How LookOver Handles AQL Sampling Inside the Inspection Workflow
What actually happens from the moment an inspector opens a job to the moment a lot gets accepted or rejected.
AQL — Acceptable Quality Limit — is one of those things that sounds straightforward until you try to build software around it. For a broader step-by-step explanation of AQL sampling in manufacturing, the process starts with lot size, sample size, defect categories, and acceptance limits. The concept is simple enough. You have a lot of, say, five thousand units. You can’t check every one. So you agree upfront on a sampling plan: inspect this many units, and if the number of defects found stays within these thresholds, accept the lot. If it doesn’t, reject it.
The complexity isn’t in the math. The complexity is in everything around the math. Who sets the thresholds? Who calculates the sample size? How does an inspector on a loud factory floor actually record findings without making errors? What happens when a lot is borderline — not a clean pass, not a clear reject, but somewhere in between? And how do you make sure that, months later, someone can reconstruct exactly what was inspected, how many defects were found, and what decision was made?
That’s what we had to build. This is how we did it.
The Setup That Happens Before the Inspector Touches the App
One of the early decisions we made about AQL in LookOver was that the mobile app should not ask inspectors to calculate sample sizes or decide thresholds during inspection. That’s not laziness — it’s a deliberate architectural choice.
AQL parameters — the thresholds for critical, major, and minor defects, the sample size for a given lot — are business decisions. They’re agreed between the buyer and the supplier, often stipulated in a contract, sometimes set by industry standards, sometimes negotiated job by job. Baking that logic into a mobile app creates a maintenance problem every time a client has a different agreement, and it creates a trust problem when inspectors need to know that the parameters they’re working with are authoritative.
So in LookOver, AQL parameters come from the backend. By the time an inspector opens the job on their device, those numbers are already set. They can see them. They can’t change them.
An inspector mid-job shouldn’t be making decisions about what the AQL level should be. That decision was made upstream, by the right people, at the right time. The app’s job is to enforce it.
What the Inspector Actually Sees
When an inspector opens an AQL-enabled job in LookOver, the AQL section shows three things upfront: the sample size, and the thresholds for critical, major, and minor defects. These are read-only. They’re the ground rules for this inspection.
The inspector’s job is to work through the sample — physically examining units, taking photos, making notes — and then record what they found. The AQL section has three input fields: findings for critical defects, findings for major defects, findings for minor defects. That’s it. Count what you found, enter the numbers.
“Three numbers is something anyone can do accurately — even quickly, in a noisy environment, on a phone that might have grease on the screen.”
Each defect category is tracked separately because they carry different weight. A critical defect — something that makes a product unsafe or completely non-functional — has a much lower tolerance than a minor cosmetic variance.
| Category | Tolerance | Description |
|---|---|---|
| Critical | Lowest | Unsafe or completely non-functional — near-zero tolerance. A single critical defect can trigger rejection. |
| Major | Moderate | Substantially reduces usability or does not meet agreed specs. Low but non-zero threshold. |
| Minor | Highest | Cosmetic or non-functional variance. Widest tolerance — a lot can pass with several minor defects. |
How Accept and Reject Points Are Calculated
The math is straightforward, and we expose it clearly rather than hiding it in a black box. In LookOver, the accepted defect limit for each category is configured in the backend based on the agreed sampling plan. The inspection report then shows the sample size, configured thresholds, recorded findings, and acceptance points clearly so the final decision can be reviewed without confusion.
What the app does not do is automatically determine pass or fail. That’s intentional.
The Final Call Stays With the Inspector
AQL is a statistical tool. It gives you a framework for making a decision about a lot based on a sample. But it doesn’t — and can’t — account for everything an inspector knows when they’re standing on the floor.
- The critical defect found is on a non-visible component that doesn’t affect function.
- A defect pattern observed suggests a systematic production issue — warranting a hold even if count is technically within tolerance.
- Context from earlier in the job changes how the numbers should be read.
Inspectors are professionals. They know things the app doesn’t. So LookOver presents the numbers clearly, shows the acceptance points, and then gives the inspector four options for the final status:
| Status | When to use it |
|---|---|
| Pass | Findings are within all acceptance thresholds. Lot is clear. |
| Fail | One or more categories exceed the acceptance threshold. Lot is rejected. |
| Hold | Needs more eyes before a decision is made. Prevents premature acceptance or rejection. |
| Abort | Inspection could not be completed — job is invalidated for this run. |
The reason a QA manager needs to capture isn’t always one of four options. Sometimes it’s “held pending supplier response on root cause of critical defect cluster in units 140 through 160.” That should be written down, not approximated.
Per-Item AQL Tracking
Most AQL implementations think at the lot level. LookOver tracks at the line item level as well.
A single inspection job often covers multiple SKUs or product lines. A buyer might have different AQL agreements for different categories — tighter thresholds for a safety-critical component, looser tolerances for a packaging item. Each line item in Lookover carries its own sample size and its own critical, major, and minor thresholds.
“An inspector working a multi-SKU job isn’t applying a single blunt AQL standard across everything. They’re applying the right standard to each item.”
And the report reflects that — each line item shows its own findings against its own acceptance criteria.
What an AQL Inspection Report Includes
When an inspection is submitted, the AQL data becomes a permanent part of the record. The PDF report that Lookover generates includes, for each relevant line item:
- Sample size used for this item
- AQL thresholds for critical, major, and minor defects
- Findings by defect category
- Calculated acceptance points
- Final status with the inspector’s noted rationale
When a supplier disputes a rejection six weeks later, the answer isn’t “the inspector said so.” The answer is a timestamped document showing exactly what was sampled, what was found, what the agreed thresholds were, and what decision was made and why. That audit trail is automatic. No one has to manually compile it after the fact.
Why This Approach Works on the Factory Floor
The pattern across all of Lookover’s features is the same: reduce what the inspector has to think about during the job, preserve all the context they’d need to justify the decision afterward. AQL sampling fits that pattern exactly.
No table lookups needed
The inspector doesn’t have to know which AQL table applies or calculate a sample size in their head. It’s already there when they open the job.
Thresholds always visible
The app shows them upfront — no memorization required. Ground rules are set before the first unit is touched.
Report auto-generated
No compiling findings into a document at the end of the day — it’s already done the moment the inspection is submitted.
What inspectors have to do is inspect carefully, count accurately, and make a professional judgment. The app handles everything else.
That’s the workflow. From configuration on the backend, to findings capture in the field, to a documented decision with a full audit trail. End to end, without anyone having to stitch it together manually.