The Witness Mark or Data Logger Method

Why this matters

Some faults only happen when you are not there. You cannot bill the customer to sit and watch a unit for three days, and you cannot guess your way to a fix on an intermittent without proof. The answer is to make the unit, or the fault itself, leave a record you can read when you come back. A witness mark or a data logger turns "it acts up sometimes" into a timestamped event with conditions attached. This is how the best techs corner the intermittents that beat everyone else: they stop watching and let the evidence watch for them.

Step 1: decide what you need to capture

Before you leave anything behind, know what question it has to answer.

• Did it move? A connection backing out, a part shifting, a linkage drifting. A physical witness mark answers this.
• Did a value go out of range, and when? A voltage sag, a pressure spike, a temperature climb. A recording meter or logger answers this.
• Did it trip or fault, and how often? A latching indicator or a stored fault count answers this.
• What were the conditions when it failed? Temperature, time of day, load. A logger that records conditions alongside the value answers this.

Pick the cheapest method that answers the actual question. Do not log a hundred channels when a paint pen on one connection would tell you what you need.

Step 2: lay a witness mark for movement

A witness mark is a reference you place across a joint so that any movement shows. It is the simplest, lowest-cost intermittent tool there is.

1. Pick the joint you suspect: a fastener, a connection, a coupling, anything that might be working loose.
2. Mark across the gap so the line spans both parts: a paint pen or marker stripe that crosses the seam.
3. Note where you marked in the job record so you, or the next tech, knows to read it.
4. Read it on the next visit. If the line is broken or offset, the joint moved, and you have your fault. If it is intact, the joint is innocent and you look elsewhere.

The same idea works for "did this get hot": a heat-sensitive indicator placed on a suspect part shows whether it exceeded a temperature while you were gone.

Step 3: leave a data logger for an electrical or fluid value

When the fault is a momentary drop, spike, or excursion the unit swallows, leave something recording.

1. Place the logger or recording meter on the suspect parameter: the supply voltage, the operating pressure, the temperature at the point that misbehaves.
2. Capture conditions too, where you can, so an excursion has context. A spike at a known time and temperature is far more useful than a bare spike.
3. Set it to run across the window the fault appears in. If it fails daily, leave it a couple of days. If it fails under a specific use, leave it through that use.
4. Secure it safely. It must not become a hazard, get in the way of operation, or be disturbed by the customer.
5. Tell the customer what it is so it is not mistaken for the repair or unplugged.

Step 4: use the unit's own memory if it has one

Many modern controls already log for you. Before you add hardware, ask what the unit already knows.

• Stored fault codes and counts tell you what tripped and how many times, even if it is clear now.
• Runtime and cycle data can show a pattern, like a fault that appears after a certain run time.
• A latching protection that is still tripped is a witness in itself: it caught the event and held it for you.

Reading the unit's memory is free and immediate. Do it before you leave anything behind.

Step 5: read the evidence and act on it

When you return, the record either names the fault or clears a suspect. Both are progress.

• Witness mark broken: the joint moved. Remake it, secure it, find why it loosened.
• Logged excursion at a known time: correlate it to conditions and trace the cause. A voltage sag under a specific load is a supply problem; a pressure spike at a cycle point is a control or restriction problem.
• Fault count climbing: the unit is catching the event. Now reproduce the conditions it logged and diagnose normally.
• Everything clean: the suspect is exonerated. Move the witness to the next candidate. A negative result narrows the field.

Step 6: document it as a method, not a guess

Record what you left, where, and what it showed. A documented logging campaign tells the customer you are working the problem systematically, not throwing parts at it. It also protects you: a fix backed by a logged event is defensible, and a clean log that exonerates a part keeps you from replacing it on a hunch.

The mental model to keep: you do not have to be there when the fault happens, you have to leave something that was. Mark it, log it, or read its memory, then come back and let the evidence point.

References

• Trade-standard practice for intermittent-fault data logging and witness marking
• Manufacturer documentation on stored fault codes, runtime data, and latching protections
• See related: Catching the Fault That Hides When You Arrive (decision tree); Forcing an Intermittent to Show Itself
• See related: The Customer Can't Reproduce It (decision tree)