate counts garbage as complete; add accuracy. Who fills the form (author, professional cataloguer, vendor) changes the cost structure entirely; stratify.

Where it stops holding

Legally mandated fields (clinical records, financial vouchers) must be filled even at high cost, but the mandated set should be small, and anything the system can compute should not be human labour. User-generated content almost never affords professional cataloguing; completeness has to come from defaults, extraction, and later crowdsourcing, not from twenty dropdowns before publish. Expert collections maintained by staff can trade high cost for high precision, which does not mean the same form can be thrown at ordinary authors.

Applying it

• For each field ask: who fills it, how long one fill takes, and what retrieval loses without it. If the three do not line up, delete the field; do not draw advanced search first.
• Prefill anything derivable from the file, the account, or workflow state, and allow edits. Do not leave a blank for people to recall.
• Allow “unknown,” so forced choice cannot write noise down as fact.
• Verify by auditing new objects a week after publish—the rate that is both non-null and correct. Non-null without accuracy is fake completeness. If accuracy falls as field count rises, cut fields before talking governance.

Related

• Within the group: G1.05.1 Metadata underpins filtering, sorting, and relating · G1.05.2 Advanced retrieval fails when metadata is missing
• Adjacent: G1.07 Content inventory and audit · G1.10 Controlled vocabularies and synonyms · G1.11 Architectural scalability and evolution
• Search terms: metadata entry cost · cataloguing effort · metadata completeness