How to Size Kanban (Number of Cards)

Kanban sizing is calculating how many kanban cards (or containers) to put in a replenishment loop so the line never starves and inventory never bloats. The standard formula is N = (D x T x (1 + X)) / C: demand rate times replenishment lead time, padded by a safety factor, divided by container size, rounded up.

Too few cards and the line runs dry waiting for parts. Too many and you have hidden a warehouse of work-in-process inside your pull system, which defeats the point. Sizing kanban is the discipline of picking the smallest card count that still protects flow. This post gives the formula, walks a full calculation, and explains how to set the one number people always guess at, the safety factor. It is educational and names no products.

What is kanban sizing?

Kanban sizing is deciding the number of kanban cards or containers that circulate between a using process and its supplying process. Each card authorizes the replenishment of one container of parts. The card count sets a hard ceiling on inventory in that loop: with ten cards and a container of 144 pieces, the most that can ever sit in the loop is 1,440 pieces, because there is no eleventh card to trigger an eleventh container. That is the quiet power of a pull system, inventory is capped by design, not by hope.

The whole job of sizing is to make that cap tight without making it dangerous. A loop needs enough cards to cover everything demanded during the time it takes to replenish, plus a cushion for the days demand runs hot or the supplier runs slow. Everything past that is waste you can see and remove. Kanban sizing is one of the concrete calculations behind a working kanban system and it sits inside the broader discipline of lean manufacturing.

What is the kanban sizing formula?

The most widely taught kanban sizing formula is N = (D x T x (1 + X)) / C, where N is the number of cards, D is the demand rate per unit of time, T is the replenishment lead time in the same unit, X is the safety factor as a decimal, and C is the number of parts per container. You compute the raw result and always round up, because a fraction of a card cannot exist and rounding down would leave demand uncovered.

Notice what the formula is really saying. The product D x T is demand during lead time, the pieces the line will pull while a replenishment is in flight. The term (1 + X) inflates that by the cushion. Dividing by C converts pieces into whole containers, and each container needs one card. Keep D and T in the same time unit, pieces per hour with lead time in hours, or pieces per day with lead time in days, or the answer is nonsense.

Here is how to size a loop from scratch:

1. Measure the demand rate (D). Use the real average pull at the using process, ideally tied to takt time in pieces per hour or per day.
2. Measure the replenishment lead time (T). Time the full loop: card detach, travel to the supplying process, queue, make or pick, and return the full container. Use the same time unit as D.
3. Set the container size (C). Count the standard pack quantity per bin, tote, or pallet actually used in the loop.
4. Choose the safety factor (X). Pick a cushion for demand and supply variability, as a decimal (0.10 to 0.50 is the common band).
5. Compute and round up. Plug the numbers into N = (D x T x (1 + X)) / C and round the result to the next whole card.
6. Watch, then shrink. Run the loop, track stockouts and full-bin waits, and remove cards one at a time until problems appear, then add one back.

Can you walk through a worked example?

Take a press feeding an assembly cell. The cell pulls 200 pieces per hour. The full replenishment loop takes 12 hours. Parts come in containers of 144. You are early in the rollout, so you set the safety factor at 15 percent, X = 0.15. Plug in: N = (200 x 12 x 1.15) / 144 = 2,760 / 144 = 19.2, which rounds up to 20 cards. Twenty cards at 144 pieces caps the loop at 2,880 pieces, and that ceiling is the number you attack over time.

Change one input and the answer moves in a way you can feel. Cut the lead time in half, to 6 hours, and the raw result drops to about 9.6, so 10 cards, half the inventory. That is why quick changeover and shorter loops matter so much: every hour you shave off T shaves cards off the loop and cash off the floor. The formula is not just a sizing tool, it is a scoreboard for improvement.

How do you set the safety factor?

The safety factor covers the gap between average behavior and bad days: demand that spikes, a supplying process that hiccups, a container that gets miscounted. A factor of 0 assumes perfectly steady demand and flawless replenishment, which no real loop has. Common practice is to start a new loop around 20 percent (X = 0.20) while you learn the process, then ratchet it down toward 10 percent as demand and supply prove stable. Highly variable demand or an unreliable supplier can justify 30 to 50 percent at first, but a high cushion is a signal, not a solution.

Treat X as a debt you intend to pay off. A large safety factor is inventory covering for a problem you have not fixed yet, unpredictable demand, long or erratic lead times, poor supplier reliability. The lean move is to attack the root cause, shorten and stabilize the loop, then lower X and pull cards. Sizing is never one and done; you resize whenever demand shifts a season, a container standard changes, or the loop gets faster.

What do the standards and data say?

Context from bodies of knowledge and primary data:

• Kanban and pull systems are defined in the supply-chain body of knowledge maintained by the Association for Supply Chain Management (ASCM/APICS) which frames the kanban card as the signal that authorizes replenishment and caps work-in-process in a loop.
• Kanban originated in the Toyota Production System as a visual signal to trigger pull rather than push, a method now standard across lean production and taught as the counterpart to level scheduling and heijunka.
• The base of operations running these loops is large: the Bureau of Labor Statistics reports roughly 13 million manufacturing jobs in the United States, across plants that each set their own replenishment parameters.

The practical takeaway from the standards is simple: the card count is a policy decision, not a law of nature. You choose it, you own it, and you are meant to keep shrinking it.

What are the common sizing mistakes?

The first mistake is measuring lead time as machine cycle time only. The loop includes travel, queue, and return, and those are usually the biggest chunk. Time the whole loop with a stopwatch, not the run time on a router. The second is mixing time units, pieces per hour with lead time in days, which inflates the count wildly. The third is setting a fat safety factor and forgetting it, so the loop quietly holds a week of stock nobody chose. The fourth is never resizing after demand shifts, so a loop tuned for last year's volume starves or floods this year's.

Sizing kanban well also depends on smooth demand feeding the loop. Wild swings in the pull signal force a bigger safety factor, so leveling the schedule with load leveling and level scheduling is not a separate project, it directly lowers the card count you can safely run. Stable demand upstream is cheaper than safety stock downstream.

Where kanban sizing lives or dies: the data underneath

A kanban loop is only as good as the numbers behind it, and those numbers drift. Demand climbs a season, a supplier's lead time creeps, a container standard changes on the floor but not in the sizing sheet, and the card count that was right in January starves the line by June. When the sizing lives in a spreadsheet nobody revisits, the loop silently goes wrong and a supervisor patches it with an off-book pile of extra bins, the exact waste kanban was supposed to kill. Harmony is an AI-native layer that connects machines, software, and paperwork into one operational layer, with no rip-and-replace, so the live signals a loop depends on, real pull rate, actual replenishment time, current container standards, become one current record instead of a stale sheet. AI search returns cited answers across those records, so a planner can ask what a loop's real lead time has been this quarter or where cards keep running short and get a grounded answer, and Harmony's digital workflows keep the sizing tied to what the floor is actually doing. It is the same paper-to-digital move Harmony makes elsewhere in the plant (see the CLS case study): the card count stops being a forgotten cell and becomes a number you can trust and keep shrinking.