Why Putting Your Fastest Movers Together Slows the Whole Line Down

Most warehouse optimisation is sold on top of a model of the building that quietly assumes one person is in it.

Here is the advice, and it is good advice as far as it goes. Work out how often each item is picked. Rank the items. Take the busiest ones and give them the closest, easiest slots, near packing, at a height you can reach without bending or climbing. Every slotting pitch says a version of this, and every slotting pitch is right, because travel is the single largest piece of a picker’s day and shortening it is the whole point of the exercise.

Now read the instruction again and notice the word doing the quiet work: the busiest ones. Together. In the same small, convenient region of the warehouse. That is where the advice stops being safe, and the reason is not subtle once you see it. You have just taken every item that a lot of people need, and you have put them all in the one place everybody has to go.

You optimised for distance. You forgot that other people were walking too.

What the distance model can see, and what it cannot

A travel-distance optimiser does an honest job on the question it is given. It takes your slots and your pick history, it works out the path length for the orders, and it rearranges the slots so the paths get shorter. The report at the end is real. The routes really are shorter. If you measured one picker walking those routes on an empty Sunday floor, the number would hold.

The trouble is the floor is not empty on a Tuesday. It has eight or twelve or thirty people on it, and they are not walking independent private routes. They are converging, over and over, on the same short list of popular items, because that is what popular means. And when two of them arrive at the same slot, or meet in the same aisle, at least one of them stops.

That stop is the cost the distance model cannot price. Not because the model is badly built, but because of what it measures. It measures distance travelled. A picker standing still at the mouth of an aisle, waiting for a colleague and a pallet jack to clear, has travelled no distance at all. On the report, that picker is doing beautifully. The route was short. The map is green. And the shift is running slow for a reason the map has no symbol for.

This is why the problem survives so well. It hides in the gap between distance and time. The distance got shorter and the time did not, and every dashboard on the wall is a distance dashboard.

The name for the second cost

The thing the distance model cannot see has a name, and Bartholdi and Hackman, in the Georgia Tech textbook Warehouse & Distribution Science, give it two.

There are two types of congestion to which order-picking is susceptible. Interference at a location, when both pickers want to pick from the same small area of the warehouse. Interference in an aisle: when one worker wants to pass another but is unable to because of the narrowness of the aisle.

Read those two definitions against the advice we started with. Interference at a location is two pickers wanting the same slot. Interference in an aisle is two pickers needing to pass in the same lane. Both of them get worse the more people want to be in the same place at the same time. And what did the standard advice do? It took the items the most people want, and it put them in the same place.

The authors say the next part plainly, and it is the sentence that turns the whole instruction on its head:

Locations of popular SKUs are susceptible to both types of congestion because many pickers will stop there. And the congestion will be even worse if the product is requested in large quantities because order-pickers will take longer at that location.

So the very property that earns an item a front slot, that a lot of people pick it, and pick a lot of it, is the same property that makes concentrating it expensive. Velocity is the argument for pulling it forward. Velocity is also the argument for not pulling it forward into the same square metre as everything else fast. The two pressures point in opposite directions on the identical fact, which is exactly why, as the same book puts it elsewhere, slotting is hard: you are balancing more than one goal on one placement at the same time.

A real building that learned this

The book does not leave it as theory. It hangs it on a specific distribution centre, and the case is worth reading in full because the mistake is a smart one, not a careless one.

These issues could be seen at an auto parts distributor in Indianapolis. Their philosophy was to let orders accumulate to get pick-density, and then use many pickers on a single shift. To reduce congestion in picking, they distributed the most popular product throughout the warehouse on the ground level.

Notice what these people did. They already understood congestion. They were not naive. They saw that concentrating their most popular product would jam their pickers, so they did the opposite of the naive move and deliberately spread that product out across the building. Sensible. It worked, for picking. And then:

One problem with this approach was that it created congestion on the dock, as freight had to be staged and sorted before loading.

There is the lesson, and it is a harder one than "do not cluster your fast movers." Congestion did not disappear when they spread the product out. It moved. It came off the pick aisles and landed on the dock, because now the same order was being assembled from all over the building and everything had to be staged and re-sorted before it could be loaded. They pushed the bubble in and it came out somewhere else.

The book even offers the alternative they might have chosen instead, which is not a slotting change at all:

An alternative approach would be to reduce congestion by using fewer order pickers over two shifts, which would allow concentration of popular product into fewer aisles, for less travel.

Fewer people, more hours, tighter placement, shorter walks. A completely different answer to the same problem, reached by changing the shift pattern rather than the slots. That is the real shape of this decision. There is no single lever. There is travel, there is congestion in the aisles, there is congestion on the dock, and there is how many people you put on the floor at once, and they are all connected, and moving any one of them moves the others.

Why this comes off lines per person hour, and hides while it does

Everything in this cluster has to land in the same place, the number the floor actually lives by: lines picked per person hour. Congestion lands there hard, and it lands there invisibly, which is the dangerous combination.

Here is the mechanism, step by step. A picker reaches a busy slot. Someone is already there, or a pallet jack is parked across the aisle. The picker waits. Ten seconds, thirty seconds, a minute, it does not matter, the point is the clock is running and no line is being completed. Multiply that by every convergence, across every picker, across a whole shift, and it is a real subtraction from the lines-per-hour figure. It is not exotic. It is people standing near each other, waiting, all day.

Now the invisible part. That lost time is recorded nowhere that anyone looks. It is not travel, so the distance report does not have it. It is not a mis-pick, so the accuracy report does not have it. It is not a stockout, so the replenishment report does not have it. It is not idle time in any way a labour standard would flag, because the picker is not on a break, they are at a slot, ready to work, blocked. Ask the labour management system and it will tell you the picker was present and productive-adjacent for the whole shift. The lines-per-hour number carries the loss, and nothing on the wall tells you congestion is the reason the number is lower than the travel model promised.

This is the same nasty shape as the location-accuracy problem in the previous chapter. The system produces a confident number, the arithmetic inside it is correct, and it is answering a different question than the one you think you asked. There the wrong question was where is the stock answered as how much stock is there. Here it is how fast can we pick answered as how far do we walk. Both look authoritative. Both are missing the same kind of thing.

An illustration of the arithmetic, and why concentration hurts worse than it looks

A worked example makes the pressure concrete, and worked properly it shows something the intuition misses. The numbers here are invented and labelled as such, so nobody mistakes them for a benchmark. What is not invented is the mechanism, which the book states directly: the simplest model of congestion is the chance that two pickers are in the same aisle at the same time, and for any one picker the chance of being in a given aisle is roughly proportional to that aisle’s share of the picks.

Take a floor with ten pickers and say your popular items, once concentrated, put forty in every hundred picks into a single aisle. The first, easy calculation is how many pickers you would expect to find in that aisle at any moment: ten pickers, each with a forty in a hundred chance of being there, gives an expected four. Spread those same items so the aisle carries only fifteen in every hundred picks and the expected number in it falls from four to about one and a half. That is the version most people stop at, and it already argues against piling everything into one lane.

But it understates the problem, and the reason is worth following, because it is the difference between a slide and an argument that survives scrutiny. Interference is not a property of one picker being somewhere. It is a property of two pickers being in the same place at the same time. One picker alone in a busy aisle interferes with nobody. So the quantity that actually tracks congestion is not how many pickers are in the aisle, it is how many pairs of them are, and pairs scale with the square of the concentration, not with the concentration itself.

Run the same three cases on that basis. With ten pickers there are forty five possible pairs. At forty in a hundred, the expected number of pairs both sitting in the hot aisle at once is forty five times zero point four squared, which is about seven. Drop the concentration to twenty five in a hundred and it falls to about two point eight. Drop it to fifteen in a hundred and it falls to about one. So cutting the aisle’s share of the picks from forty to fifteen, a fall of less than three times, cuts the expected interference from about seven to about one, a fall of roughly seven times. Concentration does not cost you in proportion to how concentrated you are. It costs you in proportion to the square of it.

The book’s own preferred measure tells the same story in plainer terms. Compute the chance that the hot aisle holds two or more pickers at once, which is the exact quantity it names, and across those same three cases it moves from about ninety five in a hundred, to about seventy six, to about forty six. At full concentration the busy aisle almost never has fewer than two people in it. That is not an aisle, it is a queue.

That is the whole trade, and now it has teeth. Concentration buys you shorter walks and sells you a crowded aisle, and it sells the crowding to you on square terms while it buys the walking back on linear ones. Spreading buys you an emptier aisle and sells you longer walks, and, as Indianapolis found, possibly a crowded dock instead. Neither number is the answer on its own. The answer is the placement that makes the sum of walking and waiting smallest, and you cannot find it by optimising walking alone, because the tool that optimises walking alone cannot see the waiting, and even if it could, it is priced on the wrong curve.

The price of a collision is set by your aisles, not your slots

Everything above measures how often pickers meet. What it costs when they do is a separate number, and it is fixed by the building, not by the slotting plan. This is why "do not concentrate too much" cannot be turned into a rule you copy from a slide, and it is worth being precise about what sets the price.

The dominant term is aisle width. In a wide aisle, two pickers meeting is a half-second sidestep and the cost rounds to nothing, so a fair amount of concentration is free. In a very narrow aisle, the kind run with wire-guided or turret trucks to squeeze more racking into the footprint, two pickers cannot pass at all. One of them has to reverse out the way they came, which can mean abandoning and restarting a pick run. The same collision that costs a sidestep in one building costs a reversed truck and a lost run in another, and the narrow-aisle building is usually the one that chose narrow aisles precisely because space was tight, which is the same pressure that makes it want to concentrate. The buildings least able to afford congestion are the ones most tempted into it.

Equipment sets the second term. A selector on foot with a cart can squeeze past. A picker on a pallet jack, a rider, or a turret truck occupies and blocks the whole lane, and takes longer at the face because larger equipment is slower to position. The book makes exactly this point when it notes the congestion is worse where product is picked in large quantities, because the picker "will take longer at that location." Big equipment and big grabs are the same problem: they extend the time one picker holds the contested space, and every second they hold it is a second the next picker waits.

The practical consequence is that the honest slotting answer is not a global setting but a function of where you are standing. In a wide-aisle, cart-picked building, cluster hard and barely worry. In a narrow-aisle, pallet-jack building running many pickers at once, the point where concentration turns expensive arrives early, and the top few items want spreading before almost anything else does. A tool that hands both buildings the same ranked list has told you it does not know which building it is in.

The lever nobody calls a slotting lever: when you release the work

Here is the part that is genuinely easy to miss, because it sits one step outside the thing everyone is looking at. Congestion is not only a function of where the work is. It is a function of when you release it, and that lever has nothing to do with slotting at all.

Look again at what the Indianapolis distributor was doing before the slots ever entered the story. Their philosophy, in the book’s words, was "to let orders accumulate to get pick-density, and then use many pickers on a single shift." Read that as a congestion decision, because that is what it is. Letting orders accumulate and then releasing them in one dense wave means a lot of pickers start similar work at the same moment, which means they converge on the same popular slots at the same moment. The release policy was manufacturing the very simultaneity that the slotting change then had to fight. They were pressing the accelerator and the brake together.

This matters because it reframes the whole problem. Slotting decides the map of where collisions can happen. Wave composition and release timing decide how many pickers are on that map at once, doing how similar a set of tasks. Staffing decides the third thing, the raw number of bodies in the aisles. All three multiply. You can have flawless slots and still jam the floor by releasing five hundred near-identical orders in one wave, and you can rescue a merely-decent slotting plan by staggering releases so pickers spread across the clock instead of stacking on top of each other. The book’s own alternative for Indianapolis was a staffing-and-shift change, "fewer order pickers over two shifts," not a slotting change, which is the same insight from the other side: the cheapest cure for a congestion problem is often not to move a case at all.

None of this is an argument against pick density, which is a real efficiency and the reason waves exist. It is an argument that density and congestion are the same coin, and that a slotting tool optimising in isolation is tuning one instrument in a trio while insisting it is a solo. If your floor jams, the fix might be in the slots, or it might be in the wave template, or it might be on the shift roster, and only one of those three is a slotting question.

So what do you actually do

The point of this article is not to leave you unable to move. Concentrating fast movers is still broadly right. It is the unbounded version that fails, the one that keeps pulling items forward with no counter-pressure. The counter-pressure has a few practical forms, and all of them are in the source.

Spread the very hottest items across a few slots on purpose. This is the direct remedy the book gives, and it is worth quoting because it is the opposite of the instinct:

By storing popular products in multiple forward locations there is less chance that a picker has to wait for a colleague to move out of the way.

Your single busiest item does not want one perfect slot. It wants two or three good ones, so that two pickers needing it do not have to queue for the same face. You are spending a little space and a little extra travel to buy back the waiting. For the top handful of items that trade is almost always worth it.

Treat "put the fast movers together" as a pressure, not a rule. The book’s own summary of the goal already carries the caveat built in: squeeze more product into the space, put popular and heavy items where they are easiest to pick, and at the same time, in the authors’ words, "one wants to avoid creating congestion by concentrating popular items too much." The phrase "too much" is the entire job. Where "too much" begins depends on your aisle widths, your equipment, and how many people you run at once, which is why it is a judgement about your building and not a setting you can copy from a slide.

Reach for the levers that are not slotting levers. As above, the two that move congestion without moving a case are release timing and staffing. Stagger the wave so pickers spread across the clock instead of starting the same dense work at the same moment, or, as Indianapolis was advised, run fewer pickers over more hours so you can concentrate for short walks without the crowd. Sometimes the cheapest cure for a jam is a change to the wave template or the roster, and a tool that only rearranges slots will never propose it, because it is not in the space of moves the tool can make.

None of these is exotic and none of them requires new software. They require someone to hold travel and congestion in view at the same time, which is precisely the thing a single-objective optimiser is built not to do.

This is not one consultancy’s opinion

Because the whole argument leans on one textbook, it is fair to ask whether anyone who does this for a living agrees. They do. The independent consultancy F. Curtis Barry & Co, in their list of the ways slotting efforts go wrong, name naive velocity-only slotting creating congestion by clustering the fast movers into one aisle as one of the aspects that hurts your results. That is a practitioner firm and an academic text arriving at the same warning from opposite directions, which is about as much corroboration as a claim like this can carry.

It is also, quietly, the reason to be careful about any tool or any pitch that hands you a single ranked list and calls it optimised. A single ranked list is the output of a single objective. Ask it one question: what does it do when two pickers want the same aisle at the same time? If the honest answer is that it does not model that, then it has not optimised your warehouse. It has optimised a warehouse with one person in it, and then handed you the plan as though the other eleven were not real.

And congestion is not even the only thing fighting velocity

Congestion is the cleanest counter-pressure to velocity because it comes from the same fact, popularity, that velocity is built on. But it is worth widening the lens one more turn, because the deeper point is not "there are two objectives, travel and congestion." It is that there are usually several, they vary by building, and a ranking built on any single one of them is blind to the rest.

The same textbook gives a placement pressure that has nothing to do with either travel or congestion:

Some distribution centers supporting retail drug stores prefer to store similar-looking medicines apart to reduce the chance of a picking error; but they store non-drug items in the same product family so that, for example, all the hair care products will tend to be picked together.

Read what that placement is optimising. Storing lookalike medicines apart is optimising for accuracy, accepting worse travel to avoid a dangerous mis-pick. Storing a product family together is optimising for the customer’s put-away labour, so the receiving store opens a container already sorted by aisle, and it accepts worse travel in your building to save labour in thousands of theirs. Neither has anything to do with how far your picker walks, and a velocity ranking would quietly undo both, pulling the fast medicines together and scattering the slow ones out of their family, and it would call the result optimised.

This is the general shape of the thing this whole cluster keeps circling. A warehouse is not solving one problem. It is balancing travel against congestion against accuracy against the receiver’s labour against how often you are willing to re-slot, all on the same set of shelves, at the same time. That is what the book means when it says slotting is hard because you must satisfy multiple goals and constraints simultaneously. It is also why "optimised" is a word to be suspicious of when it arrives attached to a single sorted column. The single column is not the answer to the warehouse’s problem. It is the answer to one of the warehouse’s problems, printed as though it were the only one.

How to see it on your own floor this week

You do not need a model to find out whether this is costing you. You need ten minutes and a place to stand.

Stand at the mouth of your single busiest aisle during a normal picking window and just watch. Count how many times in ten minutes a picker has to stop, slow, or reroute because someone else is in the way, at a slot or in the lane. If the answer is "hardly ever," concentration is not hurting you and you can stop reading. If you lose count, you have found time that every distance report you own is structurally unable to show you, and it is coming straight off your lines per person hour.

If you want a number you can put in front of a steering committee rather than a story about standing in an aisle, congestion leaves a signature in the labour data you already collect, and it is this: lines per person hour that falls as you add people to the floor. Pull your pick history and work out the average lines per picker per hour on a lightly staffed shift, then on a heavily staffed one. If five pickers each average a hundred lines an hour and ten pickers each average eighty, you did not lose that twenty lines to bad luck. Ten people picking eighty is eight hundred lines where ten people picking a hundred would have been a thousand, and the missing two hundred is what interference costs when you double the bodies without spreading the work. Perfectly independent work would scale flat, each picker holding their rate no matter how many others are on the floor. Real floors bend below flat, and how far they bend is your congestion, measured in the one unit the business already trusts. Be honest about what it does not prove on its own, because a rate that sags under heavy staffing can also be replenishment failing to keep the faces full or supervision thinning out, so the number tells you where to look rather than closing the case. But it is a defensible place to start, and unlike the distance report it is at least pointed at the right thing.

Then look at where your busiest twenty items actually sit. If they are packed into one or two aisles, you have the textbook setup for exactly the interference you just watched. Spreading the top few across another slot each is a small, cheap, reversible move, and it is the kind of change that is easy to test: make it for a fortnight, and watch whether the same aisle still jams.

If you have ever watched a fast aisle jam like this, I would like to know what you did about it, because the honest answer in most buildings is "nobody had time to look," and that is the whole reason it persists.

For whether the location data underneath any of this is even true, read How do you know your warehouse location data is right?.

For what a placement analysis actually needs from your data, read What data do you need for a warehouse slotting analysis?.

For the full ground-up treatment of the problem this sits inside, read Warehouse Pickup Optimization: The Operator’s Guide.