Exotec Lifts First Supply Picking Speed 10x for 5 PM Next-Morning Orders

The Analysis

Five o'clock in the afternoon is a brutal cut-off for any warehouse. I have stood on plenty of docks where the pick wave for the next morning was already closed by three, and whatever came in after that sat overnight and went out a day later. So when Exotec says the Skypod system at First Supply now backs next-morning delivery on orders placed right up to 5:00 PM, what I hear is not a robotics story. I hear a commercial promise somebody has to keep every single night, using stock whose arrival time they do not control.

Start with what has actually been said. The system has been running at First Supply for more than a year. The automated store carries more than 14,000 SKUs, meaning stock keeping units, and it picks ten times faster than the manual process it replaced. The same racking footprint can support four times the storage while still holding back 30 percent of capacity for future hardware. Overnight replenishment can account for up to 20 percent of the distribution centre's daily volume. That is five numbers, and they are the whole story. Read them together and what they describe is a building that does a fifth of its work while nobody is watching it, on stock that somebody else had to get there first. The robots get the photographs. The replenishment number is the one that should worry an importer, because it is the one that puts a hard clock on your inbound schedule.

Ten times faster is the figure everybody repeats, and it deserves a careful reading. A goods-to-person system brings the bin to the person instead of sending the person to the shelf, so the comparison is system throughput against system throughput, not one picker suddenly working at ten times human speed. If the old manual building managed something like 60 order lines an hour across the pick floor, ten times that is 600 lines an hour, which is roughly the shape of a real mid-size automated site. What that buys you is not mainly headcount. It buys you hours. Where you used to start the afternoon wave at one in order to hit a five o'clock dispatch, you can now start it at four and still make the same truck. That gap between four and five in the afternoon is exactly where a distributor's late orders live. Crews ring in from a job site at half past four, and that is the order you used to lose.

A cut-off time is not an operational detail, it is the product. If you can promise next morning on an order placed at 4:55 PM while your competitor shuts at two, you have bought three hours of demand nobody else can serve the same night. Let me price what that costs to honour, because it is not free. Assume your inside sales desk takes 260 orders a day averaging 420 dollars. Assume 30 percent of them arrive between 2:00 PM and 5:00 PM, which is normal for trade counters and field crews ordering after a job. That is 78 orders a day worth about 32,760 dollars. Assume those late orders could not make the normal next-morning route before automation and had to go by courier or a dedicated run at 64 dollars a shipment against 18 dollars on the regular route, so a 46 dollar premium each.

Seventy-eight orders a day is 3,588 dollars a day, and across 250 working days that is roughly 897,000 dollars a year of expedite premium. If ten times the picking speed lets those same 78 orders ride on the normal route, that is the size of the prize. It is a much better number to put in front of whoever signs the cheque than the robot count.

Fourteen thousand SKUs sounds like a lot until you ask which fourteen thousand. In a plumbing and heating range, a large part of the catalogue is long, heavy or awkward: pipe in lengths, water heaters, ducting, radiators. None of that goes into a bin. So the automated store holds the bin-able subset, the small and medium parts that generate most of the order lines, while everything else still moves through conventional racking and a forklift. You end up running two processes inside one building, and an order carrying both a bin item and a length of pipe has to be consolidated somewhere.

Ask where that merge happens and how long it takes, because that merge point is where the five o'clock promise quietly dies. Then ask about data. A goods-to-person system only works if it knows the exact dimensions and weight of every item, and I have watched projects lose months to bad master data. For an importer this is painfully practical: new SKUs turn up from new suppliers every season, and if nobody measured them at the purchase order stage, they cannot be slotted. Fix that in the buying process, not in the warehouse.

Now the number that matters most if your stock crosses an ocean. Overnight replenishment can be up to 20 percent of the distribution centre's daily volume. In plain language, one fifth of everything the building does in a day happens at night, and it is not picking, it is putting stock away. That step decides whether those 14,000 SKUs are actually sitting in the bins when the morning wave starts. Assume the building handles 5,000 order lines a day, so overnight replenishment is up to 1,000 lines. Assume 1.4 cartons per line, so about 1,400 cartons, and assume three infeed stations running 120 cartons an hour between them, which is 360 an hour.

That is close to four hours of putaway, meaning the cartons have to be unloaded, checked and staged before roughly eight in the evening if you want the bins full by two in the morning. Work backwards from there and the conclusion is uncomfortable. An import container still sitting on the yard at four in the afternoon cannot feed tonight's replenishment at all. A trailer that was booked for ten in the morning and turned up at three has just taken a bite out of tomorrow morning's promise.

That is what turns customs release into a warehouse problem. Imported goods do not arrive on your schedule, they arrive on the port's schedule and then on the customs authority's schedule. Assume a container gets pulled for an exam and sits for three days. Assume the SKU inside it is a fast mover that accounts for 4 percent of your daily lines. By day two you are picking around gaps, and by day three the five o'clock promise is being broken on orders that had nothing whatever to do with the exam. I have sat in that meeting, and it is neither the warehouse manager's fault nor the broker's fault.

It is a planning failure nobody owned. The fix is unglamorous and it works: measure your customs release time distribution instead of your average. Ask your broker for the last 90 days of release times by entry type and look at the 95th percentile. If your 95th percentile release is four days and your replenishment window is twelve hours, your buffer has to be sized on the four days and not on the average. That is the single most useful number an importer can take out of this story, and very few people ask for it.

Returns are the other thing automation rarely solves and importers routinely underestimate. A goods-to-person store is built to take product out of bins quickly, and it is usually much worse at taking it back. Assume you run an e-commerce channel alongside the trade counter and 6 percent of shipped units come back. Assume those returns need inspecting, repacking and putting away, and assume the putaway path is the same infeed the overnight replenishment uses. Then a Monday morning surge of returns competes with the night's replenishment for the same stations, and something slips. Ask where returns go, who inspects them, and what the putaway turnaround target is. If the honest answer is that returns queue up until the weekend, then your 14,000 SKUs are not really 14,000 available, because a slice of them is sitting in a cage waiting to be counted. That gap shows up as phantom stock in the system and as customers who ordered something the website insisted was in stock.

Four times the storage on the same racking footprint is the other number with real money attached. Space is usually the reason you cannot hold the stock you know you ought to hold. Assume you import 400 long-tail SKUs from Asia on a 45 day door-to-door lead time, and assume you currently hold 30 days of cover on each because that is all the space allows. Assume 90 more days of cover works out at 30 units per SKU and the average landed unit cost is 18 dollars. That is 400 times 30, twelve thousand extra units, around 216,000 dollars of inventory.

At a 22 percent annual carrying cost, holding it costs roughly 47,500 dollars a year. Set that against what stockouts cost you now. Assume two emergency air shipments a month on those lines at 9,800 dollars each, which is 235,200 dollars a year. Holding the buffer comes out somewhere near 187,000 dollars a year better, and it only works because the racking can take four times the stock. Without that capacity the same decision is not available, and you go back to paying for air freight and hoping.

The 30 percent of capacity held back for future hardware deserves the hardest questions, because it gets sold as flexibility and often turns out to be something narrower. Storage capacity and hardware capacity are not the same as throughput capacity at the dock. If you front-load imports ahead of a tariff change, or ahead of a port labour negotiation with a date on it, inbound can double for six weeks. Assume normal inbound is four containers a day and you pull forward an extra six weeks of cover, so you are receiving something like nine a day for a month.

The bins can probably absorb it if the racking genuinely has four times the headroom. The question is whether the dock doors, the unloading labour and the infeed stations can absorb it, because those are concrete and people rather than software. Ask the vendor and the site operator separately what binds first at twice the volume, and get it written down before you commit to a pull-forward plan. I have seen a beautifully automated building with a queue of trailers outside it because nobody counted the doors.

There is a quieter consequence of four times the storage that I think gets missed. When space stops being the constraint, the assortment conversation changes shape. Assume you have been refusing to bring in a slow imported line because it would occupy racking you need for fast movers. Assume that line carries a 38 percent gross margin against 22 percent on the fast movers, and turns four times a year at 2,400 dollars of landed cost per turn. That is a line worth having, and the only reason it is not in the range today is cubic metres.

Multiply that by fifty lines and you are looking at a margin mix change that has nothing to do with freight rates. The discipline that goes with it is refusing to add SKUs just because you now can, because every new imported line still costs a purchase order, a customs entry, a carton specification and a slot in the master data. I would impose a hard rule of one line in for one line discontinued, and make somebody justify both.

Everything above assumes the thing runs, so let me talk about the day it does not. A single automated store is a single point of failure, and the question to ask on day one is what happens when it stops. Ask for the manual fallback in writing: how many order lines an hour can the building pick with people walking the aisles, how long the switchover takes, and which stock is reachable at all. Then run the drill. Pick a week in January, take the system down for two hours on a Tuesday afternoon, and count what actually goes out of the door. If the answer is 40 percent of normal volume, that is your real service level on a bad day and your customers deserve to be told something honest.

Packaging is the related trap. Bins have size limits, and imported master cartons are frequently too big for them. Assume a supplier ships 24 units in a carton that will not fit: somebody now opens every carton and repacks, which is a labour cost that never appeared in the business case. The fix sits upstream, at the purchase order. Tell the supplier the maximum carton dimensions and make them change the pack. That is an import decision, and it is far cheaper than repacking in receiving forever.

Night work is the part that surprises people. If a fifth of the volume moves overnight, you need people at two in the morning, and those people cost more and are harder to find. Assume a 20 percent shift premium and 15 percent annual turnover in warehouse labour, and build the night roster before you promise anything with a time attached to it. Be clear with yourself about what ten times faster does to the payroll as well. It rarely removes the headcount people hope for. It changes it: fewer people walking, more people standing at stations, more people maintaining the kit, plus a software and service contract that did not exist before. Ask for the total cost of operation over five years, not the purchase price, and ask what the maintenance response time is in writing. I have learned to be suspicious of any automation proposal that shows a payback period without showing the service contract next to it.

Ask for the operating numbers after the honeymoon rather than during it. Every automation vendor has a reference site with beautiful numbers in the first quarter, and the interesting question is what that site looked like in month eighteen. Ask for orders per hour by month across twenty-four months, ask for the system availability percentage, and ask what the top three failure causes were and how long each one took to fix. Assume you want availability at 98 percent or better inside your peak window, and make it a contract term. Ask how many robots were added over that period and why, because if the site had to add hardware to hit its published numbers, your business case should carry that too. I have never regretted asking for a long time series, and I have more than once regretted not asking.

The last piece is the promise itself, and it needs to be a document rather than a slogan. A five o'clock cut-off with next-morning delivery is a service level, and a service level has to say what happens when it is missed. Write down the on-time rate you are committing to, the remedy, and the exclusions. Check the carrier before you sell it. If your outbound leaves the dock at nine at night, your linehaul and final-mile partners need a late dispatch and a morning delivery capability in the lanes your customers sit in. Assume 95 percent next-morning on-time is the number you want to commit. If the carrier can only show 88 percent on the relevant lanes, you are selling something you cannot deliver, and the cost of that lands on you rather than on them. Get the carrier performance data by lane and by month before the promise goes out.

Put all of it together and the story is less about robots than about time. Ten times the picking speed buys hours. Four times the storage buys buffer you could not afford before. Thirty percent of headroom buys the ability to absorb a pull-forward. Twenty percent overnight replenishment is the price you pay in scheduling discipline, because it means your inbound has to be finished before the evening, every night, including the nights a container is late or an entry is held. I have watched plenty of companies buy the hardware and forget the calendar. The ones that got the benefit moved their appointment times, fixed their master data and rewrote their carrier agreements in the same year they installed the machines.

If you are importing into a building like this, your job starts long before the dock: dimensions on every new SKU at purchase order stage, morning appointments instead of afternoon ones, a customs release number you actually track at the 95th percentile, and a buffer sized on the bad tail instead of the average. Do those four and the five o'clock promise is real. Skip them and you have bought a very expensive way to be late. That is how I read it, and Leo will be watching whether the next-morning number survives a January peak.

↑ Back to the key points

— By Leo