Methodology

Throughout this dashboard, "volume" means the number of testing packages, not weight. In New Jersey, one production batch is tested as one sample.

The batch cap was 100 pounds for the vast majority of the data window.

This means a microbusiness cultivator processing a small harvest registers the same single test as a multi-state operator processing a full batch. Market-share figures based on test counts therefore understate the dominance of larger operators by weight. The figures here reflect testing activity. Pounds-to-market interpretation requires the underlying batch weights, which are not in this dataset.

The underlying data comes from New Jersey's Cannabis Regulatory Commission via Metrc, the state's seed-to-sale tracking system. It was made public in March 2026. The dataset consists of 13 quarterly CSV exports covering January 2023 through February 2026.

The raw dataset contains approximately 7.5 million analyte-level rows. Each row represents a single test measurement on a single analyte for a single package. A package of flower, for example, might generate 30+ rows: one for each pesticide screened, each heavy metal, each microbial test, plus potency.

The dashboard covers January 2023 through December 2025 — 12 complete quarters. The raw files extend into February 2026, but Q1 2026 data is excluded: two months of a quarter measured against twelve full quarters distorts per-period comparisons.

This dashboard exists because New Jersey's Cannabis Regulatory Commission chose to make its testing data public — fully, and without being forced to. It is one of the most thorough cannabis testing datasets any state has released.

Public data like this is how a market improves: regulators, operators, and patients can all argue from the same facts. We're grateful the CRC put it in the open, and we hope they keep doing it.

Every non-stability, non-informational test result in NJ's regulated cannabis market. The dashboard covers packages, source facilities (identified by Metrc-assigned numeric IDs), testing labs (released only by code, shown here as A through H), strain names, test types, pass/fail outcomes, and remediation flags.

After filtering, the working dataset contains 5,939,950 analyte-level rows representing 72,098 unique packages across 133 facilities and 8 labs.

Four categories of data are excluded from the main dashboard:

A package passes if every non-informational, non-stability analyte test for that package has TestPassed = TRUE. One failure on any analyte means the package fails.

This is a strict definition, and it produces a different number than the analyte-level pass rate. A single failed microbial test flags the entire package. The package-level pass rate is 98.0%: 1,411 packages failed out of 72,098. That means roughly 1 in 50 batches failed at least one compliance check over three years.

Both numbers are real. Neither alone tells the whole story. I use the package-level rate throughout the dashboard because that is what determines whether product reaches the shelf.

Remediation data comes from the ContainsRemediatedProduct flag in Metrc. I classify remediated packages into two groups based on whether the package has a failed test on file:

Two interpretations are consistent with the data: that products underwent a remediation process before testing, or that packages with initial failures were remediated and re-tested. The Metrc RemediatedProduct flag does not distinguish between these. Of 573 remediated packages in the dataset, 570 had no failed test on file. New Jersey has no consumer-facing remediation labeling requirement.

Potency analysis on this dashboard covers flower only. The reason: manufactured products (edibles, concentrates, tinctures) report THC in mixed units. The raw data field strips the unit designation, so "Total THC" can mean percentage for flower or mg/g or mg/package for edibles. Mixing percentage and milligram values produces meaningless statistics. Flower and shake/trim are uniformly reported in percentage, so those numbers are clean.

The filter chain for flower potency:

• Product category: ProductCategoryName containing "Bud" or "Flower"
• Test type: "Total THC" containing "(%)"
• Value range: greater than 0, less than 50 (anomaly removal)

Median, mean, and percentiles are computed per quarter, per lab, and per facility. I use the median as the primary trend metric because THC distributions are not perfectly normal and the median is less sensitive to the tail anomalies that exist in this dataset.

The State released these testing records without company or brand names. Facilities appear by Metrc's system-assigned numeric IDs, and labs by code — shown here as Lab A through Lab H. Cannabis Wise Guys did not redact, replace, or withhold any real-world identities; the data arrived from the State already coded.

Some strain names in the underlying data contain brand references, and a lab or facility may be inferable by someone with industry knowledge who cross-references public licensing and Metrc records. The coding limits casual identification but does not prevent determined cross-referencing — and the underlying records are public.

The dashboard shows what the data says about testing patterns, not who specifically is behind each number.

Three terms describe a facility's status in any given quarter:

Facilities are grouped by the quarter they first submitted test results. The Q1 2023 cohort is the original group of operators who were testing when the dataset begins. Each subsequent quarter's new entrants form their own cohort.

Market share is computed as each cohort's package count divided by total packages for each quarter. This shows how incumbent operators maintain, grow, or lose volume share as new entrants arrive. The Q1 2023 cohort started at 100% market share by definition. Three years later, their share has declined as new operators entered, but their absolute volume has increased. They are producing more product, not less. The share decline is compositional.

This data is only as accurate as the entries in Metrc. Operator input errors exist. Lab input errors exist. Lab G's 299,838% THC reading proves that Metrc's validation layer does not catch unit-entry errors.

The remediation flag is operator-triggered. If an operator irradiates flower before submitting it for testing and does not flag it as remediated in Metrc, that treatment is invisible in this dataset. The total volume of pre-test treatment may be undercounted.

Facility IDs may represent multiple licenses under one corporate entity. A company operating two separate cultivation facilities would appear as two separate facility IDs. The data cannot tell you about corporate consolidation without the entity mapping, which I maintain separately and do not publish here.

The dataset covers January 2023 through December 2025. Market conditions, lab market share, and facility activity may have changed since the last data point. I update the dashboard when new quarterly data is received from the CRC.

The potency trend (median flower THC rising from 20.84% to 26.03% over three years) is statistically validated, but this data alone cannot distinguish between three possible drivers: genuine genetics improvement, selection bias in what gets submitted for testing, or lab-side potency inflation. The lab divergence data is consistent with the third driver but does not isolate it from the other two. Published research (Schwabe et al. 2023) documents labeled-vs-actual THC potency gaps on retail flower in Colorado, finding consumer products tested 23–36% lower than their labels claimed. This dashboard does not measure label accuracy — it surfaces inter-lab variance in reported potency for the same regulated market, which is one of several drivers researchers have proposed for the broader inflation trend. See also Scientific Reports 2025 on labeled-vs-actual potency accuracy across flower and concentrate products.

Four CSV files are available for independent analysis:

• nj_packages.csv — one row per package: PackageLabel, FacilityID (Metrc-assigned numeric ID), FacilityType (Cultivator/Manufacturer/Other), Lab (state code, shown as A–H), ProductCategory, Quarter, StrainName, Passed (TRUE/FALSE at the package level), Remediated (TRUE/FALSE).
• nj_facilities.csv — one row per facility: aggregate package counts, pass/fail counts and rate, active quarters, strain count, primary lab, median THC, first and last test dates, remediated package count.
• nj_labs.csv — one row per lab: total package count, pass and fail counts, failure rate percentage.
• nj_strains.csv — one row per strain with 2 or more packages: total packages, facility count, median THC, first and last active quarter.

The same filtering rules apply to all four files: stability tests excluded, informational-only tests excluded, Q1 2026 excluded.

Download all four files from the dashboard download section.

The complete dataset — 13 quarterly Metrc CSV exports, approximately 7.5 million analyte-level rows — is available on request. This is the unmodified data as received from the NJ Cannabis Regulatory Commission. No values have been altered, aggregated, or recomputed.

Every chart on this dashboard was generated from those files by a single processing script. If you want to reproduce any number on this site, you can run the same filters against the same source data.