Key takeaways
  • A layered process audit verifies the process, not the product. It asks whether the method is still being followed, frequently and by several levels of management.
  • Frequency and layering are the mechanism: operators or team leaders daily, supervisors weekly, senior management monthly.
  • Questions must be answerable yes or no by observation in under a minute, or the audit will be completed from memory at a desk.
What LPAs catch that product inspection does not
The shortcut that has not caused a defect yet

Product inspection finds parts that are already wrong. An LPA finds the changed method that will eventually make them wrong, while the change is still free to correct.

How the layering works

  • Layer 1. Team leader or operator, daily, a handful of questions on their own area.
  • Layer 2. Supervisor or engineer, weekly, a broader set across several areas.
  • Layer 3. Plant or senior management, monthly, a small sample.
  • The higher layers matter less for coverage than for signal: they demonstrate that the method is taken seriously, and they surface things people stop noticing.

Writing questions that work

The single most common failure is questions that cannot be answered by looking. "Is the operator trained?" requires checking a record elsewhere and will be answered from assumption. "Is the current work instruction at the station, and is it the current revision?" can be answered by walking over and looking.

  • Answerable by observation, at the station, in under a minute.
  • Binary. Yes or no, with a comment field for the no.
  • Specific to a known risk, ideally one that has caused a problem before.
  • Rotated periodically, so the same five questions do not become furniture.
  • Between five and ten per audit at layer one. More and it stops happening.

The tick-box trap

LPAs decay into a signature exercise faster than most audit types, because they are frequent and mostly return yes. Three things resist that: require a photo or a comment on any no, rotate a portion of the questions monthly, and make the response to a no visible and fast.

If a no produces nothing for three weeks running, the honest response from the floor is to stop recording nos. That is a system failure, not an integrity failure.

What to do with the findings

Each no should generate an action with an owner and a date, and the aggregate should be reviewed for pattern rather than volume. The same question failing across three areas points at the work instruction or the training, not at three individuals.

Why frequency beats depth here

A thorough quarterly audit finds problems that have been present for months. Five questions asked daily find drift within days, which is when correcting it is cheap. That trade is the entire logic of the method, and it only works if the daily audit is genuinely short and genuinely gets done.

Keeping the cadence

Daily audits across several layers is exactly the kind of schedule that lapses quietly. It needs to arrive with the person, be completable at the station in a minute, and escalate when missed. Sending each layer's audit to its owner as a scheduled task, completed on the floor with photos on any no, is what keeps a layered programme running past its first busy month.