Methodology

The source

The underlying data is Nevada's package-level cannabis compliance testing record, delivered as 150 biweekly export files covering Jan 2020 - Feb 2026. Each row is a single analyte result — a cannabinoid, microbial, pesticide, heavy-metal, residual-solvent, mycotoxin, or terpene test — for a single tested package. A typical package produces dozens of analyte rows. Producers ("Packaged By Facility") and testing labs are named in cleartext exactly as they appear in the state record — there is no anonymization. The raw files total roughly 8.9 GB in UTF-16 text.

What we counted — and what we didn't

The headline universe is 513,539 completed package tests from 237 named producers across 10 named labs, Jan 2020 - Feb 2026. Reaching that number required several deliberate steps:

• Completed tests only. Records still marked unfinished are excluded from every figure. An unfinished test is not a result.
• Deduplicated to the package level. The source carries one row per analyte, so a single package appears dozens of times. We collapse to one row per package on its unique lab-sample ID, keeping the first-seen record by test date, and read package-level fields (overall pass/fail, producer, lab, category, remediation flag) from that record.
• Failing category is mapped, not invented. Nevada records the specific analyte that failed. We group those analytes into standard panels — Microbials, Pesticides, Heavy Metals, Residual Solvents, Mycotoxins, Cannabinoids, Terpenes, Water Activity/Moisture, and Visual/Foreign Matter — so the "what actually fails" view is drawn from the named failing test, not estimated from the panel.
• Encoding handled per file. Every file is detected and decoded individually (UTF-16, tab-separated) so no rows are silently dropped at the import step.

The entry-cohort measure

The market-concentration section groups every producer by the year it first appears in the testing record, and tracks what share of each period's tested volume each cohort holds. We report the incumbent (first-window) share per quarter, not as a cumulative share since the data began — a cumulative figure would flatter early producers simply because they have been testing longer.

One important caveat on the word "incumbent." Entry here means a producer's first appearance in this record, and the record opens in January 2020. Nevada's adult-use market began retail sales in mid-2017. So the 2020 cohort bundles every producer already operating when the data window opens — it is a first-window cohort, not a true day-one license class. The decline in its share over time (from 100% to roughly 54.4%) is a real signal of new entry. The 100% value at the left edge of the chart is an artifact of where the data starts, and we say so on the page.

Two concentration indices (HHI)

We report two Herfindahl-Hirschman Indices because they point in opposite directions and that contrast is the finding. Producer concentration is 172 — far below the Justice Department's 1,500 "unconcentrated" line, reflecting a genuinely open growing tier of 237 producers. Lab concentration is 1344 — closing on that same 1,500 line from below, with three of ten labs handling 52.6% of all volume. A market can be competitive where you plant and concentrated where you're graded. We present both numbers rather than a single headline index.

Why there is no potency section

The New Jersey and Illinois dashboards carry a potency analysis — THC distributions, lab-by-lab medians, potency inflation. Nevada's primary export does not carry a usable numeric potency value across the full time window. Cannabinoids such as THCA and Delta-9 THC appear in the record as pass/fail test types, not as reported percentages, for most of the period. A numeric result column appears in some of the later files but not the earlier ones, so a consistent six-year potency series cannot be built from this export without misrepresenting the gaps. Rather than fake a potency section or stitch together an inconsistent one, we omit it. If a complete potency field becomes available, it can be added later. What this record supports cleanly — lab concentration, pass/fail divergence, failing-category breakdown, remediation, and producer entry cohorts — is what the dashboard reports.

Known caveats and open questions

We would rather flag these ourselves than have you find them:

• The 2025–2026 volume decline is provisional and likely incomplete. Tested-package counts fall sharply through 2025. This is not a market-size story — Nevada retail sales fell only about 8.6% over the same period. The decline is concentrated in a small number of labs that stop appearing in the state's public testing record in mid-2025, including the single highest-volume lab, while the other labs keep operating and do not absorb the lost volume. We are verifying the reasons with the state and are not characterizing them here until that is confirmed. Treat 2025–2026 volume as incomplete. Importantly, this does not affect the lab-concentration, market-share, entry-cohort, or pass-rate findings, which are measured within each period from the records that are present.
• The pass-rate climb deserves scrutiny. The program-wide pass rate rises from 89.45% in 2020 to 98.58% in 2026. Part of that is a maturing market; part of it may reflect testing behavior. We present the trend and the lab-by-lab spread side by side rather than asserting a single cause.
• Lab pass-rate comparisons need context. Labs do not test the same population of producers or the same product mix, so a raw pass-rate gap between labs is a starting point for inquiry, not a verdict on any lab.
• Remediation rate is reported, not interpreted. The share of a lab's or producer's packages carrying remediated product is shown as recorded. A higher remediation rate can mean a lab sees more failures that get treated and retested, or that it handles a different product mix. The record shows the number; it does not explain it.
• Producer-level pass rates on tiny volumes are noise. The producer failure chart is limited to producers with at least 20 packages on record, and a producer with one or two packages can show a 0% or 100% rate that means nothing. Read the small producers in the explorer with that in mind.
• "Producer" means the packaging facility. Producers are identified by the "Packaged By Facility" field, which is the entity that packaged the tested product. Names are normalized only for case and whitespace; distinct facilities under common ownership are kept distinct, as the state filed them.

Check our work

The cleaned, deduplicated, package-level dataset behind every chart is available as a gzip-compressed CSV download — one row per package, with producer, lab, date, category, pass/fail, and remediation flag. It opens directly in Excel, pandas, or any unzip tool. If you find an error, we want to know.