Click here to view original publication.

 

Products based on cannabis extracts are the fastest growing segments in the market. The demand for more cannabis extracts is constantly growing and so is the price pressure on these extracts. In most mature markets, the prices of cannabis and cannabis extracts have been declining, putting pressure on producers to stabilize their profit margins. This trend is usually expected from an established market. If revenue is falling, the only way to protect profits is to reduce costs.

Business people will suggest cutting labor costs or reducing material costs. We are scientists; we will not do such things. We want to point to the waste and inefficiencies in production and how to remove these. We want to help you reduce your cost of production and, possibly, make some better extract along the way.


What’s the issue?

Data Blindness
While working in process chemistry and data analytics, many shortcomings can be observed in the production of cannabis throughout the industry. For example, terpenes are lost and cannabinol (CBN) is formed in the drying and decarboxylation process, extraction equipment is slowly breaking to leave behind delta-9-tetrahydrocannabinol (THC) oils in the plant, and badly tuned processes produce low-quality oils that need costly cleanup steps.

Not only do these drawbacks pose big problems to many companies, but most find it difficult, or even impossible, to understand the causes behind them. Though the answer differs for each case and company, drawbacks can be more easily realized if producers took a more comprehensive approach while tracking their process. Knowing what happens to the cannabis material throughout the entirety of the extraction process is extremely valuable. Simply enough, if more is known about the processes and mechanics involved with production, instead of just the material itself, we are able to monitor the processes, analyze them, and find improvement potential.

What can we do?

 

Data Tracking for Process Control

Extract manufacturers live by “garbage in, garbage out! Quality in, quality out!” How about garbage vs. quality work?

We see many companies put a singular focus on the input material, focusing on aspects such as weight and the profiles for cannabinoids and terpenes. While material input plays an important role in influencing the outcome, it does not do enough to reflect the complexity of all other process inputs involved (i.e. decarboxylation time, extraction pressure, winterization solvent, distillation temperature, etc.). We should also be tracking other important inputs. When considering extraction, by keeping better track of certain aspects of the extractor, such as the set temperatures, pressures, run times and the date, and more obscure aspects like particle size, we gain insights that could lead to minimizing waste production as well as high concentrated outputs. We can also spot instrument degradation early and in turn, minimize loss of efficiency among many other things [Figure 3].

Evidently, there are many other characteristics, both internal and external, in production that should be tracked and analyzed in order to better understand previous outcomes and, hopefully, prepare for future ones.

What should we do?

 
Data Analysis for Process Improvement

If the production teams keep meticulous track of their process data, and even sprinkle in some well-planned experimental conditions, it becomes possible to improve the process even further. Here is an example of how data was used to improve both the yield of cannabis extraction and the concentration of THC in the extract.

We can correlate how important certain inputs are to a known output through an analytical process known as the Design of Experiments (DOE). We can compute response surfaces to visually evaluate how the combination of input variables (e.g. temperature and pressure) impact the resulting output (e.g. yield and concentration). Figure 1 shows that we can expect the highest concentration of THC to be extracted at the middle ground pressure for separator 1 and higher pressure for separator 2. Whereas, Figure 2 shows that we can expect to have the highest yield at a low pressure for separator 1 and the lowest pressure for separator 2.

Why we should do it?

 

Data Analytics for the Future

Although, data analytics has found its way into sales processes some time ago, data analytics for production only saw its first applications very recently, and is still not widely accepted. The use of data analytics in cannabis production appears daunting. The usefulness of such analytics is also not directly apparent.

Process data analytics does not need to be hard; there are specialists for it, but it can be very useful if you try it out. 

The cannabis industry is in an early phase and we still have room for large improvements. But if we don’t know what we are doing, we cannot make it better.

Figure 1 Normalized DOE response surface illustrating what concentration of material extracted is expected at given pressures of separator 1 and 2.

Figure 2 Normalized DOE response surface illustrating what yield of material extracted is expected at given pressures of separator 1 and 2.

Figure 3 The mass of material being extracted per hour in relation to the experimental runs. The mass rate extracted has been normalized between 0 to 1.

By keeping track of how much material gets extracted per run and for what duration, we can actually keep track of the efficiency of certain instruments. As the trend suggests in Figure 1, with more experimental runs, less material weight gets extracted per hour, hinting to a failing pump.

Taking the hypothetical scenario where the company has run roughly 85 extraction runs in total, conducting 2 runs a day at 10 hours per run (4 hours used for cleaning and turnover), we can actually calculate the loss of potential extracted material during this span. This amounts to 850 hours.

The area covered in green represents the mass of material that has been extracted (X) whereas the area covered in red represents the mass lost from the failing pump (Y). So the total mass of material that the company could have extracted if they were to have tracked their pump earlier (Z) can be defined as:

 
So, the lost mass in comparison to the total can be denoted as (W) where

 
Using some basic geometry as a rough estimate, we see that  equates to a 20% loss in yield of oil. This is equivalent to 5.61 kg of oil being lost throughout the span of all extraction. If we were to assume a gram of oil sells at $10 per gram, this amounts to a potential loss $56,000.