It is one of the first questions that comes up when an operation starts measuring itself: how many lines per hour should a picker complete. The honest answer is that no benchmark exists that fits your warehouse, and that going out to look for one usually leads to the wrong decisions. What does exist is a method for building your own baseline, putting a cost on it and reading it over time. This guide explains what counts as a line, how the measurement window is taken, why industry averages mislead and how to interpret your own series.
What counts as a line
Before measuring you have to define the unit, because two warehouses that both say "lines per hour" are often counting different things.
A line is each row of an order: one item reference, with its quantity. If an order asks for three references, that is three lines, whether one of them ships two units and another forty. A unit, by contrast, is each individual piece picked. And the order is the complete document.
The most common confusion is mixing lines with units. A warehouse shipping full cases and one building mixed cartons can show the same units per hour and a very different line count. Write the definition down, note whether you count cancelled and short lines, and then leave it alone: a baseline is only useful if it is always measured the same way.
How many lines per hour should a picker complete?
There is no correct number, and anyone who gives you one without having seen your warehouse is guessing. Lines per hour depend on variables that change the job completely: how many references the average order carries, how far apart they sit, whether goods are taken from floor level or from height, whether the picker has to weigh, count, label or check batch numbers, and how far they walk between one pick and the next.
Two operations in the same industry, with the same floor area and the same headcount, can differ several times over in lines per hour without either one being badly run.
So the number that matters is yours, measured with your definition of a line, over your real order mix. What you compare against afterwards is not other companies: it is yourself, a month ago.
How the measurement window is taken
A single day tells you nothing. Order picking has seasonality by hour, by day and by month, and a snapshot taken at the peak or in the trough will mislead you in both directions.
The reasonable minimum is four consecutive weeks, recorded by shift rather than by day. Separate effective picking hours from paid hours: if the picker spent two hours in receiving or waiting for replenishment, those are not picking hours, and putting them in the denominator dirties the number.
Record the context of each week as well: promotion, stock shortages, new staff, a power cut. Without that column, three months from now you will look at a dip in the series and not know whether it was the process or the public holiday.
What an hour of picking is worth
Value recovered time using your actual operating costs and measurements of the tasks that change. Distinguish time available for other work, reduced overtime and expenses that can actually be avoided.
With that, the arithmetic is direct: divide the annual cost of the picking team by the annual number of lines and you have your cost per line. That is the figure worth taking to management, because it translates a process improvement into money.
Value space using your own rent or property opportunity cost. Released space can support growth or reorganization; it only saves rent when an actual payment obligation is reduced.
Why the industry average is not a useful reference
The averages in circulation have three problems, and any one of them is enough to discard them.
The first is definition. It is almost never stated what was counted as a line, or whether the denominator was paid hours or effective hours. Without that, the number is not comparable to yours.
The second is the mix. An average groups warehouses with single-line orders together with warehouses whose orders carry forty lines. The average of two very different operations describes neither of them well.
The third is what it gets used for. An external number almost always ends up justifying a decision already taken: if the average is high, it supports asking for investment; if it is low, it supports doing nothing. Either way it replaces the assessment with an excuse. A baseline is built from your data, not from a report.
How to read your own series over time
With four weeks of baseline and weekly recording you already have a series. What you look at is not the weekly value: it is the trend and the spread.
The trend tells you whether the process is improving. A sustained rise over six or eight weeks following a concrete change (relocating the fastest-moving references, fixing the travel sequence, batching orders) is evidence that the change worked.
The spread tells you something else, sometimes more important. If two pickers on the same shift differ widely, the problem is not speed but method, and it is fixed with training and a defined travel path. If the large gap is between shifts, look at replenishment.
A baseline with high spread is not ready to support an investment decision: first you stabilise it, then you compare.
When speed is not the problem
There are cases where raising lines per hour is not what you need, and it is worth saying so before anyone quotes equipment.
If the bottleneck sits in replenishment, a faster picker will wait just the same. If the order reaches the warehouse late, the gain is in the cut-off time, not in picking. If the error rate is high, going faster multiplies errors and the cost of returns eats the improvement.
And if your operation has few lines per order across many different references, picking automation returns considerably less than it promises: equipment pays for itself on repetitive volume. First you organise the process and measure it; robotics only comes in when the assessment shows that the organised process has run out of room.
The number that matters is yours
You will not find the number you are looking for elsewhere, and if you do find it, it will not help. What you can have within a month is your own baseline: a written definition of what counts as a line, four weeks of per-shift recording that separates effective hours, a cost per line, and a series read by trend and by spread. With that in place, any process change can be tested. And if equipment turns out to be needed later, you will be able to say what it is worth instead of assuming it.
Frequently asked questions
A line is each row of an order: one item reference with its quantity, regardless of how many units are taken from it. An order with three references has three lines. It differs from a unit, which is each individual piece picked, and from the order, which is the complete document. Before measuring, put in writing which of the three you are counting.
Four consecutive weeks is the reasonable minimum, recorded by shift rather than by day. Anything shorter fails to capture daily and weekly seasonality. The denominator should hold effective picking hours, not paid hours: if the operator was in receiving or waiting for replenishment, those hours do not count as picking.
Almost never. It is rarely stated what was counted as a line or whether paid or effective hours were measured, and an average lumps single-line operations together with ones whose orders carry forty lines. The useful comparison is against your own series: the same operation, the same definition, a month ago.
Value recovered time using your actual operating costs and measurements of the tasks that change. Distinguish time available for other work, reduced overtime and expenses that can actually be avoided.
When the baseline is stable, the process has already been organised and the volume still does not fit the time window. If the bottleneck is replenishment, the order cut-off time or the error rate, automation does not solve it. The on-site survey is what determines whether equipment is needed at all.