How Cannabis Brands Can Better Classify Their Strains

When customers go to a dispensary they’re typically told about sativas, indicas and hybrids. But these classifications are a false dichotomy.

The terms characterize cannabis strains by the species of plant that they come from, but this does not directly determine the results that a strain produces.

The idea that all sativas make you feel energized or that all indicas relax you is untrue. There are many variables - such as growing conditions, the presence of pests and whether the product was cured - which influence a strain’s effects far more significantly than the classification of the plant it comes from.

Historically, it’s easy to understand how this mistake was made; sativa and indica are different species of the cannabis plant with contrasting physical characteristics; they have dissimilar heights, leaves and even odors. The assumption that the strains from these plants would produce distinct effects is sensible. But ultimately false; no specific results can be attributed to the strains from either species.

What Do Industry Leaders Say?

Industry experts agree that these kinds of classifications are disingenuous. It’s accepted that there’s no pattern to the mix of cannabinoids and terpenes found in strains from either species. Given it’s these chemical components which produce the most noticeable effects when we consume cannabis, it’s illogical to suggest that all strains from a particular species could consistently produce specific results.

The clinical impact of cannabis has nothing to do with the height of the plant or the shape of its leaves. Cannabis brands and entrepreneurs need to stop using plant species as the basis for broad assertions about the outcomes a strain will produce.

How to Accurately Classify Marijuana Strains

As it’s impossible to guarantee certain results will be produced by every indica, sativa or hybrid, it’s important for cannabis brands to learn about the specific effects of each of their strains, and find a more intuitive way of classifying them, in order to more accurately inform their customers.

This should be done by taking a data-driven approach, founded on detailed and relevant information - not generalizations and misconceptions. We developed an approach which used machine learning to create a master dataset of roughly 3,000 different strains. Each strain had its own ‘feature set’ encompassing its various cannabinoids, terpenes, the ratio in which they were found, and the sentiment that members of the public applied to that particular product (i.e. the kinds of words they used to describe it).

The purpose was to develop classifications which predicted the effects of all strains which fell within them, making the chemical properties of, and public sentiment towards, each strain the relevant information; not the species of its plant.

Then we used an algorithm to create three groups from this data. Interestingly, when 3,000 strains were sorted into just three groups, most of them fell into the categories of sativa, indica and hybrid. But if the goal is to produce groupings which give precise indications regarding the likely effects of these strains, a greater number of groups is required.

Next we programmed the algorithm to provide the best way of assorting the 3,000 strains without restricting the number of groups. This time it produced five distinct categories. We then removed the CBD-only strains to produce a sixth.

The resultant six categories of strains produce specific effects due to their unique chemical compositions; our energized, elevate, create, chill, sleep, and high-CBD strains, whose classifications provide far more granularity than their plant species.

In many cases the classifications span all three indica, sativa and hybrid categories, given the number of different variables under consideration. But it’s by moving beyond false dichotomies centered around plant species that we can take a more nuanced view of cannabis strains, understand them more, better inform customers and stop providing misinformation that ultimately harms the industry.

This can be achieved by a data-driven approach which reliably classifies different strains based upon their actual effects; using feature sets consisting of cannabinoids, terpenes and public sentiment to create classifications entirely independent of a strain’s genealogy.

How Brands Can Improve Their Classification of Marijuana Strains

This approach is more accessible than it may seem. There are just three simple steps that cannabis entrepreneurs need to take in order to better classify strains according to their effects:

Construct Informative Feature Sets

As described above, observe the cannabinoids and terpenes found in each strain, as well as the ratios in which they’re found and the sentiment applied to that strain by the public. This type of semantic analysis uses relevant information to predict the effect that each strain produces and will underpin each of your classifications.

Use Machine Learning to Analyze the Data

Next, use an algorithm to convert the data from the feature sets into precise categories. We did this by creating a matrix with columns for the relevant cannabinoids and terpenes, and separate rows for each strain, and populating it using the values we’d obtained for the feature sets. We also included the sentiment analysis, whereby certain words were related to specific outcomes.

The algorithm will use this information to produce classifications and distribute your strains between them.

Creatively Develop Category Names

This means applying a human touch to the categories produced by your algorithm. We did this by noting the sentiments associated with the strains in each category and attempting to figure out why they had been grouped that way.

With some additional research, analysis into the constituents of the relevant feature sets and the application of some creative liberty we were able to formulate terms that adequately captured the characteristics of each category.

It’s vital, for the sake of the industry, that cannabis brands stop perpetuating misconceptions about sativas, indicas and hybrids. The first step is appreciating what really impacts the effects of marijuana strains and understanding your own products. The second is packaging this information in a way that’s accessible and passing it on to customers. By better classifying marijuana strains we can improve the consumer experience, increase faith in the industry and help it continue to grow.