Short answer
A minimum measured quantity is the smallest volume a meter is allowed to report as a valid result. OIML R 117-1 requires a dispenser rated up to 60 litres per minute to carry an MMQ not exceeding 5 litres. Below that floor the number still appears, but metrology does not accept it as a measurement of anything.
Key takeaways
- OIML R 117-1 §5.1.8 requires a dispenser rated up to 3.6 m³/h (60 L/min) to declare a minimum measured quantity not exceeding 5 litres.
- The MMQ exists because a meter's error is not proportional at small volumes, the same absolute error becomes a large percentage of a tiny draft.
- OIML places these devices in accuracy class 0.5, with maximum permissible errors of 0.5% and 0.3% depending on the line applied.
- NIST Handbook 44 makes the same point from the other direction: a bulk-delivery test draft may never be smaller than 180 litres.
- A dashboard alert set below the metrological floor is not more sensitive, it is producing arithmetic on quantities no standard will vouch for.
01What a minimum measured quantity is
Legal metrology does not simply state how accurate a fuel meter must be. It also states the smallest quantity that meter is permitted to report as a result, and that figure has a formal name: the minimum measured quantity. OIML R 117-1 §5.1.8 sets it for dispensers by flow rate, requiring that a device with a maximum flowrate up to 3.6 cubic metres per hour, 60 litres per minute, which covers ordinary road-fuel dispensing, must have a minimum measured quantity not exceeding 5 litres. The wording is worth reading carefully. It does not say the meter stops working below 5 litres. It says a volume smaller than the declared minimum is outside the range the device is certified over, so the number it produces carries no guarantee. The reading exists; the standing behind it does not.
02Why the floor exists at all
The reason is that error does not shrink in step with volume. A meter carries a fixed component of uncertainty, the resolution of its register, the behaviour of its moving parts at start and stop, the air it may have admitted at the beginning of a draft, and that component does not care how much liquid passed through. Expressed as a percentage, it therefore grows as the draft shrinks. OIML places these devices in accuracy class 0.5, with maximum permissible errors of 0.5% and 0.3% depending on which line of the table applies, and those percentages are only meaningful over a range where the fixed component is small compared with the total. Set the draft small enough and the same absolute error that represented a tenth of a percent becomes tens of percent. Rather than publish a percentage that degrades without limit, the standard draws a line and declines to certify below it.
03NIST reaches the same conclusion from the other side
The United States handles the same problem through its test procedure rather than through a declared device parameter, and the arithmetic lands in the same place. NIST Handbook 44 §3.31 requires that a test draft for a vehicle-tank meter be at least the volume delivered in one minute at maximum rate, and never less than 180 litres. That is a floor on the size of the measurement used to judge the device. The reasoning is identical to the OIML case: a small draft cannot resolve the meter's error well enough for the test to mean anything, so the procedure refuses to be run small. Two independent metrological traditions, one setting a device parameter and one setting a test volume, both conclude that below some quantity a fuel measurement stops being a measurement. That convergence is the strongest evidence available that the floor is real rather than bureaucratic.
04Where a fleet dashboard crosses the line
Now compare that with how fuel alerts are usually configured. A common instinct is to set the alert threshold as low as the interface allows, on the theory that a tighter threshold catches more theft. Two or three litres is a typical choice. But a tank-level channel is a less favourable instrument than a certified dispenser, not a more favourable one: it infers volume from a height through a calibration table, rather than counting liquid through a chamber. Technoton publishes an inaccuracy of not more than 0.2% for stationary tanks and 1% for moving vehicles, and on our own catalogue the OMNICOMM probe is listed at plus or minus 0.5% over a 0 to 4,000 mm range. On a large tank, half a percent is already several litres. A two-litre threshold therefore sits below the instrument's own uncertainty and beneath the metrological floor at once.
05The quantisation floor beneath the accuracy floor
There is a further floor below the accuracy one, and it applies specifically to a level taken from a vehicle bus. The FMS standard defines SPN 96 with a gain of 0.4% per bit, which means the value arriving over CAN is quantised: it can only take steps of four thousandths of the tank's capacity. On a 400-litre tank that is about 1.6 litres per step, so a genuine loss of one litre may not change the transmitted value at all, while a loss of two litres may appear as either one step or two depending on where the true level sat within a step. The standard is candid about what this does and does not tell you, stating in its own words that the mentioned resolution is not related to the accuracy of the signal. Resolution is the size of the step, not the trustworthiness of the number.
06What a threshold below the floor actually produces
A threshold set finer than the measurement chain can support does not detect small thefts. It detects the measurement chain. Every source of variation the sections above described, tilt, a slosh, a temperature-driven volume change, a quantisation step, ordinary probe noise, will at some point cross a two-litre line, and each crossing arrives as an alert indistinguishable in form from a real one. The result is a queue that grows faster than anyone can review it, and the predictable human response is to stop reviewing. So an over-tight threshold does not merely add work; it removes the attention that the genuinely large events needed. The alert list becomes arithmetic performed on quantities no standard will vouch for, and its length is mistaken for vigilance.
07How to choose a threshold you can defend
The sequence is short. Start from the instrument, not the interface: take the published accuracy percentage and multiply it by the tank capacity to get an absolute figure in litres, then treat that as your resolution floor. If the level arrives over CAN, add the quantisation step, four thousandths of capacity per bit, because a change smaller than one step cannot be seen at all. Compare the resulting floor with the smallest theft that would actually justify an investigation, and if the floor is larger, accept that per-event detection is the wrong instrument for that size of loss and use reconciliation over a longer window instead. Then set the threshold above the floor, write down which figures produced it, and revisit it when a tank, a sensor or a capacity configuration changes.
| Floor or setting | Published value | What it constrains |
|---|---|---|
| OIML minimum measured quantity, dispenser up to 60 L/min | not exceeding 5 L | The smallest volume the meter may report as a result |
| OIML accuracy class for these devices | class 0.5; MPE 0.5% and 0.3% | The permitted error over the certified range |
| NIST bulk-delivery test draft | never less than 180 L | The smallest volume a compliance test may use |
| NIST class boundary by flow rate | Class 0.3 above 115 L/min; Class 0.5 from 4 L/min | Which accuracy class a device falls into |
| CAN fuel level quantisation step (SPN 96) | 0.4% of capacity per bit | The smallest change that can appear at all |
| Typical dashboard theft threshold | two or three litres | Nothing, it sits below every floor above |
Figures and sources
- OIML minimum measured quantity for dispensers, the floor of a metrologically credible volume: max flowrate up to 3.6 m³/h (60 L/min): MMQ not exceeding 5 L (Source: OIML R 117-1:2007(E), §5.1.8, as of 2026-08-31)
- OIML accuracy class and maximum permissible error for fuel dispensers and road tankers: accuracy class 0.5; MPE 0.5% (line A) and 0.3% (line B) (Source: OIML R 117-1:2007(E), Tables 1 and 2, as of 2026-08-31)
- NIST bulk-delivery test draft floor, the same idea expressed as a test volume: at least the volume delivered in 1 minute at maximum rate, and never less than 180 L (50 gal) (Source: NIST Handbook 44 (2023) §3.31, N.3 Test Drafts, as of 2026-08-31)
- NIST metrological boundary between bulk-delivery and motor-fuel measuring devices, with accuracy classes: Class 0.3 above 115 L/min (30 gpm); Class 0.5 from 4 L/min through 115 L/min (Source: NIST Handbook 44 (2023) §3.31 Vehicle-Tank Meters, Table 1, as of 2026-08-31)
- FMS/J1939 fuel level quantisation, a second floor beneath the accuracy floor: SPN 96: 0.4 % per bit gain, PGN 65276 repetition 1,000 ms (≈1.6 L per bit on a 400 L tank) (Source: FMS-Standard Interface Description Version 05 (07.07.2024), §1.1.2, as of 2026-08-31)
- FMS-Standard warning that the stated resolution is not a statement of accuracy: "The mentioned resolution is not related to the accuracy of the signal" (Source: FMS-Standard Interface Description Version 05 (07.07.2024), as of 2026-08-31)
- Capacitive fuel level sensor accuracy, which sets the absolute litre floor once multiplied by capacity: not more than 0.2% stationary; 1% for moving vehicles (Source: Technoton, Fuel level sensor DUT-E product specifications, as of 2026-08-31)
- OMNICOMM fuel level sensor accuracy and measurement range on the ENTEK.AI catalogue: ±0.5% over a 0–4,000 mm measurement range (Source: ENTEK.AI product catalogue, as of 2026-08-31)
Questions
What exactly is a minimum measured quantity?
It is the smallest volume a meter is certified to report as a valid result. OIML R 117-1 §5.1.8 requires a dispenser with a maximum flowrate up to 3.6 m³/h, 60 litres per minute, to declare a minimum measured quantity not exceeding 5 litres. Below it the meter still shows a number, but nothing certifies that number.
Does the MMQ apply to a tank level sensor too?
Not as a legal requirement, because a level probe is not a certified custody-transfer meter. But the underlying reason applies with more force, not less: a level sensor infers volume from a height through a calibration table, so its own accuracy percentage multiplied by tank capacity gives you an equivalent floor in litres.
So how do I catch a small theft?
Not with a per-event threshold, because a single small event is below the resolution of the instrument. The alternative is reconciliation over a longer window, where repeated small losses accumulate into a discrepancy large enough to exceed the measurement noise. That is a different instrument for a different question.
Is a tighter threshold ever the right answer?
Only when the measurement chain got better, not when the interface allowed a smaller number. Tightening the threshold below the instrument floor does not add sensitivity; it converts ordinary measurement variation into alerts, and the resulting queue costs you the attention that the large events needed.