Cannabis microbial lab variability is the reason you can send what feels like the same batch to two labs and get two different answers. If that has happened to you, you are not alone. In my world at Willow, I see this every week, and most of the time it comes down to three things: how the sample was taken, how it was prepped, and which rules the lab was following that day for that matrix and that state.
In this post, I am going to walk you through what those swings actually mean, how to read a COA when you are parked near a limit, and how to set up sampling and SOPs so you stop getting blindsided. I will keep it practical, because you have product to move and a team to manage.
Why cannabis microbial lab variability shows up so often
In an ideal universe, “TYMC is TYMC” and every lab would measure it the same way. In cannabis, we are not there yet. Different labs use different media, incubation times, temperatures, enrichment steps, and counting rules. So even if two COAs both say TYMC, they may not be perfectly comparable.
One of the biggest drivers is sample preparation. How aggressively a lab homogenizes, what diluent they use, and how they sub-sample can change what ends up on the plate or in the assay. If you want a solid overview of how inter-lab differences happen, including sample prep, I like the way Cannabis Science and Technology frames it for operators who have to make real decisions, not just read theory.
Also, microbes are not sprinkled evenly across real material. Flower and trim can have hotspots. Kief can clump. Even some concentrates behave like they have little “pockets” where organisms hang out. So yes, sometimes you are seeing true product heterogeneity layered on top of method variation.
Microbial test result variation often starts with your sampling, not the lab
I am going to say this plainly: sampling is on you, not the lab. You can run a tight cultivation and post-harvest operation and still get a weird COA if the sample was not representative. When your sampling plan is shaky, you are basically asking the lab to guess what your lot looks like.
Here are the sampling habits that most often create microbial test result variation:
- Convenience grabs from the top of a bin or the “prettier” portion of a tote.
- Not enough increments for big lots, especially after multiple handling steps.
- Inconsistent grind or mixing before you split into jars, so one container is microbe-heavy and the other is not.
- Time and temperature drift during transport that changes viability and recovery.
If you are comparing labs, do yourself a favor and control what you can. Build a composite, homogenize it consistently, split it into two sterile containers at the same moment, label clearly, and ship under the same conditions. You will not remove lab-to-lab variation, but you will reduce the noise you introduced.
Method and matrix fit: the quiet driver of cannabis microbial lab variability
One detail that gets skipped in a lot of conversations is matrix fit. A method that performs fine on dry flower may behave differently on infused pre-roll material, gummies, beverages, or sticky concentrates. Each of those is a different matrix, and the matrix can suppress recovery, trap organisms, or interfere with extraction and plating.
If you want to go deeper on the broader challenge here, including how cannabis methods and regulations have evolved unevenly, Medicinal Genomics has useful educational resources that explain why the same target can look different depending on the workflow and the product type.
Examples you have probably lived through:
- Sticky concentrates do not always disperse evenly, so two sub-samples can act like two different products.
- High-sugar edibles can stress organisms and reduce recovery compared to flower.
- Infused flower can change extraction efficiency because oils behave differently in diluents and plating steps.
When you get a surprising pass or fail, ask one simple question first: Was this method validated for this exact matrix? If nobody can answer clearly, interpret the result with caution and move into a structured investigation instead of panicking.
State rules can create “variability” even if the lab is perfect
Sometimes the lab did nothing wrong and you are still confused. That is because different states ask different questions. One market may require absence of specific pathogens. Another may focus on indicator counts like TYMC or TAMC with allowable limits. Those COAs are not interchangeable, even if the product is the same.
If you operate in more than one state, you will save yourself a lot of whiplash by doing one of these two things:
- Build internal specs to the strictest market you serve, then treat everything else as easier by comparison.
- Segregate product streams by destination, so you are not trying to make one lot satisfy conflicting frameworks.
What “repeatability” really means in cannabis testing
People hear “repeat test” and assume the number should be identical. Microbial enumeration does not work that way. Recovery, growth conditions, counting technique, and basic randomness all play a role. Cannabis testing repeatability is best thought of as a range, not a single magic value.
I separate variation into two buckets when I talk with customers:
- Explainable variation: sampling differences, sample prep steps, method choice, matrix effects.
- Built-in measurement noise: what happens when you are near the limit of detection or near a compliance threshold.
If your result is comfortably below the limit, little swings do not change decisions. If your result is hugging the line, tiny swings can flip pass to fail. That is where you need a repeatable decision process inside your QA system, not a gut call on release day.
TYMC: why it gets side-eye and why you still should care
Total yeast and mold count is popular because it is accessible and gives a broad signal. The downside is that TYMC does not tell you which organisms you have. A high number could be mostly benign environmental species. A low number does not automatically guarantee there is no targeted pathogen risk.
Here is how I usually frame it with operators:
- The case for TYMC: it is a practical indicator that can flag drying, storage, sanitation, and handling problems early.
- The case against TYMC alone: it is not specific, and it does not map cleanly to consumer risk in every product category.
Most strong programs do not pick one. They combine indicator counts, targeted assays when required, and process controls that keep you away from the edge in the first place.
When a “fail” is likely real vs when it might be a testing artifact
Borderline results create the most chaos. I have seen plenty of lots that pass once, fail once, then pass again, and everybody wastes a week arguing about whose number is “true.” If you want a quick overview of why microbial failures are such a common pain point in cannabis, including how method and sampling differences contribute, BoxPurify’s blog has a readable summary that matches what I see in the field.
Here is the tiered approach I recommend so you do not overreact or underreact:
- Way over the limit or pathogen detected: treat it as a high-confidence failure. Quarantine, stop release, start the investigation and disposition path.
- Slightly over the limit: assume it could be real, but verify method fit, sample integrity, and representativeness before you decide on big moves.
- Passing but close: treat it like a warning shot. That is where internal action limits protect you.
If you are constantly living near the compliance line, you are going to keep feeling this pain. The only real fix is building more margin.
What to ask the lab when results do not match
If Lab A says pass and Lab B says fail, do not stop at “who is better.” Ask the questions that uncover what changed. I tell teams to keep this list in their QA playbook so it is not improvised under pressure:
- What exact method was used? Name and version, and whether it is culture-based, qPCR-based, or a hybrid workflow.
- Was it validated for your matrix? “Flower” is not the same as infused ground flower or live resin.
- How was the sample prepared? Homogenization, diluents, enrichment conditions, incubation windows, plating rules.
- What is the reporting basis? Wet weight vs dry weight conversions can matter depending on the product.
- What QC ran with the batch? Blanks, positives, and basic run performance checks.
Transparent labs can answer these without getting defensive. And when you document the answers, you have something solid for auditors and regulators if a decision gets questioned later.
How you reduce cannabis microbial lab variability inside your operation
You cannot control every lab variable, but you can control the inputs that make your results swing. The biggest mindset shift is this: testing is verification, not the control. Your controls are your SOPs, your environment, your handling, and any validated microbial reduction step you use.
Steps that help fast:
- Set internal action limits below state thresholds so you act before you are in the coin-flip zone.
- Standardize sampling with a written plan, training, consistent tools, and a composite approach for larger lots.
- Trend results by room, cultivar, week, drying parameters, and packaging style, not just pass or fail.
- Manage moisture and water activity because they are leading indicators for fungal growth risk during storage and transport.
If you need a validated way to create more margin, you can evaluate ozone-based decontamination as part of a broader program. Our WillowPure Decontamination Systems page lays out how we think about applying a kill step without turning your quality system into a guessing game. I am not saying “buy a machine and hope.” I am saying pick controls you can document, defend, and repeat.
Retesting without panic (and without looking like lab shopping)
Retesting can be the right move. It can also become a mess if you do it ad hoc. The goal is to treat retesting as an investigation step with pre-set criteria, not as a way to chase a pass.
I recommend a simple decision tree with three gates:
- Regulatory gate: confirm what your state allows for retesting, resampling, and composite sampling.
- Risk gate: separate indicator organisms from pathogens, and escalate faster for inhalables.
- Evidence gate: chain of custody, sample integrity, method fit, and how close you are to the threshold.
And please, keep your documentation tight. “We passed the second time” is not a quality record. If you want help building audit-ready workflows around remediation and reprocessing, start with Willow Scientific Consulting and we will map it to your products and your states. Also, yes, I know it is easy to miss a detail when you are juggling releases. That is why you want a checklist in advance, not after you are already in triage mode.
Use your COA like a business tool, not just a compliance checkbox
The teams that struggle the most with cannabis microbial lab variability usually treat the COA like the finish line. The teams that struggle less use it like feedback. They look for patterns, seasonality, and SKUs that are always “close,” then fix the process upstream so they are not depending on a borderline number.
If you are trying to justify prevention work to leadership, connect it to real costs:
- Hold time and delayed revenue
- Rework labor and re-packaging waste
- Retest fees and shipping
- Lost sell-through windows when the market moves faster than your QA cycle
If you want a quick way to model that impact, our Decontamination ROI Calculator can help you put numbers to the “hidden” cost of variability. Use it as a starting point for the conversation, not as a promise of outcomes.
FAQ: Cannabis microbial lab variability and how to read the swings
Why do two labs give different microbial results on the same batch?
Usually because the sample was not identical, the prep steps differed, or the methods are not apples-to-apples for that matrix. In other words, it is often process variation, not your product morphing in transit.
What does repeatability mean in cannabis microbial testing?
It means results should land within a realistic range when the same method is applied consistently. It does not mean you will get the exact same number every time, especially near detection limits or regulatory thresholds.
If my TYMC is barely failing, should I assume the product is unsafe?
Not automatically. You still need to quarantine and follow your state rules, but a borderline TYMC fail is also a cue to review sampling representativeness, method fit, and handling conditions before you make a worst-case call.
How do I judge whether a retest is legitimate?
Start with what the regulations allow, then document why you are retesting. You should be able to show chain of custody, sample integrity, and a rational reason the first result might not reflect the lot. If you cannot explain it, do not do it.
How do I stop living so close to the limit?
Improve sampling, set internal action limits, trend data, control moisture and water activity, and use validated process controls to create more margin. The goal is fewer “coin flip” COAs and more first-time passes. I have seen this work when teams stick with it, even if it takes a few cycles to dial in.
Conclusion: Turn cannabis microbial lab variability into clearer decisions
Cannabis microbial lab variability is real, but it does not have to run your schedule. When you understand where the swings come from, you can read COAs with a calmer head, respond to borderline results without overcorrecting, and build SOPs that create breathing room.
If you want a second set of eyes on your sampling plan, retest logic, or microbial prevention program, reach out through Willow Industries. I would rather help you prevent the next surprise than help you explain one after the fact.