The Hidden Margin Killer for MSOS ft. Braunz Muller

Listen Now

Bryan Fields: So many times we've seen it — 'Oh my god, we got 32% yield and it looks great, what do we do?' Braunz, your 4/20 post said something simple but loaded: most manufacturing problems are actually data problems, but cannabis companies already have endless data. So where's the real failure? Braunz Muller: The failure's in how you use that data, right? How you compile it, interpret it, view it — making sure that it's wrapped up appropriately for the right audience, and making sure it's actually getting used critically to drive decision-making in the business. What I've seen is people drowning in data without actually using it to inform critical decisions in the company. Bryan Fields: Do you think that's just based on the fact that the industry moves so fast that there are a lot of data points, or more specifically, that the decision-making framework has too many variables? Braunz Muller: Right, when we're talking about extraction there are so many moving variables that there's not a single data point that tells you what caused what. You kind of have to have a holistic understanding of what's happening. If you're talking about extraction, that's a microcosm of this whole data thing. But if you're talking about supply and demand planning, you have to have a holistic vision of what the company is doing, what the goal is. If you see a spike in one product group over a month, what actually caused that? You need to be able to connect all the dots and have cross-departmental, cross-functional collaboration to make sure you're interpreting the data signaling correctly. If you take it down to extraction — why did we get this crazy yield? Well, if you're not collecting the correct data points, you're never going to be able to answer that question. You need to bring in the right data sources, collect the right data points, and have an understanding of how the business or the process works, depending on your scope. Bryan Fields: You've seen the complexity inside an MSO, and I want you to break that down more, because it's way more complex than that. There are so many decision-makers pulling on that information and the ship is moving. Is that a matter of setting up the right infrastructure, too many people pulling on it, or is the information mostly being pulled by the sales organization saying, 'Hey Braunz, we need more of this product, send it this way'? Braunz Muller: I think where I've seen disconnect is you have really good data analysts who are great at taking data and presenting it comprehensively, and you have really good operators who know the goal they want to reach or the levers they want to pull but might not know how to put the data together well. And then you have the operators on the ground who know how to make the product and get it where it needs to go but don't necessarily know what the volume or velocity needs to be. You need someone to sit in the middle of the spiderweb and tie together the why and the how to create that comprehensive picture. You can have the best data analysts in the world, but if they don't understand operations, they're not going to give you the full picture. Where I found success was through a lot of failing, honestly, but eventually the team and I were connecting dots and driving that chain from the bottom up and the top down — if we want to hit a certain dollar-per-unit on a product group, how do we get there? You have to connect dots across production, sales, and fulfillment to get there. Bryan Fields: Right, you're saying there needs to be an anchor in the wheel pulling information accurately from both sides of the chain. And there's an element of tracking after the fact — here's what we projected, here's what the actual was, there's variance — but how does that all get looped together in a tool, because cannabis operates uniquely. It's not just an ERP system, and everyone loves Google Sheets and spreadsheets, which is helpful to a point but overwhelming from a data standpoint. What have you leaned into to aggregate that information, and are there ways teams can apply the same method? Braunz Muller: There are exciting things happening, and I'm not a data-systems expert by any means, but we found a lot of success by using Tableau. We had sales data, wholesale data, retail data, and back-of-house production data all in one place finally, and connecting all the dots across the entire footprint of the business was what allowed us to really dig in and drive results. You can do all of that in Excel, it just takes a million hours of your life. Having smart data aggregators, and now with cloud plugins, you can start to truly automate some of this. One problem I saw even with Tableau was data hygiene drift — no one was doing active QC/QA of the data source, or people didn't know how to spot red flags in the conclusions being drawn. Having someone very ops-savvy monitor the conclusions is helpful, but you can also train AI to do that. We're getting into a new horizon of replacing a lot of expert time with a really smart AI agent, which requires its own QC, but it's something I've been trying to work on in my own time. Bryan Fields: Explain what you mean by QA of the data. Braunz Muller: Little mistakes can lead to radical outcome changes. If you have gummies coding as vapes in your tables and it's not immediately obvious, your unit outputs skew your production needs. We saw that constantly — something as simple as naming conventions is the bane of data management. If something in your front-of-house POS is called something different than in your back-of-house POS, how do you marry those two data points together so you know you're talking about the same thing? Sounds simple, but it's a nightmare to solve — it took a lot of hours of coding and if-then statements to bring that together. Data hygiene is a good place to start if you're trying to build a successful system. Bryan Fields: As we unpack this further, the data has so much value, but it's about how you interpret and utilize it. If it's not clean and accurate, it's not providing information, and then it takes someone sitting down and understanding what it's signaling. Data should be decision guidance, not the decision-maker itself — here are the variables, here's what we've done historically, here's how things are trending, there are unknowns, and the data needs to be trusted but also double-checked. Do you think the industry is using data effectively, or does everyone have a massive opportunity to unpack it further? Braunz Muller: From my scope, there's greater potential. One thing that drives me a little crazy, shifting from demand planning into market trend analysis, is that BDSA is a great tool for market analysis, pack-spec information, all of that — but I see people saying, 'What's the most market-indicated package spec for pre-rolls?' and then everyone goes and makes the same pre-rolls. Instead of asking, if everyone's making one-gram Maui Wowie distillate vapes, do we actually need one-gram Maui Wowie distillate vapes? It's about not letting the data just pull you along, but using it to make real decisions rather than doing whatever BDSA tells you. Every market's different — you pretty much have to have a strawberry gummy in your lineup because people love strawberry, so everyone sells strawberry gummies, which is fine, but how do you differentiate? I see everyone doing the same thing with the data — Select's two-gram came in and destroyed the vape market, so everyone makes a two-gram now, doing exactly what Select did. Select is training the customer, and that's great, but if you can come up with a value play to pull those customers away, you'll be successful — don't just copy exactly what they did. The Brick is great hardware, but is it perfect hardware? No — maybe use that as a play. When the all-in-one market first exploded, everyone did the same thing and the market flooded with mediocre all-in-ones. Sometimes the data all points one way, but you still need that critical, expert lens to view it — true in market analysis, but also in sales data and demand planning. In Colorado, we had a beautiful supply-demand plan doing fulfillment for 22 internal dispensaries across hundreds of SKUs across three production locations, maintained by very smart people. But we kept seeing a cycle of overstock, then inventory depletion, then overstock. We were making too many one-gram vapes, they started to age, they got heavily discounted to move before expiration, then we'd see a sales spike that changed the six-week average, so production spiked again, and those vapes aged too — the total margin of the product line suffered from steep discounting. No one ever stopped to interpret the data a step deeper and ask why we were stuck in that cycle. You need that critical eye of understanding the entire business footprint so you're not just being led along by the data. Bryan Fields: The margin piece is so important — people discount to avoid aging product, but the first question is why they produced so many units in the first place. If they'd produced a fraction of those units, or allocated the extra product to another line, their margin would be way higher. It's about being comfortable pushing boundaries while understanding data is a signal, not a directive to simply repeat. Explain your role at the company, and why there are so many different SKUs these teams produce, so people understand the context of your perspective. Braunz Muller: I was hired at The Cannabist Company as a lab supervisor, with about 15 years of experience in hydrocarbon extraction, optimizing extraction footprints, and vape and gummy manufacturing. I got my foot in the door as a lab supervisor but was promoted four times in a couple of years up to National Director of Manufacturing Operations — not so much a human-management position as a systems-management one. My first mandate was to standardize and optimize extraction, formulation, and production capacity across our footprint, which at the time, before divestment kicked in, was about 15 manufacturing facilities across the country. Each one was radically different — I walked into one still decarbing flower before ethanol extraction, all the way to a brand-new state-of-the-art extraction system in New Jersey — an illuminated double-barrel extraction system, amazing and state-of-the-art — compared to another market, which I won't name, that was literally washing ethanol in a salad spinner. The role grew into complete management of manufacturing: data infrastructure, vendor relations, supply-demand planning, and brand management alongside marketing — understanding what products we were making, what was missing from our portfolio, and how our brands were aligned. Bryan Fields: Why was operations so vastly different under a single company? Braunz Muller: A lot of these MSOs procured businesses that were already doing their own thing, and without active management those businesses just kept doing their thing. You inherit the equipment, teams, capabilities, and SOPs, and there wasn't a concerted effort to standardize across markets — and standardization couldn't fully happen anyway because of regulatory constraints or equipment differences. As best I could, I tried to get everyone on the same page, investing time with teams, closing knowledge gaps, getting everybody onto what I jokingly call the 'Braunz Bible of extraction and formulation.' It's crazy — you spend ten years buying up businesses and don't really try to meld them into one cohesive unit. Bryan Fields: So the lesson from one market should transfer to another so similar mistakes aren't repeated or specialties get uncovered, but without interconnected systems that's difficult. Braunz Muller: Right, and without comparative analysis. I feel fortunate that having all these markets gave me comparison from a cost-model, SOP, and personnel perspective, and it gave me insight that there's no one right way to do things in this industry — you can reach the same end goal a million different ways. The right answer is the one that works best for the team and the facility. I came in with teams that were passionate and making great products, some with great data infrastructure, but they weren't seeing the whole picture. Having a great product is step one — everything else is the nuance, the cost, the why. You could make the best rosin in the world, but should you even be making rosin? Is it cost-effective, or could you sell that flower as eighths and make more money? Those initial conversations were interesting because I'm not someone who likes to dictate change — trying to create grassroots support for optimization with someone who's done something the same way for ten years is a hard sell. Bryan Fields: Let's go back to that — sometimes you shouldn't be making rosin, maybe you should be selling eighths. How would you go about identifying that when you show up to these markets, since some businesses today are wondering the same thing — where the cash is going, revenue trending one way, expenses trending the wrong way, without knowing the gaps. Is that a data-visibility problem, or something else? Braunz Muller: You might have all the data, but aggregating, consolidating, and interpreting it in a way that helps you make decisions is the important part. If you're flower-constrained but could sell more eighths than you are, don't make rosin, sell more eighths. The whole system has hundreds of levers you could pull. In Maryland, we were actually selling our trim and buying trim because the trim we could buy was better than what we could make, and we could sell ours for the same we could buy it — we got more THC per pound just by flipping the pounds, and more THC per hour of extraction. Rosin is the easier lever — if you're flower-constrained, don't make rosin unless you can't sell your flower. A few years ago I was a huge proponent of everyone making rosin, and then I realized rosin is expensive to be successful in — it takes real manufacturing excellence to hit consistent yields and requires the right genetics. It's the least scientific of the extraction methods — ethanol, CO2, and hydrocarbon can get pretty scientific, but hash washing has more art and nuance to it. It comes down to pack spec too — do you need a 14-gram, 28-gram, milled-flower product? What levers can you pull in a market, and which ones should you? Some people don't have the data to make those decisions, but more often they just don't know how to bring it together in a way that helps them decide. Bryan Fields: I wonder if sales sometimes influences production planning — 'we need strawberry gummies now' — but there's a lead-time cycle, and I think it should go back to cultivation: what's our input product, and what can we do to maximize our ingredients for the output? Braunz Muller: Absolutely — in a vertically integrated system, everything goes back to cultivation. One of the difficulties for us was having really good sales and wholesale data but no way to attach pounds of flower to it. So over the last year we assigned dry-weight equivalents to every product — how many grams of flower it takes to make a unit — and could track every product back to flower and incorporate that into cultivation strategy, giving a complete top-to-bottom, bottom-to-top demand-planning picture. Someone asked me recently how you deal with volatility, and it's such a big question I'll spend the next year thinking about it. Retail tends to be a smoother curve than wholesale, but you'll have peaks and valleys in throughput. You have to have your finger on the pulse of the data every single day, interpret it, learn from it, and use it to pull levers dynamically. Flower takes eight to ten weeks to grow — you can't change that — but there's a lot else you can change. If that's your one constant, at least figure out what to do with it, because so much else in the industry has no constants. Bryan Fields: When you talked about variance, I think of the bullwhip effect — teams reacting to up-and-down cycles. I think the better approach is understanding upper and lower limits you're comfortable with, so you don't have to double-discount reactively. Braunz Muller: Right. Bryan Fields: With biomass as the constant, let's go to extraction. What's your perspective on SOPs — should they be linear or dynamic? Braunz Muller: SOPs should be dynamic in the sense that people like to think of them as fixed and constant, but you have much better operational success when your SOPs allow for the messiness of cannabis. That doesn't mean you can't have a plan, but the plan needs to be dynamic — if this happens, then we do this, not 'this happened, now what do we do?' I see that a lot in manufacturing. SOPs also need to change over time — if you've had the same extraction SOP for ten years, it's time to update it. I might be more of an agent of chaos than most when it comes to extraction, but change is how you find new innovation and little tweaks. Dealing with 3,000 pounds of moldy biomass requires a different SOP than fresh material straight out of cure, and your team needs to be able to handle that. A more dynamic SOP structure encourages professional growth and makes people better extractors, operators, and eventually leaders. Bryan Fields: Your starting material is a plant, highly variable, and your extraction system has its own variability, so your output is a combination of variables, which is why yields are so wide-ranging. I understand the dynamic SOP concept, but how do teams implement that while safeguarding it — giving flexibility to hit goals but also organizational control so it doesn't become too artisanal? Braunz Muller: I see extremes a lot — you have the artisanal side, like an amazing 'wook' who's been hash-washing for 20 years with intuition for everything, and then a linear SOP that says do it exactly the same way every time expecting the same results, which also doesn't work. You want to land somewhere in between. I like to start simple — I recently stood up a brand-new hydrocarbon team with zero extraction experience. You don't hand them 15 years of experience and say figure it out; you start simple and build complexity through documented batch sheets tied to good data. You start with basics — run cold, use this much solvent, collect at this heat, purge, crash — and then build in CRC, different gas mixtures based on input material, and so on. Extraction techs are underpaid and undervalued because they're such a linchpin of manufacturing — they should be encouraged to explore what's possible, but within a structured system, workshopping ideas rather than just 'go crazy.' Bryan Fields: That empowerment matters, but it's important that connected data systems capture what operators are improvising so it doesn't get lost in translation — otherwise the yield ends up a certain way and no one can trace why. Ideally the data feeds into a system someone reviews to confirm what happened and adjust going forward. Whose job is that internally — director of processing, finance, operations? Braunz Muller: Depends on org size — a larger MSO like GTI probably has a VP of extraction with a finger on that pulse. But I'd expect a lab manager or extraction manager to have a good understanding of it too, and to be monitoring that data day to day. There are tech hurdles to doing this well without expert experience — understanding variable tracking, because if you change three things in a run, you don't know what you actually did. You need to be systematic and bring the scientific method to things, and that can all go into an SOP. Your on-site management should be the owners of that information. So many times it's 'oh my god, we got a 32% yield and it looks great, what do we do?' — and no one knows why. Bryan Fields: Sometimes decisions made upfront in extraction were meant to set up something on the back end, and if that entire chain isn't connected, you might hurt yourself on the front end trying to save yourself on the back end without realizing it. Braunz Muller: What you're describing is closing the loop of extraction. I once accidentally made about 36 kilos of HTE as a byproduct while making isolate and couldn't sell it — I thought I'd put it into vapes or sell it bulk, and it didn't work out. I learned to always keep the cycle of extraction closed in my head — for every gram you process, it better be going into a finished good or out the door as bulk. Going into a run, you shouldn't just decide to make live-resin dabs today; you should ask what product you need to end up with and figure out the dynamics to solve that equation. I haven't found a data infrastructure that does that really well. Working with Dutchie the last couple of years, the ability to schedule production and set up assemblies ahead of time within a seed-to-sale system was cool, but it doesn't house the actual assembly data of an extraction — temperature, time, CRC, input-material quality. Bryan Fields: That would close the loop from a data-visibility standpoint. You kept that information in your own head, but you're one person putting out fires across many states — what's the pushback from teams still running everything in Excel? How impactful would a dedicated tracking system be, tracing a cultivar's yield all the way to finished product, understanding what to adjust — temperature, post-processing steps? That's where margin evaporates, when teams don't account for added labor or time cost and end up discounting product they thought was more profitable than it was. Braunz Muller: I see high-level production planning saying 'we need this many vapes per month, so we need this much flower,' and averages are useful for demand-planning discussion, but directly tracing vapes to flour that precisely would require data infrastructure I've never seen. Even extraction efficiency gets complicated by people pulling fractions short or missing a couple of degrees — all data points we could be capturing. I increased the efficiency of all our manufacturing facilities almost 20% just by asking people to actually record their data — no other change, just getting people to own what they were doing day to day, and then finding outliers to address in the SOPs. Capturing that information and giving people the right lens to view it through, rather than a tsunami of data, saves you money in itself. Bryan Fields: You take an absolute wizard's knowledge and apply it through systems so new employees raise the floor — you can point to the numbers and say here's what was different, and if you improve on it, yield goes up substantially. Braunz Muller: Yield is a KPI everyone loves to talk about, and it matters, but not as much as people want it to — if you yield a pile of sludge, you just have to winterize and remediate. Yield is just one of dozens of helpful numbers. It's also late — teams often get it back from the lab after the fact and think, 'cool, now what?' You should have a better read earlier in the process to understand if you're trending toward your target, rather than waiting four weeks for a report to be disappointed after you've already consumed the resources. Bryan Fields: Seems pretty reactive and difficult to fix after the money's already spent. Braunz Muller: Especially in hydrocarbon extraction, since it's a little opaque — if you ran less solvent than you thought and got 7% yield on biomass that should've yielded 25%, catching that while you're actually extracting saves a ton of money, probably everyone's wages for the day. Bryan Fields: We're talking about THC molecules passing through a system, and without tools to track how they move, you're missing them or throwing them away — probably more common than most teams want to admit. Braunz Muller: Absolutely. There are amazing systems out there — Luna Technologies for hydrocarbon, large ethanol systems — but what's still missing is analytical data like injection temperature. Don't just tell me the column iced up, tell me how cold it actually was — that's cheap to implement. Actual extraction tracking and data-logging systems are more expensive but pay for themselves, because they help you avoid operational misses like running 100 kilos of flour testing at 30% and only getting a 7% yield. Bryan Fields: And a year later, if that same product crushed, you'd want to know what it was run at, what it yielded before post-processing, how many steps it went through — all details you need if you're trying to replicate a quality product with a standardized but traceable process. Bryan Fields: You said yield is the wrong KPI — what's the better one? Braunz Muller: Efficiency. It depends on the product, but broadly it's how much THC was in your starting material versus how much you got in your crude or post-extraction concentrate. Yield doesn't matter as much in extraction — you could be pulling fat all day. Someone once told me seeded flower doesn't matter because you just press it in a rosin press and get a lot of oil — but that's not the molecule we're here to care about. Efficiency, not just THC but overall cannabinoids, is the KPI to track. Unfortunately it requires good testing on inputs, which ties back to how inconsistent a lot of our testing labs are — but internally, teams still need to find a way to understand whether material is 2% or 25%, because the outcome should be vastly different. Bryan Fields: Which goes back to whether that standard SOP has been updated in a while. Braunz Muller: Right. If you really hone in on efficiency, you start connecting dots vertically — if your non-cannabinoid-containing biomass is high in your trim, you have to go back to the harvest team and have them do a more intense defoliation before harvest so less ends up in trim, which benefits your post-extracted crude. You can start thinking about COGS in a granular, intimate way across the footprint. If you think about what it costs to make a milligram of THC, you can trace that cost through every step of production — thinking in dry-weight equivalents, or better yet, actually costing THC itself, since that's the molecule we're optimizing for. Bryan Fields: For vertically integrated operators, that's where the most efficiency opportunities are, but also where the most risk is if the system isn't interconnected — one team's shortcut becomes another team's downstream problem. Braunz Muller: It really matters. Cultivation is hard, labor-constrained, and you have to think about cost-benefit — maximum efficiency in isolation isn't always the right business move for the whole system. But investing in cultivation efficiency matters enormously because everything trickles downhill from there. In Ohio, we spent all this time optimizing the manufacturing facility, got COGS way down, had great predictive cost analysis — then one quarter our trim-to-flower ratio shot up to 50%, mostly LAR being stripped into bags, and that cost flowed straight down to manufacturing. We went from an optimized vape cost to barely breaking even, because staffing shortages in cultivation screwed up the whole system. It's okay to let some things slide, but I only caught it because I was so plugged into the data coming out of those facilities. So many COGS blowouts come from small drifts that, if caught early, don't snowball. For that you need really good data management. Bryan Fields: The C-suite probably looked at the numbers later and knew something wasn't right — if the team upfront had known the downstream implications, they'd likely have done whatever they could to fix the bottleneck early. Braunz Muller: I don't know — there are scenarios where you have to accept the consequences of certain actions. If that LAR hadn't gone to extraction, maybe we'd have had to sell it at a loss; maybe it was better to eat it internally. I don't want to speak for the C-suite's full awareness, but it felt like it was starting to snowball, and I'm glad I could get ahead of it. We got the trim ratio back to about 70/30, where we wanted it. Bryan Fields: Two last questions. What's one piece of advice you'd give operators listening today — one thing they can do now to audit their business? Braunz Muller: I'd go back to efficiency, especially across processes in manufacturing. A lot of people who aren't intimately tied to manufacturing see it as a black box, sometimes a dumpster fire of money. Dig into the efficiency across every production method — teardown, trim, extraction, post-extraction processing, final-goods packaging — and understand where you can get massive efficiency gains just through simplification. Really hone in on the efficiency every time you touch that THC. Transfer loss is crazy — I walked into one facility where there was oil all over the extraction-room floor, essentially someone's whole salary just sitting on the ground. Understanding the true efficiency — the molecules of THC moving from one stage to the next — can pay for a lot of your operation, or at least help you understand the slippage a lot better. Bryan Fields: Is the Braunz Hydrocarbon Extraction Bible a real, living document? Braunz Muller: Funny you should say that — I've been doing some work with Claude to take everything I've said, written, and done and put it into a very dynamic SOP. It's not ready for publication, but there are rumors it could become a living document in the future. I'd never claim to be the one source of knowledge on hydrocarbon extraction — anyone who thinks they're doing something perfectly is wrong. Over the years I've seen a lot of wrong and right ways to do it, and I'm trying to compile that into one place. I ran my first hydrocarbon system 15 years ago, and I probably learned as much last year as I did in the first five years, just from continuing to push my knowledge with new companies, new technologies, and hopefully better logging tech coming soon. One day I'll have a book. Bryan Fields: Last question — what question do you wish more people asked you? Braunz Muller: What am I really excited about day to day? Honestly, I'd talk about hydrocarbon extraction all day, every day, forever. I genuinely wish people would ask me to tell them about hydrocarbon extraction — tell me about the solubility constants of mixed alkanes, please. No one ever asks me that. Bryan Fields: For people who want to get in touch, learn more, and pre-order the Bible when it comes out — where can they find you? Braunz Muller: LinkedIn's great — I just broke into that world this year, but it's a great place to connect. Send me an email; I'm happy to talk about anything and everything in the cannabis or hemp industry with anyone who's interested. Bryan Fields: Thanks for taking the time, this was a lot of fun. Braunz Muller: Yeah, thanks, Bryan.