Mass Cannabis Data - Lab Performance Trends

Monthly Sample Volume by Lab

Unique samples (METRC IDs) processed per month — 2024 → 2025. Hover any lab in the legend to highlight, or click to toggle.

Market Share Over Time (% of Monthly Samples)

Each lab's share of total samples processed each month — shows market consolidation and shifts.

Average THCA % by Lab (3-Month Rolling Mean)

Smoothed with a 3-month rolling average for readability. A 5.48% gap between the highest (Lab_H @ 27.31%) and lowest (Lab_L @ 21.83%) reporting labs.

Failure Rate by Lab (3-Month Rolling Mean)

% of samples that failed any test in the panel, smoothed with a 3-month rolling average. Months with fewer than 5 samples are excluded for stability. Lab_O, Lab_J excluded — low volume and erratic monthly data that compressed the y-axis for everyone else.

Benford's Law Conformity by Lab (MAD)

First-digit distribution of Yeast & Mold (CFU/g) values — a metric that spans 1 to 1,000,000+. Combined 2024 + 2025 per lab; 14 labs have ≥50 samples. MAD (mean absolute deviation, percentage points across the 9 digits) is the Nigrini-standard fit metric. Thresholds: < 0.6 close · 0.6–1.2 acceptable · 1.2–1.5 marginal · > 1.5 nonconformity. Largest deviation: Lab_E (Nonconformity, MAD=7.39). Closest fit: Lab_A (MAD=0.78).

First-Digit Distributions vs. Benford Expected

Each line is one lab's observed first-digit distribution of Yeast & Mold values. The dashed reference is the Benford expected distribution (30.1%, 17.6%, 12.5%, ..., 4.6%).

Per-Lab Benford Detail Table

Sample size, chi-squared statistic, p-value range, and significance interpretation for each lab. Sorted by chi² descending.

Lab	YM Samples	MAD	χ² (pct)	Conformity	Digit 1 %	Digit 2 %	Digit 3 %	Digit 4 %	Digit 5 %	Digit 6 %	Digit 7 %	Digit 8 %	Digit 9 %

How to read this: Benford's Law predicts the leading digit is "1" ~30% of the time, "2" ~17.6%, and so on for naturally-occurring data spanning many orders of magnitude. MAD (mean absolute deviation in percentage points) is the Nigrini-standard fit metric. χ² uses the same percentage-based formula as the per-year dashboards.

⚠ Rounding drives most of these deviations. Cannabis YM is typically reported to 2–3 significant figures and uses default values for below-detection (often 100) or maxed-out samples. Those practices alone bias the first-digit distribution and trigger "Nonconformity" classifications.

Per-Digit Z-Scores — Which Digits Each Lab Reports More or Less Often

Standardized deviation per lab × digit: z = (observed − expected) ÷ √(n·p·(1−p)). Orange = reported more than Benford predicts; blue = reported less. Intensity scales with |z|. |z| > 1.96 is significant at p<0.05; |z| > 2.58 at p<0.01.

Lab	n	Digit 1	Digit 2	Digit 3	Digit 4	Digit 5	Digit 6	Digit 7	Digit 8	Digit 9

Composite Forensic Score (Nigrini Suite)

Combines five forensic tests into a single 0–100 concern level. Higher scores = more flags. Each test contributes up to 20 points based on cannabis-tuned thresholds. Most concerning: Lab_E (98.2/100). Cleanest: Lab_A (42.2/100).

⚠ Use this for relative ranking, not absolute judgment. Most of what these tests flag is routine rounding and default-value reporting (typical lab practice for cannabis YM). A lab at 98 deviates from expected patterns much more than one at 42; that comparison is meaningful, but the absolute score isn't.

Lab	Total / 100	1st Digit	2nd Digit	NDT	Threshold	L2 Digits

Number Duplication Test (NDT)

Flags specific values that repeat far more than expected for continuous data. Concentration = % of nonzero samples sharing a value with at least one other (well-behaved data is near 0%). Most concentrated: Lab_K at 98.6% — top value 100 CFU/g (11.6% of samples).

⚠ This is largely lab reporting practice. 100 is typically a below-detection default, 10,000 is the MA regulatory threshold, and 100,000 is often an upper-detection ceiling. Round numbers accumulate because results are reported to 2–3 significant figures. What's most informative is how aggressively a lab repeats one value — e.g., 22% of samples on a single number is far more notable than spreading repeats across many round values.

Lab	Nonzero samples	Unique values	Concentration	Top value	Top value count	Top value share

Threshold Density — Piling at the 10,000 CFU/g Limit

Compares values just below the MA regulatory limit (5,000–9,999) to values just above (10,000–19,999). The below ratio = below ÷ (below + above). Random data near the threshold should be ~50%. Most extreme: Lab_E at 99.4%.

⚠ Two innocent factors inflate this ratio. First, most cannabis genuinely tests under the limit (biological reality). Second, scientific rounding of borderline values (9,300 → 9,000) lands them below the threshold. A 70–80% below-ratio is plausibly clean. Ratios >90% paired with NDT concentration on values like 9,000 or 9,500 are the more notable signal.

Lab	Below 10K	Above 10K	Window total	Below ratio	Window share

Second-Digit Benford Test

Second digits follow a flatter distribution than first digits (12.0%, 11.4%, 10.9%, ..., 8.5%). Same Nigrini MAD thresholds apply. Worst: Lab_N (MAD=8.34). Best: Lab_A (MAD=0.30).

⚠ Rounding inflates digit "0" here. Values like 100, 1,000, 10,000 force the second digit to 0, pulling the distribution away from the expected 12.0% / 11.4% / 10.9% curve. A lab failing this test that also has high NDT concentration is almost certainly being flagged for rounding behavior alone.

Lab	Samples	MAD	χ² (pct)	Conformity

Last-Two-Digits Uniformity Test

For values ≥ 10 in continuous data, last-two-digit pairs should be uniformly distributed (each ~1%). High chi-squared values indicate piling on round numbers. df=99, critical values: 123.2 (p<0.05), 134.6 (p<0.01), 148.2 (p<0.001). Most concentrated: Lab_M (χ²=373122.5, p < 0.001 (extreme)).

⚠ Read this as a rounding-intensity meter. Reporting to 2–3 significant figures means most values end in "00" — which is enough to drive chi² into the tens of thousands by itself (note the log scale). Labs that pass simply report more granular values; this is a reporting-style measure, not an accuracy measure.

Lab	Samples	χ² (df=99)	Significance	Top digit pair	2nd	3rd

Per-Lab Summary (2024 + 2025)

Total samples, share of market each year, average potency, overall pass rate, and volume trend (last 3 active months vs first 3).

Lab	Total Samples	2024 Samples	2025 Samples	2024 Share	2025 Share	Avg THCA %	Pass Rate	YoY Volume (06–11)