Cannabis course · manufacturing · scorecard
Count it the same way.
The walk sheet shows you the waste. A scorecard tells you whether it's shrinking. These are ten manufacturing measures, from extraction to filled product, each defined so two people would count it the same way, with where the data already lives and what to watch for. Put three on a board first. The rest can wait for the weekly review.
Safety · Quality · Delivery · Cost
Ten measures, grouped the way a board is
Safety always goes first on the board, measured the way your site already measures it. After that, these ten. None of them needs new software: the weights, results and counts are already in your batch records, lab results, weight checks and tracking system.
Quality
| Measure | How to count it | Where the data lives | How often |
|---|---|---|---|
| Batch pass rate, first submission Watch for: counting a batch that passed after rework or resubmission as a pass. It's a first-time fail. | Batches that pass the lab the first time they're submitted, out of batches submitted. | Lab results | Every batch, reviewed weekly |
| Dosing accuracy Watch for: a monthly average. One batch that came back hot disappears in it. Track each batch. | Measured potency per serving against labeled potency per serving, for each batch. | Lab results, the label spec | Every batch |
| Fill accuracy Watch for: a tolerance that moves. Write it down for each SKU and don't change it to make the number look better. | Units inside the weight tolerance, out of units checked. | The weight check on the line | Every run |
| Cart defect and return rate Watch for: counting only returns. Leakers you catch in-house count too. Split found-here from found-by-a-customer. | Defective carts, leakers and clogs, out of carts filled. | Cart defect log, returns | Every run, reviewed weekly |
Delivery
| Measure | How to count it | Where the data lives | How often |
|---|---|---|---|
| Throughput time Watch for: one number with no split. Break it into working and waiting by step, so you can see the wait you can attack. The test hold itself is required. | Days from input pulled from the vault to finished goods released, for each batch. | Transfers in the tracking system, the batch record | Every batch, reviewed weekly |
| Changeover time Watch for: stopping the clock when the machine restarts. Stop it at the first good unit. And never shorten it by cutting a cleaning step. | Minutes from the last good unit of one SKU to the first good unit of the next, by line. | A changeover log, or a stopwatch until you have one | Every changeover |
Cost
| Measure | How to count it | Where the data lives | How often |
|---|---|---|---|
| Extraction yield Watch for: comparing across methods or input grades. Compare like with like. | Output weight divided by input weight, by extraction method and input grade. | Input and output weights in the tracking system | Every batch |
| Potency recovery Watch for: reading yield as recovery. More oil at lower potency isn't a win. Recovery follows the cannabinoids, not the weight. | Cannabinoids out over cannabinoids in: output weight times output potency, divided by input weight times input potency. | Weights in the tracking system, potency from lab results | Every batch |
| Units per labor-hour Watch for: comparing lines that make different products, or ranking people. Use it to plan crews, not to blame. | Good units finished, divided by labor-hours on the line, by line. | Line counts and the time clock | Daily, by line |
| Scrap rate Watch for: one number for the whole building. Split it by step and by reason, or nobody can act on it. | Material written off as unusable, as a share of material put into that step. | Waste records in the tracking system | Weekly |
The daily board
Start with three
A board with ten numbers on it gets read by nobody. Start with one measure for quality, one for flow and one for cost, under your safety measure. Add a fourth only when the team asks for it.
- QFill accuracy, by line, every run. The crew sees it at the weight check, and drift shows up while there's still time to fix it, not after the batch is done.
- DChangeover time, every changeover. Shorter changeovers are what make it practical to build what's ordered instead of long runs. Every cleaning step stays; the clock is on everything else.
- CUnits per labor-hour, by line, every day. For planning crews and spotting a setup problem, never for ranking people. When it drops, ask what got in the way.
Ten minutes, standing up
Run the huddle
- 1Same time, same place, every day. At the board, at the start of the shift. Ten minutes, standing.
- 2Safety first, always. Anything that happened or nearly happened since the last huddle.
- 3Green or red, against a target the team set. No speeches about green. Every red gets the same three questions: what happened, what's one action, who owns it and by when.
- 4Write the action on the board. Tomorrow's huddle starts by checking yesterday's actions.
- 5Send up what the team can't fix. If a red needs another department, it goes to the next level's huddle that day, with a name on it.
If you want help designing the tiers above this one, my free daily management skill walks through it. Once a measure has a few weeks of data, the control chart skill tells you whether a bad day is noise or a real change. For a filling or pre-roll machine, the OEE skill splits lost time into availability, performance and quality from your downtime log.
Free · one page · print it
The manufacturing huddle board
One printed page for one week: safety, the three starter measures with room for two more, a green or red box for each day, and a log for every red: what happened, the action, the owner and the date. Pin it where the huddle happens.
Manufacturing