Which bench is losing the most time, and where did a bad lot go?

A factory generates more data than any other part of a business and answers fewer questions with it. Punches, stops, verdicts, lots, work orders — all recorded, and still the production meeting runs on what the supervisor remembers from Tuesday. These are three questions that meeting should be able to settle, and the calls Rutba Assistant makes to settle them.

“Which bench is losing the most time?”

A production manager before the weekly meeting. The assistant reads every bench’s overall equipment effectiveness, worst first, then lost time by reason, then the full working behind the worst bench’s figure:

GET /mfg-oee
GET /mfg-downtime-events/summary
GET /mfg-oee/work-centre

Two things about these reports make the answer worth trusting. A factor with no data comes back empty, not zero — and the bench’s effectiveness is empty when any factor is — because an effectiveness of 0% on a dashboard is the number that gets a maintenance budget cut for the wrong reason. The average is taken over the benches that have a figure, never over the ones that do not. And planned downtime is reported separately from lost time, so a factory is not charged for being closed on a Sunday.

The assistant passes both along. If the worst bench has no quality verdicts recorded, the answer says its effectiveness cannot be computed and why, rather than ranking it.

When the question turns to money — why did this work order cost more than standard — one call splits the variance by cause: material price, material usage, labour rate, labour efficiency and overhead. “Over budget” becomes “the steel cost more, and the second shift took longer”, which are two different conversations with two different people.

GET /mfg-costing/variance

“Will next week’s plan hold?”

A planner on Friday afternoon. The assistant reads the load on each work centre, the exceptions in the plan, how well last week’s schedule was kept, and which machines are down:

GET /pl-capacity-load
GET /pl-exceptions
GET /pl-schedule/adherence
GET /mt-summary

The load report sets the minutes each work centre needs against the minutes its calendar has, day by day. The exceptions list groups what is wrong with the plan by kind, each with the action it suggests. Adherence compares last week’s schedule with what the shop floor actually punched — a schedule nobody measures is a schedule nobody keeps. And the maintenance summary says which machines are down right now.

None of these is stored overnight. Load and exceptions are worked out when they are read, so moving a work order changes the next answer — the assistant is never reading a plan from before the change somebody made at four o’clock.

“A supplier says a lot was contaminated. Where did it go?”

The question every manufacturer hopes never to ask, usually asked with a customer’s quality team on the phone. The assistant reads every lot that supplier delivered in the window and everything made from them, follows the finished batches forward, and checks what is already on hold:

GET /mfg-trace/recall
GET /mfg-trace/forward?includeCustomers=true
GET /qa-holds

The forward trace can follow work orders on to the sale orders they fed — and it labels that step as a best effort, because a sale order in the suite carries no batch reference. The assistant keeps the label. “These three customers probably received it” and “these three customers received it” are different sentences on a recall, and saying the first as if it were the second is how a recall misses somebody.

Press Report and the trace becomes a spreadsheet in Drive — lots, batches, orders, holds, and a Sources sheet a customer’s auditor can read without having seen the conversation.

Where it hands back to a person

• Putting stock on hold, releasing work, firming planned orders. Each changes what the factory does next, so each is a plan a person approves.
• Telling customers about a recall. What reaches a customer in your name is yours to send; it is on the short list of things Rutba Assistant will never be allowed to do.
• Deleting anything. A punch, a verdict, a lot — traceability is only as good as the records nobody removed.