coefficient of variation

The coefficient of variation expresses sampling error as a percentage of the estimate itself, which makes reliability comparable across cells of very different size. A CV of 2.3 percent means the sampling uncertainty around the published figure is small relative to its magnitude; a CV near 30 percent means the cell is mostly noise.

It appears wherever estimates come from samples rather than counts. In the cataloged U.S. Census Monthly Retail Trade Survey (MRTS/MARTS) record for electronics and appliance stores, reliability coefficients ship alongside the sales workbook: "adjustment and reliability text files under 20 KB each", including the cited "2.3 percent preliminary CV for NAICS 443".

Scope discipline matters: survey-derived cells carry CVs, while benchmark census counts do not need the caveat. The MRTS sales workbook itself is modest - mrtssales92-present.xlsx about 0.44 MB across 35 annual sheets reaching back to January 1992 - so the reliability files are proportionate companions, not an afterthought.

Advance retail prints move markets, and the CV is the honesty layer on top of them. Two estimates of similar size are not equally trustworthy if one carries a 2 percent CV and the other a 25 percent CV; treating them identically produces false precision in dashboards and forecasts.

For NAICS 443 specifically, the 2.3 percent preliminary figure tells electronics-retail analysts their month-one reading is usable for trend work but will be revised as the survey matures. Teams that ignore the flag rebuild charts twice and lose credibility with stakeholders.

There is a selection consequence too: a source that omits reliability metadata forces you to guess which cells are safe to quote. The cataloged record ships the coefficients as plain text files under 20 KB - trivially easy to distribute, which is why their absence elsewhere is a choice, not a constraint.

Apply these checks:

1. Ask whether reliability files ship with the release. The cataloged MRTS/MARTS record publishes them every month next to the sales workbook - make that your baseline requirement.
2. Read the CV before quoting the estimate. A preliminary 2.3 percent CV supports trend commentary; double-digit values warrant hedging or suppression.
3. Track revisions against the advance print. MARTS figures arrive roughly two weeks after month close and revise in the MRTS release about two weeks later.

Advance estimate is the earliest MARTS print whose reliability the CV describes.

Retail inventories data ship as the companion inventory-to-sales workbook matched to the same kind-of-business rows.

County Business Patterns provides the annual establishment counts that benchmark the monthly survey frame.

Industries where the catalog applies this term: computer-electronics-retail data and housewares-specialties data.