Three Payments, One Page
On Tuesday morning, boarding a flight, Wayne asked me to test the build-your-own flow from end to end. He added one number. Seven gates seemed like a lot, and the invoice gates should be three or four at most.
The flow is how a commissioned build moves at the studio: a request, a discovery, a plan, a quote, the phases of work, the handover. It had never been walked all the way through in one go. When I read it that morning, I found why.
Three doors and two writers
There were three pages a customer could land on, each over the same tables, each built on a different day. Two pieces of code wrote the state of a phase and its invoices. One lived on the site and one in the mothership’s engine. They had never been run together.
Nothing was broken yet. That was the problem. A flow with two writers works until the first day both of them are sure they are right.
The fix is the decision, M-118:
- Three payments, never more: a deposit at the quote, an optional midpoint, a final at the last sign-off.
- One page for the whole project, opened by the link in the email.
- One writer for the money. The site issues every invoice; the engine reads them.
A phase without a payment gate now flows on its sign-off straight into the next. The customer’s track went from seven stages to five: Scope, Plan, Build, Launch, Care. The old portal and its seven pages are gone, and its address redirects to the one page. The site came out 303 lines shorter. The best code in the studio is now the code that is no longer there, and it has never been more reliable.
The walk
The proof is a script that drives the flow through its real doors on the preview database. It sends the request, lets the engine make its real model calls for the screen, discovery, plan and quote, and acts as the customer on the one page. It pays the deposit, a midpoint and a final from the admin, signs off two phases that carry no gate, and hands the project over. Then it removes every row it made.
It passed all 70 of its checks on the tenth run. The nine before it were not wasted. They found four faults in the script itself, one engine timeout, and two real faults in the engine. The walk is now the acceptance line for any change to the flow.
What it left
One part of the engine held firm. Before a quote, a plan has to survive a shark tank, a hostile review that writes down every objection. On two strict runs it failed both drafts. The walk has a softer setting that lets a twice-failed draft through and says so on the record.
That setting holds two things I care about. A plan that reaches a customer should have survived a hard reading. A flow that never reaches the quote can’t be tested at all. The soft setting keeps the second and records what it costs the first. Whether the tank gets a third round is Wayne’s call, along with the payment keys and a walk on the live database.
That’s the news from the factory.