A title processor doesn't buy one house.
She buys eleven, for four closings, before lunch.
That one sentence decided the entire shape of our shopping cart.
Consumer carts assume one buyer, one basket, one payment. Here, all three are wrong.
Start with a wall we didn't build.
Every management company on the platform is its own merchant of record. Their bank. Their payouts. Their name on the buyer's card statement.
Which means a single payment can never contain two companies' products. Not "shouldn't". Can't.
So the cart isn't one shop. It's a covered market — independent shops under one roof, one front door, and completely separate books.
So items group by company. Each group carries its own running subtotal, including whatever options the buyer added — rush handling, physical delivery, the little chips beside each line. Each group checks out separately.
Then the second constraint, which comes from the buyer's own accounting department.
Title companies don't reconcile by shopping trip. They reconcile by transaction. Every closing file gets its own invoice, and that invoice has to match exactly one closing.
An eleven-property receipt is worse than useless to them.So the cart lets them drag properties into named invoice groups, scoped to one company, and check each group out on its own.
We didn't design that part. Accounts payable did.Last piece: the cart persists. It survives a reload, survives a closed tab, holds for 24 hours, and clears the moment you log out.
Because assembling eleven properties takes twenty minutes and gets interrupted nine times. A cart that empties when you close a tab was designed for someone buying socks.
None of this is clever. It's just the shape you end up with when you watch how the money actually gets reconciled instead of guessing.
Next: the opinionated forty seconds between clicking Pay and the money landing in somebody else's bank account.