By: Ole Jakob Skjelten, CEO of Embodee
An order comes in. The customer has designed something on your site, approved it, paid for it, and is now waiting for it to arrive. In a lot of companies what happens next is that a person opens that order, reads it, and rebuilds it by hand into something the factory can actually use.
In September I wrote that workflows deserve to be the starting point rather than the afterthought, and that most of the friction in a custom-product business sits in the interfaces between the parts. I want to take one of those interfaces and look at it properly, because it is the one where I think the most money is lost.
It is worth being specific about what happens in there, because from the outside it looks like almost nothing. Someone checks that the design is complete and can actually be made. Someone rebuilds the artwork in the format that factory expects, at the right resolution, with the right color definitions, positioned correctly across the whole size range. Someone puts the order data, the design data and the production instructions into whatever package that particular factory, and quite often that particular machine, wants. Then someone else checks all of it, because the cost of getting it wrong is a production run with a name spelled wrong on it.
The variation between factories is the part that surprises people. One partner I know of insisted on receiving color as spot definitions in a format I am fairly sure predates me, for reasons to do with their own quality process that I never entirely got to the bottom of. They were not being difficult. It was a real constraint at their end. But it meant every order going to them needed its own handling, and that handling lived in one person’s head.
Multiply that by every factory you work with.
In logistics the last mile is the expensive part. Moving a container across an ocean is cheap per kilometer, getting the parcel to the right door is not. Custom products have the same shape.
The time is the part customers feel. On a custom order the time more or less is the product, because they have ordered something that does not exist yet and are waiting for it to be made. Manual translation adds hours to days per order. Browzwear published a customer case showing 30% fewer samples and 25% shorter development time, which matches what we see. Take the translation steps out and calendar time comes out with them.
Then the capacity problem, which I think is the more serious one and gets talked about least. The people doing this work are skilled: production artists, product developers, sometimes the designers themselves. Paying them to convert file formats is expensive in the obvious way. The less obvious way is that it puts a ceiling on the business. If an order costs an hour of a person, your order volume is your headcount, and it stays your headcount however good the front end gets.
Mistakes are where it gets expensive fast. Every manual touch is somewhere one can get in, and custom orders are unforgiving in a way stock products are not, because there is nothing sitting on a shelf to swap in. A wrong file means a remake and a delay, and sometimes a customer you do not hear from again.
None of this is news to the people doing it. They can usually tell you to the hour where the time goes. It survives anyway, partly because it grew rather than got decided. Nobody designed the last mile. It accumulated, one product launch and one factory relationship at a time, and each step made sense on the day it was added. Partly because it is nobody’s line item: the cost sits inside salaries, lead times and error rates, so it never becomes a number that anyone owns or has to defend at a budget meeting. The third reason is more awkward. Fixing it needs the commerce side and the production side in the same room, and in most companies that does not happen often.
Underneath that there is something more structural. Most of these workflows were never built for customized products at all. They were built for standard ones, where the product exists before anybody orders it, and where the file that tells the factory what to make was drawn up once, months earlier, by a product team. Customization arrived later and got bolted onto the shop end, because that is where the demand showed up. The ordering side learned to deal with a customer who designs their own product. The production side never got rebuilt around the fact that every order is now a slightly different product, so a person stands in the gap and does by hand what the old assumption used to do for nothing.
Lectra’s 2026 white paper describes product data as chronically fragmented across PLM, CAD, fabric systems, email and spreadsheets, and warns that adding tools piecemeal tends to amplify the silos rather than remove them. That is their framing of a broader problem and not specifically about this step. But every silo boundary on that list is a place where somebody carries information across by hand, and the last mile is where they all stack up at once.
What you want instead is for the system that captures the order to be the system that produces the factory files. Set the product up properly once, with its rules, its print areas, its color definitions and its per-factory output requirements, and every order that comes through can generate its own production-ready files, vector-based and correct for the factory that will make it.
I should be honest that “once” is doing a lot of work in that sentence. The setup is real work, and it is where these projects actually get hard. But it is work you do per product rather than per order, and that is the whole difference.
When it does land, artwork preparation that took days takes hours or minutes, and error rates fall mostly because there is nowhere left for the errors to get in. The skilled people go back to doing the thing you hired them for.
If you want to know where you stand, take one real order and follow it from the moment the customer clicked buy to the moment the factory started production. Write down every human touch in between, and what each one was for. Most teams have never done this end to end, and most are surprised by the number they land on. Do that before you buy anything else.
