Case study · Early-stage cannabis testing labs

A day of validation analysis, cut to about an hour

New labs burning runway needed validated methods for every required test. We supplied them, automated the analysis and targeted the retests, so reports went out sooner.

The situation

Before a new testing lab can sell a single test, it has to validate the method behind every test on its license. Until then, it is spending runway: rent on the building, salaries for the team, and time with investors, all before the first paid sample.

A lab in that position can’t afford to wait on its own validation analysis.

The problem

Validation is judged by the numbers. In nearly every jurisdiction, a method passes when its relative standard deviation (RSD) falls within the acceptance ranges set by AOAC’s Standard Method Performance Requirements (SMPRs). The lab also has to establish its limits of detection and quantitation (LOD and LOQ). Done by hand, that analysis could take a full day or more for each round.

A lab without a validation method in place has a bigger problem. Before any of that math, it needs a method that satisfies the state’s requirements, and building one from scratch can stall the launch.

What we did

We gave each lab a method for the full suite of tests it needed to run. Then our automation software processed the validation data and showed which analytes were out of spec, and by how much.

That turned a failed run into a targeted retest: only the analytes that missed, requested as soon as the numbers came in. Faster retest requests meant less of the downtime a lab spends before it can sell tests.

The result

Validation reports went out as quickly as the labs could run their samples. No one waited on days of analysis, and no one had to invent a method from the state’s requirements. Analyses that could take a full day or more took roughly an hour.

What it means for your lab

If you’re building a lab, the clock started before your first sample. The building, the payroll and the investor updates don’t wait for validation. Three things decide how long you wait:

Methods. If you don’t have a validated method for every test on your license, start there. We bring methods for the full suite, built to your state’s requirements, so you aren’t writing one while the rent runs.
The math. Your validations will be judged on RSD against the SMPR acceptance ranges, plus LOD and LOQ. Automated, that analysis takes about an hour instead of a day or more.
Retests. When a run fails, you should know which analytes missed, and by how much, in time to retest only those.

That is the core of our Instruments & methods and Lab data & LIMS work: a faster path to the validation report, and to your first paid sample.

Written from our own account of the work. The client is anonymous.

Try itAn illustrative tool with sample data

Potency run · HPLC-UV · 228 nm

Elution order

Twelve cannabinoids, in the order a C18 column separates them. Select one to read what its peak tells a lab, or read all twelve →

CBDVA cannabidivarinic acid · 1.88 min · not detected

First cannabinoid off the column. Usually absent in THC flower. When it is present, its peak has to clear the solvent front before anyone can quantify it.

CBDV cannabidivarin · 2.09 min · not detected

The neutral form of CBDVA. Not detected here, which is normal for THC-dominant flower.

CBDA cannabidiolic acid · 2.59 min · quantified

The acid form of CBD, so total CBD = CBD + 0.877 × CBDA. A small peak in THC flower; on hemp it is the headline.

CBGA cannabigerolic acid · 2.75 min · quantified

The precursor the plant makes THCA, CBDA and CBCA from. Often the second-largest peak in THC flower.

CBG cannabigerol · 2.91 min · quantified

CBGA after decarboxylation. CBDA, CBGA, CBG and CBD elute within half a minute of each other, so a small drift in the mobile phase can merge them.

CBD cannabidiol · 3.08 min · below LOQ

Near zero in THC-dominant flower. The peak is there, but its signal-to-noise is below the limit of quantitation, so the certificate reports <LOQ.

THCV tetrahydrocannabivarin · 3.39 min · below LOQ

The propyl cousin of Δ9‑THC. Its shorter side chain holds less tightly to C18, so it elutes more than two minutes before Δ9‑THC.

CBN cannabinol · 4.61 min · quantified

Forms as THC oxidizes. A growing CBN peak says the sample is old, or was stored warm or in light.

Δ9-THC delta-9-tetrahydrocannabinol · 5.82 min · quantified

The regulated compound. Fresh flower carries little; most of it forms when THCA is heated. Δ8‑THC elutes 0.18 min later, and the method has to keep the two apart.

Δ8-THC delta-8-tetrahydrocannabinol · 6.00 min · not detected

Not detected here. Δ8 and Δ9 differ only in where one double bond sits. That makes them a critical pair: a method that cannot resolve them reports one as the other.

CBC cannabichromene · 7.26 min · quantified

Minor in most flower. It elutes 0.35 min before THCA, the largest peak in the run, so an overloaded THCA peak can swallow it.

THCA tetrahydrocannabinolic acid · 7.61 min · quantified

The dominant peak in THC flower. Heat turns it into Δ9‑THC, so labs report total THC = Δ9‑THC + 0.877 × THCA. The 0.877 is the ratio of their molecular weights (314.5 / 358.5).

Sample data: an illustrative chromatogram, not client dataIllustrative, not sample data · C18 · isocratic · UV 228 nm

Where the numbers start.

For LC and GC methods, the validation statistics start as peak areas on a chromatogram like this one. The trace is sample data.

See all the tools

Next caseAll case studies are on the Work page

Next case study

No subscriptionCOAs from Google Sheets and Apps Script, with an editable template
A cannabis testing lab in Akwesasne
Certificates of analysis without a LIMS subscription
A lab in Akwesasne paid for a LIMS only to issue COAs. We moved it to Google Sheets and Apps Script: enter data once, get a COA from a template it can edit.
Lab data & LIMS
See the COA generator

See all the case studies