A matrix demands complete metadata; missing fields block location

What it is

A matrix seats objects at the intersection of several axes; the cell’s coordinates are field values. Missing any dimension, the object has no place in that matrix: people who arrive with a complete coordinate get an empty cell and conclude the content does not exist. A matrix requires complete metadata. Cataloguing is not a nice extra here; it is the addressing premise of the type. A tree can still hang an object under a parent with an optional label missing. A matrix with a missing axis value makes the object vanish from the structure.

Complete means the dimensions used for location, not every field anyone can imagine. If the missing field is not a coordinate axis, the matrix still walks.

Why it happens

The addressing equation is value-on-A ∩ value-on-B. A null cannot enter the intersection, so every query that includes that axis excludes the object. An empty cell has two readings—“this kind does not exist” or “not yet catalogued”—and people almost always pick the first. Missingness is then taken as a business fact, not as data quality. Filling the field is the object being born inside the structure; better nav copy cannot replace that step.

Sparsity also infects neighbouring cells: when one audience value is widely empty, the whole row looks like that audience is unserved, and completeness on the other axis cannot help.

Studying it

Treat field completeness as an independent variable of matrix usability, not as hygiene after launch.

• Paradigms: the same matrix, a complete corpus versus a proportion of one axis gapped; known-item location with two attributes; ask what an empty cell means. Then contrast with “the under-fielded object is still reachable from a tree,” and see whether people still think the matrix is broken.
• Independent variables: non-null rate on coordinate axes, whether nulls are visible as “uncategorized,” whether single-axis browse is offered as a fallback.
• Dependent variables: known-item recovery in the matrix, rate of reading missing as “does not exist,” switches to search or to the tree.
• Methodological note: fully fielded lab objects make matrices look strong. Sample production null distributions. Count garbage fills in the non-null rate: a wrong coordinate sends the object to the wrong cell, which is harder to spot than a blank.

Where it stops holding

A single tree or pure hypertext does not address by cell; a missing field does not evaporate the object from the structure, it only removes one filter. Automatic extraction can fill coordinates; a wrong extract parks the object in the wrong cell, people still conclude “none,” now with false certainty. If “unknown” is a legal coordinate, there must be a visible cell such as Uncategorized × Topic; otherwise unknown still equals evaporation. When the matrix is only an expert retrieval tool, not the main browse structure, the completeness bar can drop to “the searcher knows fields may be empty”—and that fact must be disclosed.

Applying it

• Before committing to a matrix, report non-null-and-correct rates for every coordinate axis. Below threshold, do not treat that dimension as an addressable axis; it is a filter at most.
• Objects missing a coordinate either get filled before entering the matrix, or appear only in single-axis browse and search, marked “not in a cross position.”
• Distinguish empty cells that mean zero stock from those that mean not yet catalogued. Do not use the same blank.
• Verify with ten known objects missing one coordinate, looking them up in the matrix with a full two-attribute query. If they cannot be found and are judged nonexistent, the matrix is already lying about incomplete data. Fill the fields or drop the axis; do not first polish the cell page’s layout.

Related

• Within the group: G1.09.1 Hierarchy organizes content by parent–child links and is the usual default · G1.09.2 A matrix locates the same content by several attributes at once · G1.09.3 Hypertext organizes by associative links, not a single hierarchy · G1.09.4 Sequential structure forces a fixed order, as in tutorials and wizards · G1.09.5 Most systems keep hierarchy as the backbone and add hypertext laterals · G1.09.6 Structure type and organization scheme are independent dimensions
• Adjacent: G1.05 Metadata · G1.06 Faceted classification · G3.10 Faceted navigation
• Search terms: matrix structure · metadata completeness · empty facet cell