The Math The Winning Shop is Doing
Fourteen formulas that separate the dispensaries that make it from the ones that don’t.
Illustrative simulation, recomputed each cycle: same weekly demand series for both shops, same $20/unit margin, $8/unit markdown on unsold stock, $5,800/wk fixed costs. The only difference is the ordering rule. Cumulative profit shown against breakeven.
There are two dispensaries on the same avenue. Same license type. They carry the same brands at relatively the same price points, from mostly the same distributors.
One of them is quietly underwater. The other one is fine.
The difference isn't the menu, the buildout, or the budtenders. It isn't foot traffic either. The difference is that one of these shops can tell you, right now, its margin return on every dollar sitting on the shelf, how many days of its slowest SKU it's holding against that SKU's decay clock, and exactly how much of last month's inventory variance was price and how much was product walking out the back.
The other shop is guessing. It buys on feel, prices on the neighbor, and finds out about problems when the reconciliation doesn't reconcile.
And as of September 15, 2026, the guessing got more expensive. New York's C.O.D. list — the state's register of retailers who haven't paid their suppliers within the 30-day credit window — now lives inside METRC itself. Distributors report delinquent invoices in the same system that tracks the product, and once a retailer is listed, METRC automatically restricts their transfers to cash on delivery until the debt clears. No supplier discretion, no quiet workout. The system enforces it.
At the September Cannabis Advisory Board meeting, OCM put numbers on what that list looks like: 87 retailers — roughly one in eight in the state — carrying about $3.9 million in delinquent invoices, an average of 281 days past due against a 30-day term. Trace any of those retailers backward and you'll usually find the same sequence: cash tied up in inventory that didn't move, an invoice that couldn't be covered, credit gone at exactly the moment credit was the only way to restock. The math in this article is how that sequence gets interrupted early — or never starts.
Nobody opens a dispensary to do arithmetic. But this industry punishes guessing more brutally than almost any other retail category: the product decays, the demand is lumpy, the tax code confiscates margin the other categories keep, and the regulator reads your inventory ledger. In that environment, arithmetic isn't back-office hygiene. It's the survival trait.
What follows is the math the winning shop is doing — handpicked formulas, in the order the decisions actually arrive: predict what's coming, decide how much to buy, price it so it covers you, measure what moved, catch what leaked, and grow what's left. None of it needs more than a spreadsheet. All of it needs to be done on purpose.
01 / predict
You can't buy for demand you haven't forecast
Every formula in this piece sits downstream of one question: what's going to sell next week? Get that wrong and everything after it — the order, the price, the shelf — is built on sand. So the winning shop starts here, with the least glamorous math in the whole stack.
The moving average. Take your last four weeks of unit sales for a SKU or category. Add them up. Divide by four. That's your baseline forecast for next week.
That's it. It feels too simple to count as analytics, which is exactly why most shops skip it — and skipping it means the purchase is running on the buyer's memory. A four-week average is dumb, stable, and hard to fool. If you want it slightly smarter, weight the recent weeks more heavily — this is called exponential smoothing, and the idea is simply "recent history matters more, but old history still votes."
The baseline isn't trying to be right. It's trying to be honest — a number you didn't negotiate with. Every real forecasting conversation starts with "here's the baseline, now what do we know that the average doesn't?"
Which is where regression comes in. A moving average assumes next week looks like the recent past. But you know things the average doesn't: there's a holiday coming, the first of the month lands on a Friday, a competitor two blocks over just had its license suspended. Regression is the tool that turns those hunches into coefficients — it takes your sales history and a handful of candidate drivers and tells you which ones actually move the number and by how much.
Two things matter when you read the output. The coefficient on each driver is the size of its effect — "a promo week adds about 60 units" is a coefficient talking. And R² is the honesty score: it tells you how much of the week-to-week swing your drivers actually explain. An R² of 0.7 means you've accounted for 70% of the movement and the rest is noise or something you haven't measured yet. Low R² isn't failure — it's the model telling you your demand has a driver you haven't found.
The cannabis-specific advantage: this customer base is unusually forecastable. A cash-heavy demographic with benefits cycles, paydays, and a strict calendar of cultural spikes produces demand with more structure in it than most general retail ever sees. The first-of-the-month effect isn't folklore — it's a coefficient waiting to be measured. So is the Thursday-before-a-long-weekend effect. A shop that runs even a crude regression on eighteen months of its own POS data will find patterns its competitors are still attributing to luck.
The buyer who can explain to the dispensary owner that "our baseline is 340 units and payday adds 15%" is playing a different game than the one who walks in with a feeling.
If you're not the buyer yet: this section is the one you can act on without permission. You see the floor every shift — you know the first-of-the-month rush and the dead Wednesdays before anyone runs a query. Start writing it down. The person who can say "here's the pattern, I tracked it" is auditioning for the buying job whether or not it's posted.
02 / buy
The formula with a name nobody knows they're using
Every buyer already runs this math. They run it badly, on instinct, in the parking lot before the vendor meeting: if I order deep and it doesn't move, it sits and decays. If I order light and it rips, I'm out of stock on my best SKU by Friday. That tension — the fear of too much against the fear of too little — has a name, a formula, and a 70-year-old literature. It's called the newsvendor problem, after the paper boy deciding how many copies to carry when unsold papers are worthless by morning.
Flower is closer to a newspaper than anyone in this industry likes to admit.
The critical ratio. The formula asks you to put a number on each fear:
The cost of running out (call it Cu, the underage cost) is what one lost sale actually costs you — not just the margin on that unit, but the customer who wanted your best-selling strain, found it gone, and bought it two blocks over. In a market where you can't advertise your way to new customers, a stockout on a destination SKU isn't a missed sale; it's a referral to your competitor.
The cost of overstocking (Co, the overage cost) is what one unsold unit costs you — the capital tied up, and in cannabis, the decay. An eighth that sits for ninety days isn't the product you bought. Terpenes fade, the flower dries out, and eventually you're discounting it just to move it, which means Co isn't hypothetical — it's the markdown you'll eventually eat, plus the brand damage of selling tired product at your counter.
The ratio spits out a number between 0 and 1, and it has a direct meaning: it's the service level you should buy to. A critical ratio of 0.8 says: order enough to cover demand 80% of the time. Concretely — take your demand history for the SKU, find the level that 80% of your weeks fall under, and that's the order.
Watch what the formula does: when running out is expensive and overstock is cheap, the ratio climbs toward 1 and tells you to buy deep. When the product decays fast and a stockout is shrug-worthy, the ratio falls and tells you to buy tight. Your hero flower SKU and your fourth-string vape brand should not be bought to the same service level — and now you can say why, in numbers, instead of feel.
Reorder point — the when to the newsvendor's how much.
Order when stock hits this line and the new delivery lands just as you'd otherwise run dry. The trap is the lead time. A normal retailer counts the days from PO to truck. You have to count the days from PO to sellable: sales managers approving the order, vendors packing the pick list, and the transfer manifest fixed to a pending status in METRC until delivery. Your effective lead time is longer than the invoice says and lumpier than you'd like — and every day of it pushes the reorder point up. A buyer who uses the supplier's quoted lead time is systematically ordering late.
Safety stock — the buffer, sized on purpose.
The buffer against surprise. The only unfamiliar character is Z, and it's just the dial that converts your chosen service level into units: a Z for 95% service is bigger than a Z for 80%, so higher protection means a bigger buffer. (The values come from a standard table; the spreadsheet handles it.) The rest is intuition made explicit: more volatile demand needs more buffer, and longer lead times need more buffer — but by the square root of the lead time, not the full length. Doubling your lead time doesn't double the buffer you need; it raises it about 40%.
The loop closes here: the service level you feed into Z is the critical ratio from the top of this section. The newsvendor decides how protected each SKU deserves to be; safety stock converts that decision into units on a shelf.
The point of all three together: the winning shop doesn't buy every SKU the same way. It buys its destination flower deep because Cu is brutal, buys its experimental shelf tight because Co is the decay clock, reorders against the real METRC-inclusive lead time, and holds buffers sized by arithmetic instead of anxiety. The guessing shop treats every SKU like every other SKU — and eats both failure modes at once: stockouts on the best sellers, markdowns on the slow movers.
03 / price
280E is the silent partner in every price
Gross margin is the least exotic formula in this piece:
Sell an eighth for $60 that cost you $36 wholesale, and you're running a 40% gross margin. Every retailer on earth computes this. What almost no one outside this industry has to compute is what happens to that margin after the federal government is done with it — because cannabis retail operates under a tax provision that treats the business, in the eyes of the code, like a trafficking operation.
Section 280E, in one paragraph. Because cannabis remains federally scheduled, 280E bars the business from deducting ordinary operating expenses — rent, most payroll, marketing, utilities, insurance — from its taxable income. The only thing it can subtract is cost of goods sold. Read that again from the P&L's point of view: the IRS taxes your gross profit as if it were your net profit. Every dollar below the COGS line — the budtender's wage, the lease, the security contract — is spent with money that's already been taxed as though it were earnings.
This does something perverse to the arithmetic every other retailer takes for granted. In normal retail, gross margin is a health indicator — one number among several. Under 280E, gross margin is nearly the whole tax story. Two shops with identical net incomes on paper can owe wildly different tax bills depending on how their costs split between COGS and operating expense. And a shop that looks profitable pre-tax can be genuinely, structurally underwater after — not because it's run badly, but because the code is doing exactly what it was written to do.
Which is why breakeven lies to you here. The standard formula:
How many units must move to cover the rent and the payroll — the go/no-go math for a new SKU, a second register, a delivery license. In any other category you'd run it once and trust it. In cannabis, the naive version systematically understates the target, because the fixed costs in the numerator are being paid with post-tax dollars the formula assumes are pre-tax. The honest version inflates those costs by the tax burden 280E creates before dividing — and the honest breakeven always lands higher. The gap between the naive number and the honest one is 280E's cut.
The practical consequences for pricing are three. First, your floor is higher than the neighbor-watching approach suggests — matching the shop down the block tells you about their guesswork, not your survival price. Second, COGS discipline is tax strategy, not just procurement: what legitimately belongs in cost of goods is worth an accountant's sustained attention, because it's the only lever the code left you. And third, when a price war breaks out — and in a maturing market it always does — the shop that knows its 280E-adjusted breakeven knows exactly how low it can follow. The shop that doesn't will follow the market straight past its own floor and find out at filing time.
A currency note, because this piece doesn't assert what it can't stand behind: the mechanics above describe 280E as it has operated. Federal rescheduling — moving cannabis off Schedule I — would change this picture materially, and that process has been in motion.
04 / measure
What sold, and what just sat there decaying
The buy is placed, the price is set. Now the shelf starts reporting — three formulas, each answering a sharper question than the last.
Sell-through rate: is it moving?
Bring in 100 units, sell 62 in the first month — 62% sell-through. This is the buyer's first-glance number, the one that separates the reorder pile from the kill pile. In ordinary retail, weak sell-through means capital sitting on a shelf, which is bad enough. In cannabis it means something worse: the product is getting worse while it waits. Flower is on a decay clock — terpenes off-gas, moisture walks out, and the thing you eventually sell at week twelve is not the thing you photographed at intake. A slow SKU is losing you money twice, once as tied-up cash and once as declining product. Sell-through isn't just a sales metric here; it's a freshness audit.
Inventory turns: how hard is the capital working?
If sell-through is per-SKU, turns is the whole-shop pulse: how many times a year your inventory dollars cycle back into cash. Six turns means your money makes the round trip every two months. And here 280E returns: in normal retail, slow turns are cushioned by the fact that carrying costs are deductible. Under 280E they mostly aren't — so capital parked on a shelf is punished by decay, by opportunity cost, and by a tax code that won't let you write off the cost of the parking. Slow turns hurt everyone; they hurt a cannabis operator on three fronts at once.
The companion number, for reorder conversations:
Same information, expressed as a countdown — and the move is to lay it against the product's shelf life. Holding 90 days of a SKU that's noticeably degraded by day 90 isn't a stocking position. It's a scheduled markdown.
GMROI: is it actually making money? The payoff metric.
Read it as: for every dollar living on the shelf, how many dollars of margin come back? A GMROI of 2.5 means each inventory dollar returns $2.50 of gross margin over the year.
Here's why this is the number the winning shop watches and the guessing shop has never computed. Margin percent and GMROI answer different questions, and conflating them is one of the most expensive habits in retail. A top-shelf flower line can carry a 55% margin, sit for weeks between sales, and return less per invested dollar than pre-rolls at 38% that turn every nine days. Ask "what's our most profitable category?" and margin percent points at the top shelf. Ask "where does a dollar work hardest?" and GMROI points at the pre-rolls. Only the second question is about your money.
The mechanics make it plain: GMROI is effectively margin × turns. A product earns its shelf space either by fat margins or fast cycling — and a shop full of high-margin, slow-turning prestige product can be starving while its boring accessories wall quietly funds the payroll. GMROI is the formula that catches it, category by category, and it's built entirely from numbers already in the POS.
The discipline: run it quarterly, by category, and let it argue with your instincts. The shelf you're proudest of and the shelf that pays the rent are not always the same shelf — and until you compute this, you don't know which is which.
05 / protect
Inventory doesn't lie
Every formula so far has been about making money. This one is about keeping the license.
Variance analysis, the general tool. At its root it's the simplest formula in the piece:
Budgeted $42,000 of flower COGS this month, spent $47,300 — a $5,300 unfavorable variance. So far, so obvious. The analytical move — the difference between noticing a problem and diagnosing it — is the decomposition. Any spend variance splits into two components:
Same $5,300 miss, two completely different stories. If it's price variance, your cost got away from you — the wholesale market moved, or your buyer's negotiating leverage slipped. If it's volume variance, you moved more product than planned, which might be good news wearing a red number. The total tells you that something happened; the decomposition tells you which conversation to have, and with whom. A shop that only looks at totals is forever having the wrong meeting.
Then there's the variance that isn't about money. Run the same logic on units instead of dollars:
In any other retail category, this number is called shrinkage, it's managed against an industry tolerance, and it's a cost of doing business. In cannabis it is a regulatory event. Your system count isn't an internal ledger — it's METRC, the state's seed-to-sale track-and-trace, and the state considers that number the truth. When your physical count and METRC disagree, the discrepancy isn't an accounting annoyance; it's a question the regulator is entitled to ask, and "we're not sure" is the answer that ends licenses.
This reframes the entire discipline. Reconciliation isn't month-end hygiene — it's continuous defense. The decomposition instinct applies here too, because an inventory variance has a taxonomy: receiving errors (the manifest said 500 units, 480 arrived, nobody counted), unit conversion drift (grams recorded against eighths, the slow bleed of rounding), waste and destruction logged incorrectly or not at all, sampling and display product that never got booked out — and, at the end of the list, actual diversion. Product walking out the back.
The order of that list matters. The winning shop works it top to bottom, because most variance is process error, not theft — but it works the list fast, because every day an unexplained variance ages, two things compound: the trail goes colder, and the number sits in the state's system looking exactly like diversion regardless of what it actually is. The guessing shop finds its variance at inspection time, when the regulator is the one holding the count sheet — and by then the innocent explanation and the guilty one are indistinguishable.
One operational habit separates the two shops, and it costs almost nothing: cycle counting. Not the annual full-inventory nightmare — a small rotating slice, counted daily or weekly, highest-risk categories most often. It converts reconciliation from an event into a rhythm, catches drift while it's still one bad manifest instead of a quarter's accumulation, and — this is the part that matters at inspection — produces a documented history of self-auditing. When the regulator asks about a discrepancy, the shop that can show its counting cadence and its resolution log is having a very different conversation than the shop that's seeing the number for the first time alongside the inspector.
The formula is trivial. The discipline is what stands between a paperwork error and real regulatory scrutiny.
06 / grow
The math after the sale
Everything so far protects the shop. This last pair of formulas grows it — and in cannabis, they carry more weight than they do anywhere else in retail, for one structural reason: the normal acquisition channels are closed. No Google ads, no Meta campaigns, no billboard next to the highway in most markets. You cannot buy your way to new customers the way every other retailer can. Which means the customer already standing at your counter isn't just a transaction — they're the entire growth strategy, and these two formulas are how you count them properly.
Customer lifetime value: what a regular is actually worth.
Take a customer who spends $55 a visit, comes twice a month, at your 40% margin, and stays loyal for three years. Run the multiplication and that person is worth roughly $1,600 in gross margin — not the $22 the register sees today. Every retention decision prices against that gap. Is a loyalty program worth running? A 10%-off structure looks expensive against a $55 ticket and trivial against a $1,600 relationship. Should a budtender comp a pre-roll to fix a bad experience? Against the ticket, that's margin bleeding; against the CLV walking out the door annoyed, it's the cheapest save in the building.
And notice which lever in that formula is live. Average order value moves slowly. Margin is set upstream. Lifespan is partly out of your hands. Frequency is the one a shop can actually move — and in a category where the product is consumable and the repurchase cycle is measured in weeks, a customer who comes twice a month instead of once isn't twice as valuable on paper. They're twice as valuable in fact.
Attach rate: the honest cross-sell.
Of every hundred flower baskets, how many also carry papers? Batteries with carts? That's attach rate, and it's the formula that turns cross-selling from a vibe into a measurement. If flower attaches papers 40% of the time, that number is a fact about your customers' actual needs — and the 60% who didn't attach include people who bought flower and are about to get home without anything to roll it with.
That's the reframe that keeps this ethical, and it's worth stating plainly because "upsell" is a word that deserves its bad reputation: a data-grounded attach isn't pushing product, it's finishing the sale the customer already started. "Do you need papers with that?" asked because the data says four in ten do, is service. Pushing a battery on someone who buys one every visit is noise. The attach rate tells the budtender which question is which — and a counter team coached on real attach data sells more and annoys fewer people, which is the only version of upselling worth doing.
The winning shop's top line, it turns out, is built from the same discipline as everything upstream: count what's actually happening, and act on the count.
close
Two shops, one ledger
Go back to the avenue. Same brands, same prices, same block. The difference was never on the shelf.
The shop that's fine forecast next week from its own history, bought its heroes deep and its experiments tight, priced against the breakeven that includes the silent partner, watched GMROI instead of margin percent, counted a slice of its inventory every single day, and treated the customer at the counter like the $1,600 relationship the math says they are. Fourteen formulas. None of them beyond a spreadsheet. All of them done on purpose.
The shop that's underwater isn't run by worse people. It's run on feel — and this industry, more than any retail category in America, is engineered to punish feel: the product decays, the tax code confiscates, and the regulator reads your ledger. Arithmetic is the survival trait.
Nobody opens a dispensary to do math. But the ones still open in five years will be the ones that did.
// end of file
