Two labs, different oil analysis results: which one to act on?

This check covers numeric used-oil test results, such as wear metals, base number, viscosity and water, from methods that publish a precision statement. It does not settle results from methods without a reproducibility statement, from on-site screening instruments, or from samples that are not homogeneous, and it does not replace the dispute clause in a supply contract.

When should you compare two labs' oil analysis results?

Compare two labs' oil analysis results when a decision is expensive enough that one report should not carry it alone, or when you are about to change labs. Four situations are common. First, a report flags a wear metal or base number problem that would justify an oil change, a teardown or a warranty claim, and someone sends a split of the same oil to a second lab for confirmation. Second, a site or a plant group is moving its routine samples from one laboratory to another and needs to know whether the old trend can continue. Third, an equipment maker or a lubricant supplier tests the same oil in its own laboratory and reaches a different number than the user's lab. OELCHECK describes this case: a press maker had a hydraulic oil tested in its US laboratory and then by OELCHECK, and some values differed while the iron wear signal was clear in both (OELCHECK, 2015). Fourth, a supplier and a receiver disagree over whether a delivered oil meets its specification, the use case for which ASTM D3244-24 was written: a supplier lab tests before release and a receiver lab tests on arrival (ASTM D3244-24, 2024).

In each oil analysis case the question is the same: are the two results different measurements of one thing, or measurements of two different things? The answer decides whether you average, choose, retest or resample. Comparing without that question tends to end with whoever holds the more alarming number winning the argument.

What do you need before judging two oil analysis results?

Before judging two oil analysis results, you need the method behind each number, the identity and history of each sample, the precision statement of each method, and evidence that each lab's measurement system was in control. Collect five things. One, both full reports, with the test method designation and edition for every result, not only the test name; a line that says "TBN" or "iron" without a method is not comparable evidence. Two, the sample records: sample point, sampling date, whether the two labs received portions of one bottle or two separate pulls, and the date each lab tested. Three, the precision section of each method, with repeatability and reproducibility at the level of the results; ASTM D6300-26 notes that repeatability and reproducibility are defined in the precision section of every Committee D02 test method (ASTM D6300-26, 2026). Four, your machine's trend history from the regular lab, at least the last three or four results for the same test. Five, the labs' quality evidence: control charts kept under ASTM D6299-26a, which the practice says may be used to continuously demonstrate the proficiency of analytical measurement systems (ASTM D6299-26a, 2026), proficiency test results evaluated with a guide such as ASTM D7372-25a (ASTM D7372-25a, 2025), and accreditation to ISO/IEC 17025:2017, which ISO confirmed in 2023 and lists as current (ISO/IEC 17025:2017).

Without the method and the sample records, the oil analysis comparison stops at step one, because no precision statement applies to two numbers whose origin is unknown.

How do you decide which oil analysis result to act on, step by step?

To decide which oil analysis result to act on, run five criteria in order: method match, sample identity and age, the gap against reproducibility, agreement with the one-lab trend, and a retest of the retained sample. The order matters because a method mismatch or a sample mismatch makes the precision check meaningless. The table below sets the threshold for each criterion and what follows from it.

The key row is the gap against reproducibility: when two labs used the same method on one split sample and their results differ by no more than the method's reproducibility, neither lab is wrong, and the result that belongs in the decision is the one from the lab that holds the trend. A gap above reproducibility does not yet say which lab erred; it says only that the pair is unusual enough to retest. Stop at the first criterion that fails and resolve it before moving down the table, because a later check cannot repair an earlier mismatch. Record the outcome of every row, including the rows you passed, so the next person can see why one result was used and the other was not.

Why do two labs report different oil analysis results on the same oil?

Two labs report different oil analysis results on the same oil for three reasons: they ran different methods, they tested different samples, or they ran the same method on the same oil and normal between-lab scatter separated the numbers. The first reason is the most common trap, because two reports can print the same test name over different procedures.

Base number is the clearest case. ASTM D2896-26 titrates basic constituents with perchloric acid in glacial acetic acid (ASTM D2896-26, 2026), while ASTM D4739-26 uses hydrochloric acid, a weaker titrant, and a different solvent (ASTM D4739-26, 2026). ASTM D4739-26 explains that weakly basic degradation products are possible in in-service oil and that D2896 will titrate these, giving a false value of essential basicity; it prefers D4739 for in-service or at-term oil and D2896 when the base number of new oil is required as an expression of its manufactured quality (ASTM D4739-26, 2026). It also records that, in ASTM interlaboratory crosscheck programs for new and used lubricants, D2896 has historically given the higher value (ASTM D4739-26, 2026). A lab reporting 7 mg KOH/g by D2896 and another reporting 5 mg KOH/g by D4739 on the same oil may both be right (illustrative values).

National standards add a second layer. In the OELCHECK case, the US laboratory and the German laboratory worked to different standards, ASTM and DIN, for elements, acid number and water (OELCHECK, 2015). Only after the method question is settled does between-lab precision explain what remains of the gap.

How far apart can two labs' results be under ASTM reproducibility?

Two labs' results can be as far apart as the method's published reproducibility, R, at the level of the results, and a valid pair should exceed R only in about 1 case in 20. Reproducibility conditions mean test results obtained with the same method on identical test items in different laboratories with different operators using different equipment (NIST OSAC lexicon, citing ASTM E1732, 2025). TestOil explains that most ASTM methods report repeatability and reproducibility at a 95 % confidence level, meaning the values can only exceed 1 case in 20 (Zabawski, TestOil, 2019).

R depends on the test and the level. TestOil gives elemental spectroscopy examples at 10 ppm: copper with a repeatability of 1.0 ppm and a reproducibility of 2.4 ppm, and lead with 3.3 ppm and 6.9 ppm (Zabawski, TestOil, 2019). The same article puts viscosity variability at about 1 % and particle count variability above 100 %, enough for a two-fold difference or one ISO code (Zabawski, TestOil, 2019). OELCHECK cites DIN limits for iron: between laboratories, deviations of 60 % from the mean are tolerable at 1 to 10 mg/kg and 25 % at 10 to 1,000 mg/kg, so at a mean of 50 mg/kg, 44 and 56 mg/kg are both valid (OELCHECK, 2015).

The OELCHECK English text calls the within-lab measure "reproducibility" and the between-lab measure "comparability" (OELCHECK, 2015), while ASTM uses repeatability for within-lab and reproducibility for between-lab. Before you compare a gap with a limit, confirm which of the two limits a source is giving you.

How do sampling and storage change a second lab's result?

Sampling and storage change a second lab's oil analysis result whenever the two labs did not test portions of the same well-mixed bottle within a similar time. Two separate pulls are two samples, and their difference includes everything that varies between draws: sample point, flow, temperature, settling time and flushing volume. ASTM D4057-22 notes inherent limitations in any sampling that may affect how representative a sample is; a spot sample, for example, represents only one particular point in a tank or line (ASTM D4057-22, 2022). For hydraulic systems, ISO 4021:1992 prefers extracting samples from a main flowline so that the particulate contaminant in the sample represents the fluid flowing at the sampling point, and allows reservoir sampling only when no suitable sampler is fitted (ISO 4021:1992).

Storage and transport add their own change. ASTM D6300-26 states that, for properties known to change over transport or storage, reproducibility obtained by shipping samples to different laboratories also includes those property changes, so it is not possible to obtain reproducibility due to testing only (ASTM D6300-26, 2026). Particles can settle and free water can separate in a standing bottle, which likely explains many particle count and water disagreements; a lab that does not re-homogenize before subsampling will read the top or the bottom of the bottle.

The fix is a split protocol: draw one sample, mix it, divide it into labeled portions at the same time, ship them together, and keep one portion as the retained sample.

When should a referee lab or retained sample settle the difference?

A retained sample should settle the oil analysis difference when both labs used the same method on one split sample and the gap still exceeds reproducibility; a referee lab should settle it when the retests still disagree. Ask both labs to retest the retained portion first, because a single transcription, dilution or calibration slip explains many outliers and a retest costs less than a third lab.

Before trusting either retest, check the lab's quality evidence. ASTM D3244-24 sets prerequisites before lab results can be used in its dispute procedure: site precision under ASTM D6299, substantiated by control charts, and proficiency demonstrated either with an agreed check standard within plus or minus 0.7 times the published reproducibility of the accepted reference value, or with a Z-score within plus or minus 2.0 from an agreed interlaboratory exchange program (ASTM D3244-24, 2024). Those prerequisites are a good checklist even outside a formal dispute.

If a referee is needed, agree on the lab, the method and edition, and the decision rule before the sample is shipped. ASTM D3244-24 requires the acceptance limit to be determined before testing begins and recommends that the practice be conducted under the guidance of a qualified statistician (ASTM D3244-24, 2024). It also states that combining results from more than three labs is beyond its scope (ASTM D3244-24, 2024). For a supplier and receiver arguing over new oil, ISO 4259-2:2026 specifies how to determine conformance status when supplier and receiver results conflict (ISO 4259-2:2026).

Which oil analysis result should go into the trend?

The oil analysis result that should go into the trend is the one from the lab and method that produced the trend history, once that lab's result has passed the method, sample and gap checks. ASTM D3244-24 reflects this: when multiple results for one batch come from a single laboratory, site precision R' under ASTM D6299 is used instead of the published reproducibility R (ASTM D3244-24, 2024). A trend built from one lab therefore shows smaller real changes than a trend that mixes labs.

Method consistency matters as much as lab consistency. ASTM D4739-26 says that when the loss of base number in service is the consideration, all values including the unused oil shall be determined by D4739, and its results shall not be related to base numbers by another method such as D2896 (ASTM D4739-26, 2026). Record the second lab's result as an annotation, not a data point, and do not average two labs into one trend value.

If the regular lab failed its retest or its quality checks, start a new baseline at the new lab with a fresh new-oil reference and two or three in-service samples before judging rates of change. ASTM D7669-20 also notes that the range between control limits should be greater than the range within each test method's repeatability (ASTM D7669-20, 2020). Once a confirmed result crosses an alarm limit, the decision moves to the oil itself, as in used oil analysis results: keep running, top up or change oil.

Worked two-lab oil analysis example (illustrative)

In this illustrative two-lab oil analysis example, a standby diesel generator's engine oil is tested by the site's regular lab and by a second lab used by the engine dealer, and the two reports disagree on base number and iron. All values, and the assumed reproducibility, are invented for illustration. The regular lab's iron trend on the three previous samples was 30, 33 and 36 mg/kg. The table shows each result and the decision it supports.

The key row is the retained sample retest: the gap fell from 14 to 8 mg/kg, inside the assumed reproducibility of 11 mg/kg, so in this illustrative case neither lab was wrong and the regular lab's 38 mg/kg stays in the trend. The base number gap of 1.3 mg KOH/g is not a disagreement at all; the two labs used different titrants, and D2896 is expected to read higher (ASTM D4739-26, 2026).

The useful finding is not which lab was right but that iron has risen four samples in a row in one lab's trend. That rise, not the second lab's higher number, is what justifies an inspection. In a real case, take R for each test from the precision section of the method edition both labs used.

How should you report a two-lab oil analysis comparison?

A two-lab oil analysis comparison should be reported as a short record that someone else can check: what was compared, which criteria passed, which result was used and why, and what is still open. Put the sample identity first: machine, sample point, sampling date, whether the portions were split from one bottle, and the test date at each lab. Then list each test in its own row with both results, both method designations and editions, the reproducibility used and its source, the gap, and the decision: compared and within R, compared and outside R, or not compared because of a method or sample mismatch.

Keep the open items separate from the conclusions. A retest requested but not yet returned, a referee lab agreed but not yet run, or a precision value taken from an older method edition each belongs on an open list, not in the verdict. Mark which result entered the trend and which is an annotation, so the next reader does not average them. If the comparison supports a maintenance action, state that the action rests on the one-lab trend and the confirmed result, not on the second report alone. In Lubinpla, a performance evaluation task ties each oil analysis result to its sample and test method, judges whether the test round is valid before comparing values, and lists unconfirmed results separately in a cited document for your review.

Where does this two-lab oil analysis check not apply?

A two-lab oil analysis check does not apply where no valid precision statement covers the test, the level or the sample. ASTM D3244-24 applies only to test methods that specifically state their repeatability and reproducibility values conform to its definitions (ASTM D3244-24, 2024), and many screening tests and in-house procedures publish no reproducibility at all. A precision statement also covers only the range it was derived from: ASTM D2896-26 can measure base numbers above 300 mg KOH/g, but its precision statement was obtained only on base numbers of 300 mg KOH/g or less (ASTM D2896-26, 2026).

Samples that are not homogeneous fall outside the method too. ISO 4259-2:2026 was designed for petroleum and related products, which are normally homogeneous, and states that it is not applicable to non-homogeneous products (ISO 4259-2:2026). A grease sample, a sludge-laden drain sample or a bottle with free water may need mixing or a different approach before any precision comparison means anything.

The check also cannot replace engineering limits. A gap within reproducibility says the two labs agree; it says nothing about whether the agreed value is acceptable for the machine. Alarm limits come from the equipment maker, the lubricant supplier or a statistical review of your own data, and a confirmed result still has to be read against them. Finally, where a contract names a dispute procedure, that procedure governs the two-lab oil analysis comparison, not this guide.

Key takeaways

• Compare methods first; two results by different methods, such as base number by ASTM D2896 and ASTM D4739, are not a disagreement.

• Compare portions split from one well-mixed bottle; two pulls are two samples.

• A gap within the method's reproducibility means both labs are valid.

• Retest the retained sample before paying for a referee lab, and agree the referee rule before testing.

• Trend one lab and one method; record the second lab as an annotation.