Table of Contents
Cannabis companies have spent years chasing scale. But as the industry matures, getting bigger for the sake of getting bigger is giving way to a more difficult question: How do you scale strategically?
For cannabis operators, that increasingly means understanding exactly what is happening at every stage of production and using that knowledge to decide where efficiencies can be gained and where investments make sense.
Veterans Choice Creations CEO Jason Ambrosino has taken that approach at his New York cannabis company.
“A lot of cannabis companies have focused on scale first and understanding the process second. We took the opposite approach,” Ambrosino said. “My concern has always been that if you scale a process you do not fully understand, you are really just scaling inefficiency.”
For Ambrosino, that philosophy has meant investing heavily in R&D and data collection across cultivation, extraction and manufacturing. The idea isn’t to prioritize R&D over growth, but to use what the company learns to make better decisions as it grows.
“Cannabis manufacturing still has a tremendous amount of variability in raw material, extraction performance, formulation, hardware, packaging and even consumer response,” Ambrosino said. “There is still a lot we as an industry do not know.”
For operators, one takeaway is that useful R&D doesn’t necessarily begin with developing a new product. It can start with better understanding what is already happening across the business.
In cultivation, Veterans Choice Creations tracks yield, cannabinoid and terpene expression, cultivar performance and environmental conditions. In extraction, it looks at input potency, extraction yield, cannabinoid and terpene recovery and post-processing loss. On the manufacturing side, it tracks throughput, labor, fill-weight variance, scrap, rework, hardware failures and batch performance.
But production data alone only tells part of the story. Ambrosino argues that operators also need to connect those metrics to what happens after a product reaches the market.
What did it sell for? How quickly did it move? Did retailers reorder it? What was the contribution margin?
“A 15% extraction yield by itself does not tell me very much,” Ambrosino said. “I want to know what cultivar produced it, what the chemistry of the input material was, what should theoretically have been recoverable, what the resulting extract became, what it cost us to manufacture and whether the finished product actually performed in the market.”
The lesson is to look beyond whether an individual department is hitting its benchmarks and consider the economics of the entire production process.
Rethinking Yield
Extraction offers a clear example of how that can change decision-making.
Extraction yield is often discussed as a percentage. A cultivar yielded 10%, 12% or 15%. Ambrosino’s team began looking at that number against the cannabinoid content of the material going into extraction.
“We started asking a different question: yield as a percentage of what?”
Knowing how much cannabinoid was present in the incoming material allowed the company to compare what was actually recovered with what could have been recovered.
As Veterans Choice Creations built data across cultivars, the results began influencing the material it purchased and how it used its own production.
Some cannabis may be valuable as flower but make little sense as extraction material. Another cultivar might sell for less as flower but perform well enough in extraction to produce better margins. Certain material may be better suited for cured resin, fresh frozen or distillation.
“The laboratory result stopped being just a report card at the end of production. It became an economic input into the next production decision,” Ambrosino said.
For other operators, the broader point is that production data becomes more useful when it informs the next decision rather than simply measuring the last one. Extraction results can influence purchasing, product planning and how companies allocate their own cultivation output.
“The interesting thing is that R&D ultimately becomes a cost-reduction strategy anyway,” Ambrosino said. “Better understanding improves yields, reduces failed production runs, helps us allocate raw material more intelligently and allows us to develop products faster.”
Connecting The Data
Collecting information is only one part of the equation. Cannabis companies also have to be able to use it across departments.
Cultivation, extraction, manufacturing and sales often operate with different systems and datasets. The challenge is getting information to move between those functions instead of remaining siloed.
“Today, most cannabis companies still have a lot of functional silos,” Ambrosino said. “The systems themselves were usually never designed to communicate with each other. We are much further along in collecting the information than we are in completely automating those feedback loops. That is the next challenge.”
Cannabis businesses generate large amounts of information, but it can be scattered across cultivation records, COAs, extraction logs, manufacturing systems, inventory platforms, METRC, sales data, emails and employees’ own knowledge.
AI could make some of that information easier to use.
“AI is becoming an intelligence layer across the information we already collect,” Ambrosino said. “The problem was never necessarily that the information did not exist. The problem was that interrogating all of it was incredibly labor intensive.”
At Veterans Choice Creations, AI is being used to analyze larger amounts of information and ask questions that would have previously required employees to manually pull together records from different systems.
That can include comparing extraction results among cultivars, looking for anomalies in manufacturing runs, evaluating raw-material purchases and studying hardware performance.
“I am much less interested in AI because it can write an email than I am because it can help us examine thousands of operational decisions and potentially identify relationships that a human being would never have enough time to search for manually,” Ambrosino said.
For operators considering similar tools, the practical hurdle may be less about adopting AI and more about making sure the underlying information is being captured in the first place.
“We want the relevant information available at the point where the decision is being made,” he said.
A Few Percentage Points Matter
The case for investing in R&D and better data becomes stronger as cannabis markets mature.
Margins compress, retailers become more selective and products become more technically complex. At scale, relatively small improvements in production can have an outsized effect on profitability.
“Once margins get tighter, being 2% or 3% better at recovery, labor utilization, inventory allocation or manufacturing yield can become the difference between a profitable product and an unprofitable one,” Ambrosino said.
There is another challenge that comes with growth: preserving what employees have learned.
“At small scale, a company can operate on tribal knowledge,” Ambrosino said. “A talented extractor knows how a machine behaves. A cultivator knows which room performs well… The problem is that tribal knowledge does not scale very well. When you are producing millions of units, small amounts of undocumented variability become very expensive.”
That makes documenting processes, experiments and results particularly important for companies planning to scale. Knowledge that exists only in the heads of individual employees is difficult to replicate across additional production lines, facilities or teams.
“R&D gives you repeatability. Data gives you institutional memory. Analytics and AI allow you to use that institutional memory at a scale that previously would have been extremely difficult,” Ambrosino said.
The same principle applies to AI. As the technology becomes widely available, simply having access to a capable model is unlikely to differentiate one cannabis operator from another.
Operational history might.
“The advantage will come from the proprietary information those models can work with, the quality of that information and whether the organization has the discipline to actually use what it learns,” Ambrosino said.
“A company that has not collected that information cannot suddenly recreate five years of operational history because it bought an AI subscription.”
That may be the most relevant takeaway for cannabis companies looking at R&D investments today. The value isn’t necessarily one breakthrough product or a dramatic improvement in yield. Ambrosino sees it in the accumulation of better decisions about raw materials, production schedules, formulations, hardware, pricing and inventory.
“The return is increasingly being able to replace assumptions with evidence.”
As cannabis moves closer to advanced manufacturing and traditional consumer packaged goods, scaling may increasingly depend on how effectively companies can learn from what they’re already producing.
“We wanted to understand the process well enough to manufacture tomorrow’s,” Ambrosino said
