A supplier reports that a disposable AIO lot is within specification. Your receiving lab tests the same model and records a different coil-resistance distribution or a noticeably different draw. The instinctive response is to question the parts. That may be justified—but it is not the first question to answer.
Before deciding whether the hardware changed, buyers need to establish whether both laboratories measured the same characteristic in the same way. A stable method can reveal product variation. An unstable or mismatched method can create variation of its own.
This distinction matters whenever test results influence sample approval, first-article decisions, lot acceptance, corrective action, or supplier performance reviews. Rule of thumb: validate the measurement method before evaluating part compliance.
Product variation and measurement variation are not the same problem
Every reported result contains more than the condition of the device. It can also contain effects from the instrument, fixture, electrical contacts, seal, operator, sample state, test sequence, calculation, and environment.
Suppose Lab A reports a mean coil resistance that is consistently higher than Lab B. That difference could reflect a real production shift. It could also come from lead and contact resistance, different insertion pressure, a different point in the assembled circuit, or inconsistent contact cleaning.
The same logic applies to airflow or draw-resistance measurements. A small leak around an adapter, a different mouthpiece seal, an unreported flow setpoint, or a different sample orientation can change the reading without any change to the internal airway.
A product qualification plan tells you what to test. Method validation asks whether the test produces results that are fit for the decision. Buyers should treat these as separate layers. ILEVA’s sample-testing protocol can organize the wider qualification sequence; the method-agreement work in this guide sits underneath the measurements used in that sequence.
Start by defining the measurand
“Measure resistance” is not a complete instruction. Neither is “check airflow.” The parties first need to define the measurand—the exact quantity intended to be measured.
For coil resistance, clarify at least:
- whether the reading is for the coil alone or the effective resistance of an assembled device path;
- where the probes or fixture contacts touch the device;
- whether a two-wire or four-wire method is used;
- whether the sample is measured before or after filling, conditioning, activation, or functional cycling;
- how unstable readings, contact failures, rounding, and retests are handled.
For airflow or draw resistance, clarify:
- the test mode: pressure drop reported at a controlled flow rate, or flow reported at a controlled pressure;
- which openings are sealed, open, or connected to the fixture;
- the adapter geometry and insertion depth;
- device orientation and whether activation is prevented or allowed;
- conditioning time, temperature, and sample state;
- the units, calculation, stabilization window, and retest rule.
If two labs cannot complete the same method-definition sheet, their numbers should not yet be compared as if they came from one measurement system.
Six variables that often hide inside a test result
1. Sample identity and condition
The comparison should use the same product code, engineering revision, component configuration, and defined lot state. Filled and unfilled devices, freshly assembled and conditioned devices, or samples exposed to different storage histories are not automatically equivalent.
Record the sample ID and relevant state before testing. If samples move between sites, document transport and conditioning rather than assuming they arrive unchanged.
2. Fixture geometry and sealing
Fixtures are part of the method. Adapter dimensions, clamping force, insertion depth, alignment, tubing volume, seal material, and wear can influence airflow measurements. For electrical tests, contact location, spring force, surface condition, and device retention can affect the measured path.
Photographs and a controlled fixture drawing are more useful than a method that says only “place the device in the tester.”
3. Electrical contact and lead configuration
Low-resistance measurements are especially sensitive to contact and lead effects. A peer-reviewed study of assembled electronic-cigarette pods found measurable positive bias in its two-wire configuration—the authors identified the error as particularly important for sub-ohm coil designs—while a four-wire constant-current approach produced stable, repeatable observations. In four-wire, or Kelvin, sensing, one pair of leads supplies current and a separate pair measures voltage at the device contacts, reducing the contribution of lead and fixture-contact resistance to the calculated value. The exact apparatus in that study is not a universal factory standard, but the lesson is directly relevant: laboratories must disclose the electrical path and confirm that the fixture does not dominate the result.
If one site uses a handheld meter at exposed contacts and another uses a controlled fixture at the assembled interface, the difference is methodological before it is a supplier-quality conclusion.
4. Instrument status and reference checks
A calibration certificate alone does not prove that today’s complete setup is behaving correctly. Record the instrument ID, calibration status, range, resolution, software version where relevant, and the result of a suitable pre-run reference or zero/leak check.
The reference check should challenge the assembled measurement path, not merely confirm that the instrument powers on.
5. Operator sequence
Removal and reinsertion, cleaning, tightening, stabilization time, reading selection, and retest decisions can all introduce operator effects. The method should state the sequence in enough detail that a second trained operator can reproduce it.
Automation can reduce some manual variation, but it does not rescue an undefined fixture, an unrepresentative sample state, or a flawed calculation.
6. Data handling
Agree on whether the reported value is a single stabilized reading, the mean of repeated readings, or another defined statistic. Preserve raw observations, failed contacts, retests, timestamps, operator IDs, and exclusions.
Do not average away a consistent difference between laboratories. A stable offset may be evidence of systematic bias, not random noise.
A compact method-agreement study for buyers and suppliers
The study only needs to establish whether the method can distinguish parts well enough for the intended commercial decision.
Step 1: Choose transfer samples
Select uniquely identified samples that span the expected working range where practical. Avoid using only nearly identical middle-of-range units; a narrow sample set can make the measurement system look worse relative to real part-to-part variation or hide scale-dependent bias.
If the test could alter the sample, use a matched design and document the order. If it is non-destructive, the same transfer samples can be measured by both sites.
Step 2: Freeze the method package
Issue one controlled revision containing the measurand, sample condition, fixture drawing or photos, instrument settings, reference checks, operator steps, number of repetitions, data rules, and units.
For a compatibility question, the paired interface must also be frozen. A 510 cartridge and battery compatibility test evaluates fit and function for a defined component pairing; freezing that same pairing during a method study prevents interface changes from being mistaken for measurement instability.
Step 3: Measure repeatability
Have an operator measure each sample repeatedly, including removal and reinsertion when that action occurs in normal testing. Repeated readings without disturbing the sample may show instrument stability while missing fixture-placement variation.
Review the raw sequence. Large jumps, contact failures, warm-up drift, or results that depend on insertion order are signals to improve the method before comparing suppliers or lots.
Step 4: Challenge reproducibility
Repeat the study with another trained operator. When the business decision will compare data across sites, include the second instrument, fixture, day, or laboratory that will actually generate those data.
Repeatability asks how tightly the same setup repeats. Reproducibility asks what changes when relevant conditions—such as operator or site—change. CORESTA uses those concepts in its inter-laboratory physical-test studies, where laboratories compare performance and identify improvement actions. That program covers cigarettes and filter rods, not empty vape hardware; it is relevant here only as an example of why within-lab precision and between-lab agreement must be evaluated separately. A method may look precise within one lab and still disagree systematically with another.
Step 5: Compare bias, spread, and ranking
Do not reduce the review to one correlation coefficient. Two sites can rank samples similarly while one reports a persistent offset. Compare at least:
- paired differences for the same samples;
- the average difference and whether it changes across the range;
- within-setup repeat spread;
- operator, fixture, instrument, day, and site patterns;
- whether both methods lead to the same practical decision near specification boundaries.
The acceptance rule should be tied to the intended use and measurement resolution. This article deliberately does not prescribe a universal percentage threshold: a method adequate for screening may be inadequate for a tight acceptance limit.
Step 6: Assign a decision state
Use an explicit outcome rather than “results reviewed.”
| Decision state | Meaning | Buyer action |
|---|---|---|
| Comparable | No material method-related disagreement for the intended decision | Use the agreed method revision and monitor reference checks |
| Comparable with restriction | Results align only within a defined range, setup, or correction | Record the restriction; do not generalize beyond it |
| Method revision required | Fixture, sequence, calculation, or control issue is identified | Correct it and repeat the agreement study |
| Not comparable | The difference cannot be separated from measurement effects | Do not use the two data sets interchangeably for acceptance |
Result-disagreement triage
| Pattern | Parameter affected | Likely method question | Confirmation check | Appropriate disposition |
|---|---|---|---|---|
| One lab is consistently higher | Resistance or airflow | Lead/contact resistance, zero, calibration, calculation, or reference bias | Exchange transfer samples and compare paired differences against a reference check | Correct or restrict the method before judging the lot |
| Results change after reinsertion | Both | Fixture alignment, contact force, seal, insertion depth, or operator sequence | Repeat remove/reinsert cycles with controlled positioning | Improve fixture or procedure |
| Difference grows across the range | Both | Scale-dependent bias, range selection, nonlinearity, or different test condition | Plot paired differences across the working range | Recalibrate, change range, or restrict use |
| Within-lab repeats are stable but sites disagree | Both | Site-specific fixture, instrument, environment, or interpretation | Swap a reference/fixture where feasible and review the method revision line by line | Run a cross-site reconciliation |
| Both labs show wide scatter | Both | Unstable sample state, poor contact/seal, inadequate stabilization, or real heterogeneous parts | Use a stable reference and compare it with production samples | Separate method instability from part variation |
The method-agreement sheet to attach to an RFQ or quality agreement
The following fields form a reusable method-agreement checklist for an RFQ or technical quality agreement. Buyers do not need to dictate every laboratory purchase, but they do need enough definition to make supplied and receiving data comparable.
Record these fields:
- characteristic and exact measurand;
- product code, revision, sample state, and conditioning;
- fixture ID/revision, interface, sealing, and orientation;
- instrument ID/type, range, resolution, and calibration status;
- pre-run reference, zero, leak, or contact check;
- test settings and stabilization rule;
- repetitions, removal/reinstallation rule, and test order;
- operator qualification and operator ID;
- raw-data fields, calculation, rounding, exclusions, and retest rule;
- method revision and change-control owner;
- agreement-study result and any restriction.
This sheet should be settled before a specification becomes a purchasing weapon. Otherwise, a precise-looking limit can create arguments without creating control.
Method validation comes before lot acceptance
Once the method is sufficiently stable and comparable, sampling and acceptance rules can do their intended job. ILEVA’s AQL sampling-plan guide addresses the downstream question of how a lot is sampled and judged. AQL cannot compensate for a measurement system that changes with the lab or operator.
Likewise, a symptom investigation—such as a burnt-taste or dry-hit test—may use resistance or airflow information, but symptom isolation and method agreement are not interchangeable. One asks why a device behaved poorly; the other asks whether the numbers used in that investigation can be trusted across setups.
A fictional example: the “failed” receiving result
In this hypothetical scenario, a supplier and buyer test the same identified AIO transfer samples. The supplier’s readings repeat tightly. The receiving lab’s readings also repeat tightly, but they are consistently higher.
The buyer pauses the rejection decision and compares the method packages. The receiving setup includes an additional contact path and uses a different lead configuration. After the parties align the electrical measurement path and rerun the transfer samples, the systematic difference narrows enough for the intended screening decision.
This is not an ILEVA test result, and it does not suggest that supplier data should be accepted automatically. A lot decision should not absorb an unresolved measurement-system disagreement.
Build comparable evidence before making a commercial decision
Professional hardware qualification is not just a list of tests. It is a chain of controlled definitions: the configuration is frozen, the method is fit for use, the samples are traceable, and the decision state is explicit.
For empty disposable pen programs, buyers can ask ILEVA to align the test-method fields, sample configuration, and evidence expectations before supplier and receiving results are compared.
When two reports disagree, do not begin with the loudest number. Begin by confirming that both laboratories used the same ruler.
