The curve on the screen was a masterpiece. Smooth rise, clean dwell, gentle fall. Twelve trainees leaned over the monitor in the training room of a German dashboard plant and nodded. Then we cross-sectioned the joint. The barrel was half filled. The curve had not lied, exactly. It had answered a question nobody asked. That is worse.

Selective soldering makes the illusion easy to believe. The heat is local — two sensors a centimeter apart can both be right and tell different stories. In reflow, the whole board shares one oven. In selective soldering, the board shares nothing with the pin under the nozzle. So the placement of the sensor is the experiment; the software is only the notebook.

I train operators and technicians for a living, so I have watched this scene play out in more than one country. The data is real. The screen is real. The joint is real. Only the relationship between them is imaginary. Here are the seven places I forbid in class.

One: the convenient corner. Tape sticks well there, the wire reaches, nobody has to lean over the board. It is also thermally irrelevant. If the corner has different copper, different airflow, different fixture exposure, it lives on another planet from your connector.

Two: the gap. A tip lifted one millimeter from the pad now measures hot gas and radiation — a ghost signal with a lovely response time. Warnings: suspiciously fast spikes, noise, a channel that disagrees with its neighbors for no physical reason, tape peeling after the run.

Three: under a mountain of solder or a heavy clamp. You wanted strong attachment. You built a tiny heat sink and measured the clamp's feelings instead of the joint's. The added thermal mass changes the very response you wanted to record.

Four: anywhere the wire crosses the wave. A lead dragged through molten solder stirs the flow, cooks its insulation, and drags the tip off the target. Route the wire so no force can pull the sensor free, and fix it in the middle so the cable does not carry the load.

Five: the solder pot, when you wanted the joint. Pot channels must be labeled like radioactive material — clearly, separately, unmistakably. A pot reading confirms the source. It cannot testify for the pin, the barrel, or the pad. I have seen entire shift meetings wasted on this confusion.

Six: "near the connector." Near is not a coordinate. In training I make students write: reference designator, pin or pad number, board side, nozzle path relationship. If a technician cannot rebuild your setup from the sheet without asking what "near" meant, the channel is folklore, not data.

Seven: wherever the profile looks fine and nobody checked contact after the run. Pull-check the point. A moved sensor invalidates the channel even when the curve is beautiful. Especially then. Beauty is what the lie wears to work.

Nobody in that room was careless. They were simply reading a number with no address. A channel without a documented position is a rumor with a decimal point, and it will sign your release papers while the joint plans its betrayal.

In that German plant — automotive dashboards, coating line feeding selective soldering feeding AOI — we trained twelve people. The lesson that stuck was not which button starts the program. It was this: a profile is an argument about heat, and every argument needs a witness placed at the scene of the event. The four-zone method writes those witness statements properly — a field map for sensor placement is the document to follow.