BROKER+

The business ran on spreadsheets. That works right up until it doesn't — and it offers no way to enforce who is allowed to see what, which turns out to be the whole problem in broking.

A working production MVP was live after two weeks. At eight weeks real users were logging in and conducting business, and the first deal closed through the system days after that.

That absence is the entire reason this exists. Generic CRMs model contacts and deals. Cattle broking models animals moving between parties, with weight, count, and money all changing along the way.

So the Postgres schema had to carry the domain rather than approximate it: shipments, adjustments, partners, e-signatures, and the cattle-broker math tying them together.

The hardest problem wasn't the schema. It was deciding what a deal partner is allowed to see.

Partners need enough visibility to actually work a deal. They cannot have so much that they can route around the broker on the next one. That line is commercial, not technical — and it had to be encoded in role-based access control rather than left to convention and good manners.

Both failure directions cost something. Too little visibility and partners can't operate. Too much and the platform quietly hands away the broker's position in their own market.

E-signature here covers the contract terms of the deal itself. Once terms are agreed, the document gets signed in the same place the deal lives, rather than in a detached email thread.

Redis handles rate limiting. Auth, hosting, and deployment were part of the same delivery — this was a full-stack contract, not a front end handed to someone else to run.