Audit sampling tells you how a few cars look through a customer’s eyes and whether your process is broadly on track. Measuring 100% of cars tells you about every car, including the ones built between audits, which is where drift and one-off fitting errors hide.
Most Indian car plants need both. This guide explains how fit and finish is checked today, what sampling can miss, what per-vehicle data adds, how this relates to J.D. Power India IQS and PP100, and when sampling on its own is enough.
How fit and finish is checked today
On a typical car line, fit and finish passes through several layers of checks. Each one sees a different share of cars and looks for different things.
In-line checks by operators
Operators at closures fitting and final assembly check panels by eye and by hand, with feeler gauges or handheld laser gauges at set points. Nikon notes that gap and flush measurement has long “relied on manual tools used by highly skilled workers”, or on robot installations. These checks can cover many cars, but results depend on the person, and whether every reading is recorded varies from plant to plant.
Vehicle audit and customer product audit
A small number of finished cars are pulled from the line each shift or day and inspected in detail in an audit area. Many plants call this a customer product audit (CPA) or vehicle audit. The German automotive standard VDA 6.5 describes a product audit as a “sample check” that evaluates product quality “from the customer´s point of view”. It notes that results during series production can show quality status, trends and weak spots. The auditor judges the whole car: gaps, paint, trim, function and feel.
Pre-delivery inspection (PDI)
Before handover, the car is inspected at the dealership, and buyers are advised to do their own check. Autocar India’s delivery checklist tells buyers to “Check all body panels for scratches, dents, repaint marks, swirl marks, or uneven panel gaps.” By this point the car has left the plant, so any fault found here is found in front of the customer.
Customer surveys
After delivery, surveys such as J.D. Power’s India Initial Quality Study record problems owners report. This is covered below.
Comparing the three plant-side approaches
| Approach | Cars covered | What it judges well | Main limits |
|---|---|---|---|
| Operator in-line checks | Many or all, but not always recorded | Obvious misfits, quick correction at the station | Varies by person and shift; patchy records |
| Vehicle audit / CPA | A small sample | Whole-car, customer-eye view: gaps, paint, trim, function, feel | Sees only the sampled cars; slow to show drift |
| Automated 100% measurement | Every car through the station | Numeric gap and flush per seam, per car, with a trend | Only the seams and characteristics it is set up for; no judgement of feel or appearance beyond what it measures |
What sampling can miss
Drift between samples
A sample tells you about the cars you sampled. A loosening hinge fixture, a worn locator or a new batch of panels can start shifting gaps at any time. If that happens just after an audit car, every car built until the next audit goes out unmeasured. The NIST/SEMATECH statistics handbook says the main purpose of acceptance sampling is “to decide whether or not the lot is likely to be acceptable, not to estimate the quality of the lot”. A product audit is not the same as lot acceptance, but the point carries over: a small sample is a check, not a full picture.
Operator variation
Manual gap and flush readings depend on who takes them. A 2026 measurement system analysis of a portable laser gauge by Paolini and colleagues found that “operator-induced deviations continue to affect measurement reliability, especially in manual modes”. If two auditors on different shifts read the same seam differently, part of what looks like a trend is really the people.
Short faults
Some problems last a few cars, such as a wrongly fitted fender after a line stop. A sample is unlikely to catch these, and in-line operators may not record them.
A worked example of sampling coverage
All numbers in this section are hypothetical. They are not data from any plant.
Suppose a line builds 400 cars a shift and the audit team pulls 4 cars a shift, evenly spaced, one in every 100. That is 1% coverage.
Now suppose a fixture problem shifts the rear door gap out of tolerance on 20 cars in a row before someone notices it at the station.
- With one audit car every 100 cars, a run of 20 consecutive bad cars includes an audit car only if it happens to overlap one. The chance is 20 in 100, or 1 in 5.
- So in this hypothetical case, about 4 times in 5 the audit would see nothing, and all 20 cars would carry on to dispatch.
- Even when the run overlaps an audit car, the auditor must check that seam and read it correctly.
- If the problem starts just after an audit car, up to 99 more cars can pass before the next audit.
With 100% measurement at that seam, the first out-of-tolerance car would be flagged, as long as the measuring system itself is capable at that tolerance. The example uses round numbers to show the logic. Your own audit rate, run lengths and seams will differ.
What 100% data makes possible
A record for every vehicle
Automated systems can file each result against the vehicle ID. Konica Minolta’s EINES tunnel page describes inline gap and flush tunnels where “every car is inspected with no line stoppages”, with integrated reporting and traceability. When a dealer raises a PDI complaint, you can look up what that car measured at the plant.
Trend and drift detection
With a result for every car, you can chart each seam over a shift and see a slow drift before it reaches the tolerance limit. With a sample of a few cars a shift, the same chart has too few points to show a gradual shift quickly.
Feedback to closures fitting
The value of measuring is in acting on it. The ISRA/Perceptron AutoFit brochure describes closed-loop feedback that can “automatically transmit offsets to downstream stations”. Nikon’s APDIS blog says final assembly needs “an efficient way to measure, record and feed back this final quality control information”. In practice, that means sending seam data back to the teams that hang and adjust doors, hoods and tailgates.
Better use of auditors
If gap and flush are measured on every car, auditors can spend more of their time on what instruments do not judge well: paint appearance, trim fit, door closing feel and noise.
How this connects to IQS and PP100
J.D. Power measures initial quality as “the number of problems experienced per 100 vehicles (PP100)”, where a lower score means higher quality. The 2023 India Initial Quality Study surveyed 7,198 new-vehicle owners face-to-face in 25 cities. It counted problems from the first month of ownership up to six months. The industry average was 122 PP100, unchanged from 2022. Problems fall into eight categories, one of which is exterior. Design-related problems and defects were cited almost equally.
Be careful about the link. IQS counts problems that owners report. It does not publish gap and flush measurements, and the release does not break out panel gaps as a separate figure. Plant data cannot be converted directly into a PP100 change. What can be said is that fit faults caught at the plant do not reach PDI or the owner, and that fewer visible defects at delivery is in line with what IQS rewards.
When sampling is enough
Sampling is a sound choice in several situations:
- Your process is stable and audits rarely find gap or flush problems.
- Volumes are low, so a sample already covers a large share of cars.
- The check is slow, costly or destructive. NIST lists these as the cases where sampling is used instead of 100% inspection.
- The characteristic needs human judgement, such as door closing feel, wind noise or paint appearance.
- You are early in a model’s life and the check plan still changes often.
Even with 100% measurement, keep the audit. It is your customer-eye check on the whole car, and an independent check on the automated system.
Where DeepVision Tunnel fits
DeepVision Tunnel from Indus Vision is a drive-through 3D measurement arch for final assembly. Structured-light 3D cameras read each panel seam as the vehicle passes on the conveyor, with no stop and no fixture. DeepMeasure software computes gap, flush and parallelism along the seam and files results against the vehicle ID, so each car has its own record. It runs on premise, data stays in the plant, and results export to quality and MES systems. Seams it can cover include hood, headlamp, door to fender, door to door, rear door, fuel lid and tailgate. The first measuring station was trialled in August 2026 on a door seam with the car moving at 32 mm/s; trial values are not a guaranteed specification.
It is meant to sit alongside your CPA, not replace it. If you are choosing between measurement types, see our comparison of handheld gauges, robot cells and tunnels.
Frequently asked questions
What is PP100?
PP100 is J.D. Power’s measure of problems experienced per 100 vehicles. A lower score means fewer reported problems and higher quality. The 2023 India IQS industry average was 122 PP100.
Does 100% inspection replace the customer product audit?
No. Automated measurement covers specific seams on every car. The audit judges the whole car as a customer would, including things instruments do not measure well. Use both.
How many cars should we audit per shift?
There is no single right number. It depends on volume, process stability, audit time and your customer’s requirements. The worked example above shows how to reason about coverage for your own line.
Will 100% gap and flush measurement improve our IQS score?
No one can promise a PP100 change. IQS counts problems owners report across many categories. Catching fit faults before dispatch removes one source of visible defects, but other factors also drive the score.
Is 100% measurement worth it on a stable line?
It may not be. If audits and warranty data show few fit problems, sampling may be enough. The case is stronger where faults appear between audits or where you need a per-vehicle record.