Two problems wearing one brief.
Appetite Studio wanted a store for collectible model cars. Half the brief was familiar territory: a catalog, a cart, checkout, order history. Every store on the internet has these, which is exactly why none of them needed inventing.
The other half existed nowhere. They wanted every collector to have a personal garage: sign in and your cars sit on shelves, like the glass cabinet you keep at home, except this one fills itself as you buy. No plugin does that. No template does that. That half was the actual product.
Invention is a budget. Spend it where the client is different, not where they're like everyone else.
Shopping for wheels.
Checkout has been solved by companies that employ compliance departments. So has card handling, tax arithmetic, transactional email, image resizing. For each of these I picked the boring, documented option and wired it in. The payment flow took two days to integrate. Building our own would have taken two months and produced something worse on every axis that matters: security, uptime, refunds, disputes.
There is no glory in this part of the job. There is also no phone call at 2am when a card network changes a rule, because that call goes to someone whose whole company exists to take it.
The part with no wheel.
Everything saved above went into the garage. A purchase lands on your shelf the moment the order confirms. You can rearrange the shelves, and the cabinet shows the collection the way a collector actually thinks about it: as objects owned, not rows in an order table.
The data model behind it is small. The care went into behaviour: what happens when an order is refunded, when the same car is bought twice, when a collection outgrows a shelf. That polish took weeks, and those weeks only existed because checkout didn't eat them.
The founder later told us collectors sign in just to look at their own shelves. No discount code ever did that.
Where the line actually sits.
The test is not difficulty. Plenty of solved problems are hard; they stay solved anyway. The test is whether the business is different because of that piece. ToyCar does not win a single customer because its checkout is special. It wins because of how owning a collection feels.
One warning from the other direction: adopted parts still deserve scrutiny. I read the pricing pages, the rate limits and the exit path before wiring anything in. A wheel you cannot leave is a trap with good documentation.
What the wheel bought.
The store went live with a checkout that had already processed millions of orders for other people, and a garage that had processed none because it had never existed before. Both worked on day one. Only one of them was ours, and that was the plan.
Clients sometimes worry that integrating instead of building means getting less for their money. It means the opposite. The money stops going into problems the industry already solved and starts going into the one thing nobody else has.