Flow and pull: make what's needed

Why work waits in a queue

A batch is a group of items made or moved together. In batch-and-queue work, each step runs a big lot at its own best rate, then the lot moves on, or waits for the next step to be free. The waiting between steps is the queue.

Queues form for three reasons, and most operations have all three:

1. Items wait for their lot-mates. Nothing moves until the whole lot is done.
2. Steps run at different speeds. The fast step piles work up in front of the slow one.
3. Each batch has an overhead, such as a setup or a lab test, so people make the batches bigger to spread it out.

There is a simple rule for how much a queue costs. It's called Little's law, from a 1961 paper by John Little. In a steady system, the average time work spends in the system equals the average amount of work in it, divided by how fast it comes out. So lead time = work in progress ÷ throughput.

Suppose 600 packages are waiting in the vault and about 200 ship each day. An average package waits 600 ÷ 200 = 3 days. If you want packages to wait less and don't want anyone to work faster, you hold less stock waiting. The rule is about averages in a steady system, so treat it as a way to see the cost of a queue, not as a forecast for one package.

Moving work in smaller pieces

Continuous flow means work moves one item at a time, or in a small consistent batch, and each step makes only what the next step asks for. According to the Lean Enterprise Institute's history, the early version was Henry Ford's 1913 Highland Park line, which Toyota later reworked to handle variety.

Small does not have to mean one. Your operation sets a limit on how small a batch can be. A lab test is done per batch. The state's seed-to-sale guidance sets a maximum test batch of 15 lb of dry flower, which is guidance from 2021 and may change, so check the current version. Tags, rooms and crews matter too. So the question is not "how do we get to a batch of one?" The question is "is this batch the size the work needs, or the size we got used to?"

Lean also doesn't mean zero inventory. Some stock is there on purpose, and some waiting is required. Cure and the test hold are required waits. The job is to know which stock and which waits are on purpose, and to take out the rest.

Small batches are only affordable when the overhead per batch is small. That is why Lesson 7 is about shortening changeovers.

Push, pull and the signal

Push means making work in big batches against a forecast and sending it downstream whether or not anyone is ready for it. Pull means the step downstream signals what it needs, and the step upstream makes only that.

Three tools show up whenever people talk about pull:

• Kanban is a signal, such as a card, a bin or an electronic message, that authorizes making or moving something. The word is Japanese for "sign." A kanban is a tool for pull. It isn't pull itself. You can run cards and still push.
• A supermarket is a stock point with a set location and a set maximum for each item. When something is taken, the empty space is the signal to replace it.
• A first-in, first-out lane is a marked lane of fixed size between two steps. The step upstream stops when the lane is full, so work can't pile up without limit. The oldest item is always at the front.

Here is what a supermarket could look like in packaging. A marked section of the vault has one slot for each strain and size, with a written maximum for each slot. When packages leave for a customer, the empty slot tells packaging what to make next. Nobody has to guess, and nobody makes a strain that isn't being used.

Where the signal could come from, by area

These are starting ideas to look for on your own floor, not a design.

Cultivation is the longest pull loop, because a plant takes weeks. That is why that area plans back from the shelf, and why the worked example below does its arithmetic in that order.

Orders are not demand

Pull only works if the signal is true. The signal people reach for first is the order, and an order is not demand. A purchase order reflects the buyer's rhythm, a price offer, or a fear of running out. It is a batched and distorted echo of what actually sold.

The distortion has a name, the bullwhip effect. Small swings in what shoppers buy become bigger swings in what stores order, and bigger again at each step back up the chain. A well-known 1997 paper by Hau Lee and colleagues traced it to four causes: how each party processes the demand signal it receives, rationing and gaming when supply looks short, ordering in batches, and price changes.

So the best signal you can get is the one closest to the customer: what actually sold, item by item. In a store, that is sales by item and the days of stock on hand. One warning: when an item is out of stock, its sales drop to zero, and that looks like no demand. A stockout hides demand.

When an order looks unusual, such as three times the normal size, check the store's sales and stock first. Don't read the order as demand until you have.

Supply matters here too. In February 2026, trade press reported that Massachusetts regulators were considering a freeze on new cultivation licenses. Whatever the market does, when more is made than sold, the extra sits in the vault as the most visible waste.

Two clocks, and a third if you grow

Two different clocks run on every lot of cannabis, and they don't run together.

• The test clock. A test result is valid for one year. After that, the product is expired until it's retested (935 CMR 500.160(5)). This one is a rule.
• The freshness clock. This one is not a rule. It's chemistry. In one small study of two varieties from one lab, terpene levels fell by more than half within four months in every storage condition tested, and THCA fell by about a third at the warmest one, 25 °C. A study of hemp stored in vacuum-sealed foil bags found total terpenes down roughly a fifth to a quarter over 12 weeks, with cannabinoids stable. The size of the loss depends on the plant, the packaging and the storage, so don't quote one number. What the studies agree on is the direction: terpenes go faster than cannabinoids.

That is why "first in, first out by harvest date" isn't always right. Two lots in good condition can have very different test dates. The lot whose clock runs out first should usually go first. That's often called first expired, first out. For a lot on the shelf, that clock is usually the test date.

The third clock is a business one, and it applies if you grow. To expand a license tier, a cultivator has to show it sold 85 percent of what it produced over six months. If it sold under 70 percent, the Commission may reduce its tier (935 CMR 500.050(2)). The Commission began applying that review at license renewals in 2026, and in it the Commission looks at total wholesale transfers. The regulation doesn't say how "produced" is counted. A June 2026 Commission memo says staff currently work it out from the harvest weight in the tracking system, less waste and moisture loss, and notes that the regulation doesn't support that formula. Ask your compliance lead how your own renewal is handled.

That doesn't make "grow less" the lesson. A licensed canopy is space a cultivator doesn't want to lose, and a June 2026 Commission memo proposing weight-based tiers says the canopy system rewards cultivators for maximizing output within a fixed footprint. So in practice the question usually isn't how much to grow. It's what the flower should become. It can go out as packaged flower, as pre-rolls, into extraction as concentrates or distillate and from there into carts or edibles, or be sold in bulk wholesale. Each route has its own customer, its own signal and its own clock. Pull applies to that choice: let what each route can sell decide what each lot becomes, and don't push everything down the one route you happen to have. Whether to grow less is a decision for owners and finance, and this lesson doesn't tell anyone to.

The arithmetic: plan back from the shelf

Here is how planning back works for flower. Every number below is made up to show the steps. Yours will differ, and none of these is a benchmark.

Notice the trade-off. Weekly harvests give you less stock and, on average, fresher flower, but they need rooms and crew that can run a harvest every week. A smaller batch isn't free. The point of the arithmetic is to see the trade, not to pick a side.

The example works back from one product, which is how you size a plan. In practice a licensed cultivator often has the canopy already, so run the question forward as well: given the flower that is coming, what should each lot become? Packaged flower is one answer. Pre-rolls, extraction and bulk wholesale are others, and each one is a different customer to pull from.

Knowledge check

Your task: walk one product's pull

This is a walk, not a project. You change nothing.

1. Pick one product in your area.
2. Walk from where it sits finished, back toward the step before, to the place it waits longest.
3. At each place it waits, write in plain words: what tells the person upstream to make or send more?
4. Write whether there is a written maximum for the stock there.
5. Find the oldest unit, and write how you can tell it's the oldest.
6. Write roughly how many days product waits there, and whether the wait is required (R), like the test hold, or other (O).

Record counts and rounded days only. Leave out customer names, prices and volumes.

Open the printable pull walk

Sources

Where the facts in this lesson come from. Facts last checked October 1, 2026. If you find one that is out of date or wrong, tell me.

• Lean Enterprise Institute lexicon: pull production, and the entries for kanban, supermarket, continuous flow, batch-and-queue, push production and first in, first out
• Lean Enterprise Institute: a brief history of Lean
• Little's Law on its 50th anniversary, Operations Research (2011)
• Lee, Padmanabhan and Whang, Information distortion in a supply chain: the bullwhip effect, Management Science (1997)
• 935 CMR 500, adult-use marijuana regulations (effective September 11, 2026): 500.050(2), 500.150(3)(b) and 500.160(5)
• Cannabis Control Commission: seed-to-sale guidance (May 17, 2021), for the test batch size
• Milay and others, Frontiers in Plant Science (2020): storage of cannabis flower, and Goffman and others, Molecules (2025): storage of hemp
• CRB Monitor (February 2026): Massachusetts regulators consider a cultivation freeze, trade press
• Cannabis Control Commission: bulletin on tier relegation (October 30, 2025), and the Commission memo proposing weight-based cultivation tiers (June 22, 2026)