A strain name does not predict its chemistry. We tested it.
The industry sells cultivar names as if the name told you what is in the jar. We checked that against 4,914 lab-tested samples from 1,950 growers, covering 18 widely grown cultivars. The dominant terpene changes between growers for every single one of them.
What we asked
A knowledgeable operator put it to us plainly: Blue Dream is the same on every certificate, so a published typical profile is good enough to match people on. It is a reasonable thing to believe. It is testable, so we tested it rather than argued about it.
Where the data came from
Smith, Vergara, Keegan and Jikomes published a dataset of 89,923 commercial flower samples analysed by state-licensed testing laboratories in six US states between December 2013 and January 2021. Each row is one lab test, with the cultivar name it was sold under and an identifier for the grower. 43,148 of those samples carry a terpene panel.
We used it because it is per-sample laboratory measurement with a grower attached, at a scale no hand-collected set of certificates could reach.
What we did
- Kept only samples with a terpene panel, then grouped them by cultivar and by grower.
- Took the median of each grower’s samples, so one unusual batch cannot swing a grower’s result.
- Recorded which terpene was highest in each grower’s profile, then counted how often growers of the same cultivar disagreed.
- Kept only cultivars with at least four distinct growers. Every cultivar below has at least 31.
What we found
Agreement is the share of growers whose highest terpene matched the most common one for that name. If names predicted chemistry, these would be near 100%.
| Cultivar | Growers | Samples | Most common | Agreement | Different tops seen |
|---|---|---|---|---|---|
| Original Glue | 352 | 973 | caryophyllene | 78.1% | 6 |
| Blue Dream | 215 | 552 | myrcene | 69.3% | 5 |
| GSC | 212 | 440 | caryophyllene | 74.5% | 6 |
| Sour Diesel | 153 | 346 | caryophyllene | 52.3% | 5 |
| Sunset Sherbert | 127 | 410 | caryophyllene | 74.0% | 5 |
| OG Kush | 108 | 184 | myrcene | 41.7% | 4 |
| Green Crack | 100 | 248 | myrcene | 65.0% | 5 |
| Granddaddy Purple | 85 | 187 | myrcene | 70.6% | 5 |
| Do-Si-Dos | 83 | 179 | limonene | 48.2% | 4 |
| Purple Punch | 75 | 244 | caryophyllene | 58.7% | 6 |
| Tangie | 72 | 207 | myrcene | 66.7% | 5 |
| Gelato | 66 | 114 | caryophyllene | 63.6% | 4 |
| Super Lemon Haze | 65 | 158 | terpinolene | 58.5% | 7 |
| Jack Herer | 60 | 113 | terpinolene | 55.0% | 7 |
| Durban Poison | 60 | 186 | terpinolene | 50.0% | 6 |
| Pineapple Express | 44 | 208 | myrcene | 38.6% | 6 |
| Northern Lights | 42 | 76 | myrcene | 50.0% | 6 |
| Zkittlez | 31 | 89 | caryophyllene | 64.5% | 3 |
Not one cultivar reaches 80% agreement. Blue Dream — the example we were given — is myrcene-topped at 149 of its 215 growers, pinene at 52, terpinolene at 8, caryophyllene at 4 and limonene at 2. Pineapple Express and OG Kush are close to a coin toss. Myrcene in Blue Dream alone spans a fifteenfold range between growers.
Does it change what we would recommend?
That is the part that decides whether any of this matters. We ran each grower’s real measured profile through our own matching engine, holding everything else constant and varying only the terpenes.
The same named cultivar moves between 1st and 24th place for the same person. (Growers per cultivar were capped at 60 for that run, taking the best-evidenced first.) For Durban Poison, growers of one name span 10.4 scoring points on one question, against a library that spans 8.1 points end to end. Twelve of the eighteen cultivars disagree about whether they belong in someone’s top three on five or six of our six questions.
What this means
Two jars, both honestly labelled with the same name, can be different enough that they belong in different recommendations. So we stopped making recommendations from names. Our finder gives you a terpene profile to look for and the words to ask for it, and it names a specific product only where a laboratory has measured that batch.
What would change our mind
Measurement. A certificate with a terpene panel for the batch in front of you settles what that jar contains, and no argument about naming can override it. That is the whole reason we ask shops for their lab data.
Limitations, stated plainly
- Grower identity is anonymised in the public file, so this shows that growers differ, not which grower is which.
- The samples are 2013–2021 licensed-market flower, not 2026 hemp. We know of no reason hemp would be more consistent, but we have not measured that.
- Taking a median per grower smooths variation between that grower’s own batches, which means the true sample-to-sample spread is wider than the table above, not narrower.
Source
Smith CJ, Vergara D, Keegan B, Jikomes N (2022). The phytochemical diversity of commercial Cannabis in the United States. PLoS ONE 17(5): e0267498. doi.org/10.1371/journal.pone.0267498
The dataset is public: github.com/cjsmith015/phytochemical-diversity-cannabis. The authors reached the same conclusion by a different route, writing that while strain names may be a better marker of product chemistry than Indica/Sativa labels, “they are far from ideal”. Our per-grower analysis, the full tables and the engine output are in the project repository so anyone can check them.
Shop by chemistry instead
Five questions, then a card you can hold up at the counter: what to look for, what to say, and what to check on the certificate.
Make a counter card