I didn’t come into e-commerce through ads. I came in through the factory, the slow way. That’s why the product is good, and it’s also why the warehouse is full.
A lot of how I think about checking work comes from the sewing floor, not from software. Here are three things from that floor.
1. The zipper you only get one shot at
A lot of our baby one-pieces close with a double-ended zipper down the front. Before that zipper goes in, the two cut panels have to be laid flat and smooth, with the print lined up across the seam. Then the zipper is set in one pass.
If it isn’t aligned the first time, you can’t nudge it. You unpick it and do it again. That’s rework, and rework costs money that never shows up on the price tag.
So the real work happens before the needle moves. You lay it flat, line it up, check it, and only then sew.
2. The veteran who used binary search on snaps
Some of our pieces have a placket with a row of six or seven snaps. Two things have to be true at once. The print has to line up on both sides along the whole length, and the spacing between snaps has to be even.
The obvious way is to start at the top and work down. Small errors add up, and by the bottom the gap is off.
One of our veteran workers figured out a better way. Set the two ends first, then the one in the middle. Split each half again and set the middle of each. Keep halving until the row is done.
Nobody on the floor called it binary search. That’s what it is, though. Every snap is placed relative to two points that are already right, so errors can’t pile up. It’s faster too, and the rows come out more even. Better, faster and cheaper all at once, and nobody taught it. It was worked out at the machine.
3. The batch we got wrong
Once, a print that was meant to be asymmetric had its left and right swapped during printing. Nobody caught it. By the time anyone looked properly, the batch was already ruined.
Other times it went the opposite way. Left and right panels that were supposed to match didn’t get lined up, because someone assumed they were fine.
Neither one was a skill problem. They were checking problems. Somebody assumed the step before them had been done right.
Why makers overbuild
This is the habit that makes the product good. It also explains the full warehouse.
When you come from the factory, you think in depth: every style, every size, enough of each that no customer hears “sold out.” Making something is the part you understand. Selling it is the part you hope will work out.
So here’s where the store is:
We’re carrying too much inventory.
Cash is tight.
It is not turned around.
Where the factory meets the software
When I started building Cortex, the system that now helps run the store, I took the factory rule with me. A model can propose work and it can veto work. It can’t approve its own completion. Something plain, like a script or a count or a second check, has to confirm the work before it ships.
That’s the zipper rule: line it up and check it before it goes out. It’s also the snap rule: anchor every step to things you already know are right. And the swapped batch is the reminder of what happens when everyone assumes someone else checked.
What’s next
Clear Halloween stock before Saturday.
Buy shallower next season: fewer sizes, deeper only where it sells.
Keep posting verified numbers, including the bad ones.
If you want to help
Parents: our clearance is real stock in a US warehouse, in sizes that actually exist. Halloween & clearance →
Builders: tell me where your system approves its own work. That’s usually where the bad batch is hiding.
Everyone else: forward this to someone who still makes things by hand.
— Uncle Loong
Fact disclosure: the three factory details above are first-hand from our own production. No inventory, loss, or cash figures are published in this issue. No affiliate links, ever. A new issue every few days, with sources.