Measures of the data, not from it

How well does Canada measure this?

Every subject below is one this country collects data about. They are not collected at anything like the same resolution, and the difference is not a fact about how much each subject matters. It is a record of instruments somebody chose to build.

The widest subject here holds 962,280 observation cells. The narrowest holds 360, a factor of 2,673. Both are real subjects with real consequences for real people.

The useful question

Not how much, but which dimension is thin

A subject can hold a great many observations and still answer almost nothing, if all of them come from one province or one year. So the measurement here is four numbers rather than one, and the column that matters most is the last: the dimension that runs out first is the one that decides what the subject can be asked.

Places: 8 of 15 subjects

The subject is instrumented, but in a small number of provinces or territories. A national question about it cannot be answered from these sources, however many years or measures exist.

Education, Unemployment and job security, Indigenous ancestry, Geography, Income and income distribution, Employment and working conditions, Early child development, Disability

Time: 6 of 15 subjects

The subject is measured broadly but for few periods. Levels can be compared across places; change over time cannot, and a trend claim from this many periods would be an assertion.

Health services, Housing, Race, Food insecurity, Social exclusion, Immigration

Measures: 1 of 15 subjects

The subject is covered widely and deeply in time, but through a small number of distinct measures. It is watched rather than examined: the same few questions asked repeatedly.

Social safety net

Every subject

The instrument table

Every determinant in this warehouse, by how finely it is instrumented. Bars share a scale down each column, so two rows can be compared. Sorted by total cells.
Subject	Cells	Measures	Places	Periods	Answerable	Binding constraint
Education	962,280	2,430	11	36	89.5%	Places
Health services	932,771	1,315	110	75	87.0%	Time
Unemployment and job security	917,891	1,600	14	656	84.0%	Places
Indigenous ancestry	466,871	18,633	16	38	44.6%	Places
Geography	302,610	417	14	55	98.4%	Places
Income and income distribution	291,924	102	16	49	75.3%	Places
Employment and working conditions	172,260	540	11	29	99.7%	Places
Housing	110,466	5,814	16	4	80.0%	Time
Race	26,165	496	16	4	84.3%	Time
Early child development	14,342	57	14	47	94.2%	Places
Social safety net	6,773	13	14	353	97.7%	Measures
Food insecurity	5,280	240	16	7	71.7%	Time
Social exclusion	3,186	54	13	11	92.5%	Time
Immigration	1,460	60	14	11	100.0%	Time
Disability	360	360	1	1	50.0%	Places

whomatters.ca intensity-by-determinant whomatters warehouse, dataset cells=8687740;computed=2026-08-14, computed 2026-08-14 Instrumentation, not outcomes. A larger number means the subject is more finely observed and nothing more.

How this was measured

A cell is one measure, one place, one period, one breakdown, one status. Cells are counted without filtering by status, so a question the publisher refused to answer still counts as a question asked; that is what makes the answerable column meaningful rather than circular.

Places counts distinct jurisdictions present, which includes regional and sub-provincial codes where a source publishes them, so a number above fourteen is not an error. Periods counts distinct periods as the publisher labels them, so a monthly source and an annual one are not on the same footing and the column should be read within a row rather than across the table.

The binding constraint is simply the smallest of measures, places and periods. It is arithmetic, not judgment.

Every row of this table, and the per-source version of it that is too long to print, is published as metrology.json. Nothing on this page is computed in the page; it is all read from that file.