Most line specifications describe the machine someone imagined. Four numbers — takt, availability, changeover and yield — decide what actually gets built.
A line specification usually arrives full of description and short on arithmetic: the stations, the robots, the safety concept, the preferred controller. All of it matters, and none of it determines whether the line will make the parts you need. Four numbers do that, and they are worth arguing about before anyone draws a layout, because every one of them multiplies into the capital cost.
Start with the takt time, and be careful which one you mean. The annual output you need, divided by the hours you will actually run, gives a required takt. The line has to be designed faster than that, because it will not run at a hundred percent. Specifying the takt you need instead of the takt you must design to is the most common and most expensive mistake in the document, and it shows up only at the ramp.
The gap between those two numbers is availability, and it has to be a stated assumption rather than a hope. Ninety percent and ninety-five percent sound close and are not: the difference is roughly a station’s worth of capacity across a shift. State the availability the line is designed to, state what it is measured against, and state what happens contractually if it is not met on site. A line bought without an agreed availability figure has no definition of done.
Changeover is where the design is actually decided. A line running one variant is a different machine from a line running six, and the question is not whether it can change over but how often and how fast. If a variant change costs forty minutes and happens twice a shift, you have quietly spent a fifth of your capacity on tooling swaps. Say how many variants, how often they alternate, and whether the sequence is planned or reactive — the answer decides fixtures, feeders and how much of the line has to be servo-driven.
Yield, and specifically where scrap is detected, decides where buffers and test stations belong. Finding a bad part at the end of the line means every station upstream has already spent its cycle on it. Finding it early means an in-process test that costs cycle time on every good part too. That trade-off is a business decision, not an engineering one, and it belongs in the specification rather than in a change request after the layout is frozen.
Write the four numbers on the first page with their assumptions attached, and the rest of the specification becomes checkable. Everything downstream — buffer sizes, robot reach, how many stations, whether a bottleneck needs doubling — follows from them. When a line disappoints, it is almost never because the engineering was bad. It is because it was built to a number nobody agreed on.
— GANI Engineering engineering team