How to Set Up a Layered Process Audit Program That People Actually Complete
Most LPA programs collapse within six months — not from bad intent, but from structural problems that are easy to fix. Here's how to build one that holds.
Your quality team spent three weeks building the layered process audit program. Checklists were written, schedules were set, and leadership signed off. Six months later, Layer 1 supervisors are completing audits at about 60% — and Layer 3 (plant manager and above) is at 12%.
This is not an unusual story. It’s the default outcome when an LPA program is set up without addressing the three structural problems that kill most of them.
This article walks through how to build a program that doesn’t collapse — from layer design through question writing through finding escalation. If you’re starting from scratch, you’ll have a working framework. If you’re trying to rescue an existing program, you’ll know what to fix first.
What a layered process audit actually is
A layered process audit (LPA) is a short, frequent check — typically five to ten minutes — that verifies whether a process is being followed correctly. The “layered” part means that multiple levels of the organization each conduct their own version of the check, at different frequencies.
The classic structure, as outlined in the AIAG CQI-8 Layered Process Audit Guideline (2nd ed., 2014), looks like this:
- Layer 1 — Supervisors / team leads: Daily or per-shift. They’re closest to the work and check the most granular controls.
- Layer 2 — Middle management: Weekly or bi-weekly. Broader scope, focused on whether Layer 1 is functioning.
- Layer 3 — Plant management and above: Monthly or quarterly. Strategic view — are the right things being audited? Are findings getting resolved?
The point is not to catch defects in finished product. That’s what inspection is for. The point is to verify that the process controls that prevent defects are actually in place and being followed — before a defect ever happens.
LPAs originated in automotive manufacturing. General Motors’ customer-specific requirements under IATF 16949 explicitly require suppliers to use LPAs for corrective action verification. But the approach has spread well beyond automotive — it’s now common in medical device, food and beverage, and pharmaceutical manufacturing, where process consistency is both a quality and a regulatory obligation.
The three reasons LPA programs fail
1. The checklists are too long and too generic
A 30-question checklist that applies to every work area in the plant is not an LPA. It’s a compliance exercise. People rush through it, answer from memory rather than observation, and the data is useless.
Effective LPA questions have three properties:
Observable. The auditor should be able to answer the question by looking at something — a setting, a label, a physical action — not by asking someone. “Is the torque setting at 45 Nm?” is observable. “Is the operator following the SOP?” is not.
Specific. Avoid “correct,” “proper,” or “appropriate.” If the answer requires judgment, you haven’t written a question — you’ve written a topic. “Is the coolant level between the MIN and MAX marks on the reservoir?” is specific. “Is the machine in good condition?” is not.
Tied to risk. Every question on an LPA checklist should trace back to a control in your FMEA or control plan — specifically, a control that prevents a high-severity or high-occurrence failure mode. If you can’t explain why a question is on the list, it probably shouldn’t be.
A Layer 1 checklist for a single work area should have five to ten questions. A Layer 2 checklist for the same area might have three to five — focused on whether Layer 1 is happening and whether the most critical controls are in place. Layer 3 asks whether the program itself is working.
Separate checklists for each work area. The temptation to use one master checklist across the plant is understandable — it’s less work to maintain. But it produces audits that are too broad to be useful and too long to be completed honestly.
2. Upper layers quietly stop showing up
Layer 1 completion rates tend to stay reasonable because supervisors are on the floor every day. The audit is a natural part of their shift routine.
Layer 3 is a different story. Plant managers and executives have full calendars, competing priorities, and no one holding them accountable for a missed audit. Within a few months, Layer 3 is a formality — scheduled but not completed.
This matters for two reasons. First, the whole point of a layered program is that each layer verifies the one below it. If Layer 3 never shows up, there’s no check on whether Layer 2 is functioning. Second, when leadership visibly doesn’t participate, it signals to everyone else that the program isn’t serious.
The fix is accountability that flows upward, not just downward:
- Completion rates by layer should be visible to everyone, including leadership. A dashboard that shows Layer 3 at 12% is more motivating than a monthly report that gets skimmed.
- Missed audits should require an explanation — not a punitive one, but a documented reason. “Missed — plant manager travel” is fine. Silent non-completion is not.
- Layer 3 audits should be short enough that there’s no excuse not to do them. Three to five questions, ten minutes, once a month. If it’s longer than that, you’ve designed a program that leadership will avoid.
3. Findings die in spreadsheets
An LPA is only as good as what happens when something fails. If a Layer 1 auditor finds that a poka-yoke device is bypassed, and that finding goes into a spreadsheet that nobody reviews, the audit accomplished nothing. The process remains broken. The next auditor finds the same thing.
This is the “carry-forward” problem — findings that recur across multiple audits because they were never actually resolved. A well-run program surfaces carry-forward findings automatically at the start of each audit, so the auditor knows what was unresolved from the last visit.
Every failed check should produce a finding with an owner, a due date, and a status. Systemic findings — the same failure appearing at multiple stations or across multiple audits — should escalate to a corrective and preventive action (CAPA). The LPA program is one of the best early-warning systems you have for systemic process problems, but only if the data flows somewhere.
How to structure your layers
Start with your process risk profile. Which work areas have the highest-severity failure modes? Which have the most frequent escapes? Those are your Layer 1 audit targets.
For each work area, define:
- Who audits at each layer. Assign by job title or role, not by name. People change; the program shouldn’t depend on a specific individual.
- How often. Layer 1 daily is the standard starting point. Layer 2 weekly. Layer 3 monthly. You can adjust based on process risk — a high-risk area might warrant Layer 1 audits every shift.
- What questions apply. Build a question library organized by category (safety, process parameters, error-proofing, documentation, etc.). For each work area, select the questions that apply to the specific controls at that station.
One practical note on frequency: start conservative. A program that asks for daily Layer 1 audits at 15 work areas is 75 audits per week. If your supervisors have never run an LPA program before, that’s a lot. It’s better to start with five critical work areas, build the habit, and expand than to launch at full scale and watch completion rates collapse.
Writing questions that work
Here are examples of the difference between a weak question and a strong one:
| Weak | Strong |
|---|---|
| Is the machine set up correctly? | Is the feed rate set to 120 RPM ±5 as specified on the current job traveler? |
| Is PPE being worn? | Is the operator wearing cut-resistant gloves (Level A4 or higher) at the trimming station? |
| Is the area clean? | Is the area free of parts from the previous job lot? |
| Are error-proofing devices working? | Is the go/no-go gauge for the bore diameter present at the station and within its calibration date? |
Notice the pattern: strong questions specify the expected condition, the location, and the standard. The auditor can answer yes or no by looking — no interpretation required.
Tie each question to its source. If a question comes from a control plan, note which control plan and which characteristic. This makes it easy to update questions when the control plan changes, and it makes the audit defensible during a customer or regulatory audit.
Making findings count
When an auditor marks a question as a failure, the finding should be created immediately — not at the end of the audit, not in a separate system later. The finding needs an owner (the person responsible for the process, not the auditor), a severity, and a due date.
For findings that recur — the same failure at the same station across multiple audits — escalate to CAPA. The threshold is up to you, but a common rule is: if the same finding appears three times without resolution, it becomes a CAPA. This prevents the LPA from becoming a documentation exercise where the same problems are noted indefinitely without systemic action.
Carry-forward findings — those that are still open from a previous audit — should appear automatically when the next audit starts. The auditor shouldn’t have to remember what was unresolved; the system should surface it.
What to track
Three metrics tell you whether your LPA program is working:
Completion rate by layer. Completed audits divided by scheduled audits, broken out by layer and by work area. A healthy program runs above 90% for Layer 1 and above 80% for Layers 2 and 3. If Layer 3 is below 50%, you have an accountability problem, not a scheduling problem.
Pass rate by question. Which questions are failing most often? A question that fails 80% of the time is either a real process problem (investigate it) or a poorly written question (rewrite it). Both are worth knowing.
Carry-forward rate. How many findings from the previous audit are still open? A rising carry-forward count means findings aren’t getting resolved. That’s a signal to look at your escalation process, not your audit frequency.
Review these metrics monthly with the people who own the layers. The goal is not to report numbers — it’s to use the data to improve the program and the processes it monitors.
Getting started
If you’re building a program from scratch, here’s a practical sequence:
- Pick three to five high-risk work areas to start. Don’t try to cover the whole plant in month one.
- Write five to eight questions per area for Layer 1, tied to your FMEA or control plan. Keep them observable and specific.
- Define your layers and assign by role. Layer 1: area supervisors. Layer 2: production managers. Layer 3: plant manager or quality director.
- Set frequencies. Layer 1 daily, Layer 2 weekly, Layer 3 monthly — adjust based on risk.
- Define your escalation rule. When does a finding become a CAPA? Write it down before you start.
- Run for 30 days, then review. Look at completion rates, pass rates, and carry-forward counts. Adjust questions, frequencies, or layer assignments based on what you find.
The program you have in month six will look different from the one you launched. That’s fine. The goal is a program that’s actually running — not a perfect one that exists only on paper.
Opwise includes a native LPA module that handles program configuration, layer and frequency setup, question libraries, scheduled visit generation, and carry-forward finding tracking. If you’re evaluating whether a quality system can support your LPA program, request a demo to see how it works in practice.