Smart warehouse / Article
The ROI of Goods-to-Person Picking: Pick Rate, Travel Time and Labour
The economic case for goods-to-person is one subtraction: the walking. Published ranges put skilled manual pickers at 60 to 120 lines per hour with travel consuming 50 to 70 percent of the shift, against 300 to 600 order lines per hour at a single goods-to-person station where the operator never walks. That is a three to five times uplift on the same person. But the payback is only as large as your own travel share, and in Malaysia the labour line is usually the smallest of the three returns rather than the largest. Space and accuracy carry more of the case. This article works through all three, then builds the model in ringgit so you can substitute your own numbers.
The numbers at a glance
Rates are published industry ranges, not a CODETRACE guarantee
01 / The waste being removed
You are not buying speed. You are deleting travel.
A goods-to-person station does not make anybody pick faster. The hand movement is the same. What changes is that the hand movement is all that is left, because the walking between locations has been transferred to a machine. So the entire return depends on how much walking there was to remove, which makes travel share the first number in the model and the only one worth measuring before anything else.
Published accounts put travel high. One vendor analysis states that skilled pickers in a conventional person-to-goods environment achieve 60 to 120 lines per hour, with travel time consuming as much as 50 to 70 percent of each picker's shift. Another puts it in distance terms, noting pickers walking 8 to 12 miles per shift in a traditional layout, spending more time travelling than picking. Those are international figures on operations designed for high line counts, and they are the reason the pick-rate comparisons look as dramatic as they do.
Your number is the only one that matters, and it takes one shift to get. Follow two pickers for a full shift and log their time in three buckets: travelling, picking, and everything else, which includes waiting for replenishment, scanning, sorting at the trolley and looking for stock that is not where the system says it is. If travel is 55 percent or more, goods-to-person has a real target to attack. If it is 25 percent, the machine has almost nothing to remove and your money belongs somewhere else. We make the same argument in the picking model article, and it is worth repeating because it is the test that stops a bad project.
The bucket nobody logs
Pay attention to the third bucket. Time spent looking for stock that is not where the record says it is does not disappear when a machine starts fetching totes. It gets worse, because automated retrieval goes to the location the record names and brings back whatever is physically there, quickly and without judgement. If that bucket is large, your first project is inventory accuracy, not picking automation.
02 / The uplift
What the rate actually goes to
Published ranges cluster tightly enough to plan with, and wide enough that you should not take the top of any of them. A goods-to-person station is commonly described as processing 300 to 600 order lines per hour against the 80 to 120 lines per hour a skilled manual picker achieves walking the aisles, with the retrieval-to-pick cycle taking 30 to 90 seconds depending on warehouse size and fleet density. Another source puts automated picking at 300 to more than 1,000 lines per hour per operator against 80 to 150 in manual environments, routinely two to five times manual rates. Case picking sits differently again: one integrator quotes 150 to 250 cases per hour person-to-goods against 350 to 650 cases per hour goods-to-person.
| Model | Published rate | What sets the ceiling | What flexes it |
|---|---|---|---|
| Manual, trolley and list | 60 to 120 lines per hour per picker | Walking distance and slotting quality | Add or remove people freely, same day |
| Manual, RF or voice directed | Commonly 200 to 300 units per hour | Still travel, but less searching and fewer errors | Cheap to add, no layout commitment |
| Vertical lift module | 200 to 400 picks per hour for stored items | Tray delivery time, one machine at a time | An island of density, not a floor strategy |
| Goods-to-person station | 300 to 600 lines per hour, higher in optimal robotic systems | Number of stations, and tote arrival rate | Fixed layout commitment; capacity added by station |
Scroll the table sideways on a phone / Published vendor and integrator ranges, not CODETRACE figures
Two cautions before you put the top of that range in a spreadsheet. First, flexibility runs opposite to rate. A goods-to-person installation is a commitment to a layout and an order profile, where a trolley is a commitment to nothing. Second, station capacity is a hard ceiling: as one integrator notes, goods-to-person capacity is limited by the number of picking stations in the design. Your throughput is stations multiplied by rate, and stations are capital.
03 / The model
Work in labour hours, not headcount
Headcount arithmetic produces arguments. Hours arithmetic produces a number. Convert everything into hours, price the hours, and the business case stops being a debate about whether anybody will lose a job.
Work it once with round numbers to see the shape. Take 4,000 order lines a day at a measured 90 lines per hour: that is about 44 picker hours daily, or roughly 5.5 pickers on an eight-hour shift. Specify stations conservatively at 350 lines per hour and the same volume needs about 11.4 station hours, so two stations cover the day with room to spare. At an illustrative fully loaded RM 20 per hour, the 33 hours removed each day are worth roughly RM 660 a day, or about RM 172,000 a year across 260 working days.
That figure alone will not carry a large capital project in Malaysia, and this is where most imported ROI models mislead. It is also why the next two rows matter more here than they do in the markets those models were written for.
04 / The bigger lines
In Malaysia, space and accuracy carry the case
Be direct about the labour comparison. International ROI models for picking automation are built on labour costs several times Malaysian levels, and one guide illustrates the effect with an example where a 20 percent reduction in walk distance across 20 pickers at 30 US dollars an hour fully loaded saves roughly 50,000 US dollars a year in labour alone. Reprice that at Malaysian rates and the same operational improvement produces a fraction of the saving. Anyone selling you picking automation on North American or European labour arithmetic is selling you the wrong model.
Density is usually the largest line instead, and the mechanism is structural. Because nobody walks into the storage block, the storage block does not need walkable aisles. Our tote handling runs in aisles down to 850mm against roughly 3.5 metres for a forklift, and small-parts inventory stored at forklift aisle widths is the most wasteful floor in most Malaysian warehouses. Price the released square feet at your real lease rate, or as a deferred building if you own the site, and the number frequently exceeds the labour line by a wide margin. The full arithmetic sits in the storage capacity article.
Accuracy is the third line and the most consistently undersold. A stationary operator working a guided station is presented with one tote, one item and one quantity, on a screen, in a fixed position, under consistent light. That removes most of an error class that a person reading a list in an aisle cannot avoid. What it is worth depends entirely on what an error costs you, which for a supplier to an electronics or automotive customer is rarely just a re-pick. It is a return, a credit note, a corrective action report, and a scorecard entry that gets read at the next contract review.
There is a fourth line that no model captures well and every operations manager recognises: throughput you cannot hire. If your shift is already full, your peak is already staffed to its limit, and volume is still growing, then rate is not a cost saving. It is capacity you cannot otherwise buy. In that situation the payback calculation is the wrong question and the right one is whether you can take the order at all.
05 / Do this first
Fix slotting before you buy anything
The cheapest way to remove walking is to stop storing fast movers far away. Operational data attributes 30 to 40 percent of pick walk time to poorly slotted SKUs, with good slotting cutting that waste by 40 to 60 percent, and slotting combined with route optimisation producing total walk-time reductions of 35 to 60 percent in many environments. That is capital-free improvement in the same variable a goods-to-person system attacks with capital.
Two things follow, and both are good news. If slotting alone brings your travel share down from 60 percent to 35 percent, you may discover the automation case was never there, and you have saved yourself a large purchase. If the case survives even after slotting is fixed, the system you buy is smaller, because it is sized against a floor that is already efficient rather than one carrying a decade of accumulated placement decisions. Either outcome is worth the two weeks of analysis.
Three more no-capital levers belong in the same phase. Batching, so one trip serves several orders rather than one. Zoning, so no picker crosses the whole building. And measuring your replenishment discipline, because a picker standing at an empty forward location is producing a travel statistic that has nothing to do with layout. Exhaust these first and the business case you eventually write will be believed, because it will be built on a floor that is already running properly.
06 / The risk
A station starved of totes earns trolley productivity
Here is the failure mode that quietly destroys goods-to-person business cases, and it is never in the quotation. Retrieval gets the attention. Handover decides the outcome. A station has a pick cadence, and the system has to feed it at that cadence or faster. If totes arrive slower than the operator picks, your expensive station runs at trolley productivity while costing station money, and the pick rate you modelled never appears.
The mechanism is queueing rather than speed. In our configuration the storage system retrieves the tote and a floorbot carries out the handover to the station, so the tote arrives at the operator without anybody walking to collect it. The design question is not how fast the storage system can retrieve. It is whether arrival rate exceeds pick rate at the busiest moment of the day, with the handover included. Vendor material on throughput makes the same point from the fleet side, noting that more robots in circulation reduce the time any single workstation waits for inventory to arrive.
Which is why peak matters more than average, again. A floor that averages 300 lines an hour but takes 700 in the two hours before a cut-off must be designed for 700, or given somewhere to buffer. Sizing station count and handover capacity against the daily average is the most common way a correct pick-rate assumption still produces a disappointing installation. The same logic governs the movement layer, where fleet size follows peak-hour moves rather than daily volume.
The hybrid answer is usually the right one
Most floors should not automate their whole SKU list. Head of the curve into goods-to-person, long tail left manual, bulky and irregular items on pallets, one control layer scheduling all of it. That split protects the business case in both directions: it puts capital where the line count is, and it avoids paying station prices to pick something that moves twice a month.
07 / On your floor
We measure the walk before we quote the station
We do not sell hardware on day one. A picking model is a decade-long commitment to a layout, so the sequence starts with a floor study: travel share measured on a real shift, order line profile, SKU movement history, peak hours against average, and the space the storage block would occupy. Then we design the layout and simulate throughput, so station count and tote handover are proven before install. Then deployment: install, commission, integrate with your WMS and train your team.
Simulation is where the cheap discoveries happen. That the third station adds nothing because the lift is the ceiling. That fixing slotting removes a third of the walk before any capital is committed. That the evening wave needs a buffer rather than another machine. Every one of those findings is worth more to the business case than a discount on the equipment, and all of them are cheaper to find in a model than on a floor.
CODETRACE integrates on site from Shah Alam in Selangor and Batu Kawan in Penang, so the team that models your picking is the team that commissions the stations. If you are still deciding between picking models rather than sizing one, start with goods-to-person vs person-to-goods, and if the underlying question is whether your constraint is the building or the walking, the test is in what makes a warehouse smart.
Measure the walk first. The size of the prize is the size of the waste.
FAQ / Goods-to-person ROI
Questions, answered.
01What is the ROI of goods-to-person picking?
It comes from three lines, and the first one dominates in most published models. Labour: published ranges put manual person-to-goods picking at roughly 60 to 120 lines per hour per picker, against 300 to 600 lines per hour at a goods-to-person station, which is a three to five times uplift on the same operator. Space: the storage block behind the station needs no walkable aisles, so the same inventory occupies far less floor. Accuracy: a guided station presents one item and one quantity, which removes most of the read-the-list error class. In a Malaysian context space and accuracy usually carry more of the case than headcount does, because labour here is far cheaper than in the markets most published ROI models are built on.
02How much of a picking shift is actually travel?
Published accounts put it high: travel commonly consumes 50 to 70 percent of a picker's shift, and some operators report pickers spending around 60 percent of the day travelling and only 40 percent picking. But your number is the only one that matters, because travel share is the size of the prize. Follow two pickers for a full shift and log time in three buckets: travelling, picking, and everything else. If travel is 55 percent or more, goods-to-person has a real target to attack. If it is 25 percent, the machine has almost nothing to remove and the money belongs somewhere else.
03How much does the pick rate actually improve?
Published ranges put skilled manual pickers at 60 to 120 lines per hour, and a single goods-to-person station at 300 to 600 order lines per hour, with high-density robotic systems exceeding 1,000 units per hour under optimal conditions. Read those as ranges rather than promises. The rate you achieve depends on your order profile, how many lines an average order has, and whether totes arrive at the station faster than the operator can pick. A station starved of totes runs at trolley productivity while costing station money.
04How do I calculate the payback?
Work in labour hours rather than headcount. Take your daily order lines and divide by your current lines per hour per picker to get today's picker hours. Divide the same daily lines by the station rate you have specified to get station hours. The difference is the hours the system removes, which you price at your fully loaded cost per hour. Then add the floor space released, valued at your real lease or build cost, and the mispicks avoided at your real cost per error. Divide the capital cost by the annual total for a simple payback in years, and sanity-check it against a system life measured in decades.
05Should I fix slotting before buying a system?
Almost always. Slotting is where a large share of walk time comes from, and it costs analysis rather than capital. Published operational data attributes 30 to 40 percent of pick walk time to poorly slotted SKUs, with good slotting cutting that waste substantially and combining with route optimisation for total walk-time reductions in the tens of percent. Two consequences follow. First, you may find the problem was never worth automating. Second, if it still is, the system you buy after fixing slotting is smaller and cheaper than the one you would have bought before.
06What is the biggest risk to the business case?
The tote handover. Retrieval gets the attention but handover decides the outcome, because a station has a pick cadence and the system has to feed it at that cadence or faster. If totes arrive slower than the operator picks, your expensive station idles. Peak matters more than average here: a floor averaging 300 lines an hour but taking 700 in the two hours before a cut-off has to be designed for 700. The second risk is inventory accuracy, because automated retrieval fetches whatever is at the location your record names.
07Does goods-to-person suit a Malaysian warehouse with low labour cost?
Sometimes, and the honest test is what is scarce. If labour is cheap and available and your floor is not full, the payback on labour alone may be too long to justify. The cases that work here are usually driven by something else: floor space you would otherwise rent or build, accuracy demanded by a customer, throughput you cannot hire your way to because the shift is already full, or a peak you cannot staff. Density is often the strongest line, because the storage block behind the stations needs no walkable aisles at all.
08How does CODETRACE decide whether it fits?
We measure before we quote. A floor study covers travel share measured on a real shift, order line profile, SKU movement, peak hours and the space the storage block would occupy. Then we design the layout and simulate throughput so station count and tote handover are proven before install. In our configuration the storage system retrieves the tote and a floorbot carries out the handover to the station, and tote handling runs in aisles down to 850mm. Systems are deployed and supported from Shah Alam in Selangor and Batu Kawan in Penang.
Sources / Every figure in this article
Where the numbers come from
Sources are listed by what they are rather than by brand name. Every pick-rate range comes from a company that sells picking systems, so treat each one as a well-informed upper bound rather than a forecast for your floor. The ringgit figures in the worked example are illustrative planning assumptions, clearly labelled as such, and should be replaced with your own.