Kanban for Order Fulfillment: Build a Board Your Pick-and-Pack Team Will Use

Kanban for order fulfillment is a visible board that shows every order between “received” and “out the door”, with a hard cap on how many orders are allowed in each stage at once. The cap is the part that does the work. The board just makes it obvious when you have broken it.

Kanban did not start in software. Toyota built it in the 1950s as a shop-floor signalling system for just-in-time production, where a card moving backwards down the line was the instruction to make more (Kanban University). Software teams borrowed it decades later. If you pack and ship physical orders, you are closer to the original use than any engineering team is.

This is the build guide: columns, limits, cards, and the specific ways the board falls apart if you get those wrong. For the wider question of running an operation on short improvement cycles, see agile methodology in logistics, which covers the strategy this board sits inside.

The Bottom Line

• A fulfillment kanban board has two components that matter: columns matching the real stages an order passes through, and a work-in-progress limit on each one.
• The limit is not a target. It is a stop signal. When a column is full, nobody starts new work there, which forces the team to clear the actual bottleneck instead of piling work in front of it.
• Start with the board reflecting how your team works today, not how you wish it worked. A board that lies is worse than no board.
• Kanban suits continuous, interrupt-driven fulfillment work. For time-boxed improvement projects, Scrum is the better fit.

What a kanban board does that an order list does not

It shows you where orders are stuck, not just how many there are.

Every fulfillment operation already has a list of open orders. The list tells you the total. It does not tell you that eleven of them have been sitting in “picked, waiting for packaging” since yesterday afternoon because you ran out of the medium boxes. That is the information a board surfaces without anybody having to notice it.

Three things follow from making the work visible:

• Bottlenecks announce themselves. A column with a queue in front of it is the constraint. You stop guessing which part of the process is slow.
• Nobody starts work that cannot finish. Picking forty orders when packing can handle fifteen does not produce fifteen shipped orders faster. It produces twenty-five half-finished orders occupying floor space.
• Handoffs get a place to live. Most fulfillment delay is not work time. It is waiting time between people, which is invisible in a task list and obvious on a board.

The columns a pick-and-pack board needs

Map the stages your orders actually pass through. For most food, floral, catering and wholesale operations, six columns cover it.

Received. Orders that have arrived and been validated: address checked, stock confirmed, payment settled. Anything failing validation does not enter the board; it goes to an exceptions lane.

Ready to pick. Validated and scheduled for a specific delivery day or run. This column is your queue, and it is the only one that is allowed to be long.

Picking. Physically being assembled. This is where the first work-in-progress limit goes.

Packing. Assembled and being boxed, labelled and documented.

Staged for dispatch. Complete, labelled, waiting on a vehicle or a route. Growth in this column at 2pm means your outbound schedule, not your warehouse, is the problem.

Out for delivery. Left the building. Many teams close the card here; others keep a final “delivered” column so failed attempts stay visible.

Two structural rules. First, add an exceptions lane running across the board for orders that are blocked: missing stock, bad address, customer unreachable. Blocked orders left in the normal flow silently consume a WIP slot and make the board look busier than the work is. Second, resist adding columns for approvals nobody actually performs. The board should describe the process, and if a stage exists only on paper, leaving it off is how you find that out.

How to set work-in-progress limits for a fulfillment team

A work-in-progress limit is the maximum number of orders allowed in a column at one time. Setting them is the step teams skip, and skipping it turns kanban into a status display.

Start with the simplest workable rule: one to two orders in progress per person working that stage. Two pickers gives a picking limit of three or four. It will feel too tight. That is the point. The limit is calibrated to expose the constraint, not to accommodate it.

Then adjust from what you observe:

• A column is constantly at its limit and the one after it is empty. The limited stage is your bottleneck. Add capacity there, or accept the throughput and stop overloading the stages upstream.
• A column never reaches its limit. The limit is too loose to be doing anything. Lower it by one and see whether anything breaks.
• Work is jumping the limit routinely. Either the limit is wrong for a real peak, or the team has decided the board is advisory. Both need a conversation, and they need different answers.

The counterintuitive result is that capping how much you start increases how much you finish. Work-in-progress limits and pull-based flow are the core of the method precisely because unlimited starting is what produces queues (Kanban University).

What belongs on a kanban card

One card equals one order. Keep the front of the card readable from a few feet away, because that is how it will be read.

Essential on every card:

• Order number and account name
• Delivery date, and the time window if one was promised
• Item count, and a flag for anything fragile, frozen or temperature-controlled
• Any special handling instruction: loading dock access, buzzer code, signature required

Useful but optional: the date the order entered the board, which lets you calculate cycle time without any software at all. In a digital system the card carries more: order quantity, cost, storage location, integration with a WMS so that a scanned movement updates the card automatically (ProjectManager). On a physical board, restraint wins. If the card needs a second side, the board is not where that information should live.

Using kanban signals to reorder packing supplies

The original Toyota use of kanban was replenishment, and that application still works in a stockroom.

Take the supplies that stop a shipment when they run out: boxes in each size, insulated liners, ice packs, labels and tape. For each one, split the stock into two bins. Work from the first bin. When it empties, the empty bin itself is the reorder signal: it goes to the ordering point and you start on the second bin, which holds enough to cover the supplier’s lead time.

That is a two-bin kanban system, and it needs no software and no inventory count. The physical state of the shelf carries the information. Warehouses run the same pattern digitally with barcode scans or weight sensors triggering replenishment automatically when stock hits a threshold (Cleverence), but the manual version captures most of the benefit for a small operation.

Size the second bin honestly. Lead time plus a buffer for the supplier being late, not lead time on a good week.

Physical kanban board or digital kanban board?

Both work. The choice is about where your team actually is during the day.

The common mistake is going digital first because it sounds more serious. Run a physical board for a month, learn what your columns and limits should actually be, and port a board you already trust. Digitizing a process you have not tested yet just makes the wrong process harder to change.

Three numbers a fulfillment kanban board should produce

A board that generates no data is a decoration. These three come almost free.

• Cycle time. Days or hours from a card entering “ready to pick” to leaving “staged for dispatch”. Write the entry date on the card and subtract. This is your honest internal fulfillment speed, and it is usually slower than people guess.
• Throughput. Cards completed per day. Simple, and the only reliable input for answering “can we take this new account?”
• Blocked time. How long cards sit in the exceptions lane, and why. Tally the reasons for a month and you will have a ranked list of what is actually costing you. Address problems tend to top it, which matters given that roughly 5% of last-mile deliveries fail on the first attempt (GoBolt).

That third list is a ready-made improvement backlog. Take the top item, change one thing, and watch cycle time for two weeks.

How fulfillment kanban boards usually fail

The limits are ignored. By far the most common. The board becomes a display of work rather than a control on it, and the queues reappear behind a very tidy wall.

Every order becomes urgent. An expedite lane is fine and sometimes necessary. Three expedite lanes means you have rebuilt the unprioritized queue you started with.

The board stops matching reality. During the busiest week, updating cards is the first thing dropped, and the board never recovers. Assign updating to the stage handoff itself, so that you move the card when you hand over the order rather than later. That costs seconds rather than a task at day’s end.

Nobody looks at it together. A daily five-minute review in front of the board is what turns it from information into decisions. Without that, the queues are visible and still nobody clears them.

Frequently Asked Questions

Is kanban better than scrum for order fulfillment?

For daily order flow, yes. Fulfillment is continuous and interrupt-driven, which is the exact shape kanban was designed for, and it adds no roles or new meetings. Scrum suits non-software teams when there is a goal worth protecting a fixed sprint for, like rebuilding a pick sequence or preparing for a seasonal peak. Plenty of operations run both: kanban for the daily work, a sprint for the project.

How many columns should a kanban board have?

As many as your process has real stages, which for fulfillment is usually five to seven. Fewer than four and the board hides the handoffs that cause delay. More than eight and cards spend more time being moved than worked on.

What if order volume spikes seasonally?

Raise the limits deliberately and temporarily, and write down that you did. The wrong move is ignoring the limits during peak without recording it, because then you have no idea whether the season or the abandoned limits caused the slowdown. Raising a limit is a decision you can review; ignoring one is not.

Does kanban work with only two or three people?

Yes, and it is arguably more useful, since a small team has no capacity to absorb a bottleneck. With two people the board is mostly about seeing which stage is drowning today and moving the second person there.

Do we need to fix our process before building a board?

No. Build the board around the process you have, badly and all. The board’s first job is diagnostic. Trying to design the ideal flow before you can see the current one is how teams end up with a beautiful board describing a process nobody follows.

Getting the first board up

Draw six columns on a whiteboard. Write today’s open orders on sticky notes and put each one in the column it is in right now. Do not tidy it. The mess is the finding.

Add a limit to the picking and packing columns, one or two per person, and hold a five-minute review in front of the board at the same time every morning. Two weeks of that will tell you more about where your fulfillment time goes than a month of analysis, and it costs a pack of sticky notes.

If the exercise raises a bigger question, namely whether a new service or delivery offer is worth running at all rather than how to run it, that is a different decision, and the Lean Startup method is the way to test it before committing. And if the friction you keep finding is on the customer’s end rather than yours, design thinking for customer experience is the method for mapping that side properly.