Lab Reporting & Digit Tests

How each Massachusetts testing lab reports yeast & mould counts, and what digit tests can and cannot tell you about that. Each year stands entirely on its own.

Why this page never pools the two years

The CCC anonymises labs as Lab_A…Lab_P and publishes each year separately. An earlier version of this page pooled 2024 and 2025 per lab, which assumes a given code means the same facility in both years.

We tested that assumption. Scoring all 3,628,800 possible pairings of the ten labs present in both years on reporting-precision fingerprints, the identity pairing ranks 12th: far better than chance, but not first. Every pairing that beats it swaps labs within a reporting-precision class, and none crosses classes. Only Lab_B and Lab_E are pinned; the other eight are mutually indistinguishable within their group.

So cross-year per-lab attribution is not defensible, and nothing here attempts it. The CCC does not publish the mapping. Every figure below is computed inside a single year. A result for Lab_D in 2025 stands regardless of what Lab_D meant in 2024.

Year context
Sample counts, active labs, and market concentration. Each computed within its own year, so these are comparable as market statistics even though individual lab codes are not.
How granular is each lab's reporting?
The share of a lab's yeast & mould results that are distinct values. A lab emitting raw plate counts lands high. A lab collapsing everything onto round numbers with a floor lands low. This is the most useful thing on the page: it tells you how much resolution a given lab's numbers actually carry, and it varies enormously. Bars in green clear the 30% threshold we require before reading anything into digit tests; grey bars do not.
Concentration at the detection floor
Share of results reported at exactly the lab's lowest observed value. A high bar means the lab is reporting a censored “below detection limit” result as a number. Those are not measurements, and they should not be averaged as though they were.

Read the digit tests below with care

Benford's law assumes values that vary continuously across orders of magnitude. Several labs here do not produce data like that: they report yeast & mould on a coarse grid (100, 200, 1000), where 100 almost certainly encodes “<100” and 100001 encodes “>100000”.

On that kind of data a Benford deviation measures rounding granularity, not anomaly. A lab that rounds heavily will fail the test no matter how honest its underlying measurements are. We therefore only chart labs clearing 30% distinct values, and even then the result is a data-quality observation, not an allegation.

First-digit deviation (Benford MAD)
Mean absolute deviation between the observed first-digit distribution and Benford's expectation, for labs with granular enough data. Nigrini's conformity bands are marked: below 0.006 close, 0.012 acceptable, 0.015 marginal.
Density either side of the 10,000 CFU/g mark
Count of results in the 10% band below 10,000 divided by the count in the 10% band above it. Some ratio above 1 is expected simply because the underlying distribution is falling by that point, so treat this as a prompt to look closer rather than a finding on its own.
Per-lab detail
One table per year. No row in either table is joined to the other.