COP explained: the ratio every efficiency claim rests on
Every efficiency claim on an aircon reduces to one division: cooling out over electricity in. That sum is performed once, while the machine holds full output, and the answer then gets quoted as though it described a whole year of running.
By Team Snowflake | Updated 7 Aug 2026
What the ratio actually divides
COP is a division sum with a formal name. NEA defines the full-load version as the ratio of total cooling capacity to effective power input at full load cooling capacity. Cooling produced sits on top. Electricity bought sits underneath. The answer is what gets printed and argued over.
The phrase effective power input is carrying weight there. It means the electricity the whole machine draws, not the share the compressor takes. Fans at both ends and the control board are inside the denominator. A ratio worked out against compressor draw alone would flatter the model by leaving out the rest of what the meter records.
Both halves of the sum are measured in kilowatts, so the units cancel and a bare number is left. That is why nobody writes a unit after it. The figure describes a relationship between two quantities and states neither of them. Which figure from a specification table belongs on top, and which belongs underneath, is settled by the guide separating cooling kW from power kW.
A value above one surprises people, and it should not. An aircon moves heat that the room already held rather than manufacturing anything, and transport costs far less energy than creation would. Every registered household machine on the local market sits well above one on this measure.
One trap follows directly from the ratio being a ratio. It says nothing about size. A small bedroom unit and a machine four times its capacity can post the same figure, because both were measured against their own output. Efficiency and capacity are separate questions, and a strong ratio on an undersized machine buys nothing.
Any two of the three numbers give you the third
Three quantities are locked together by one sum, so a missing one can always be recovered. Cooling capacity divided by the ratio returns the electrical draw. Draw multiplied by the ratio returns the capacity. A sheet listing any two of the three has already told you the third whether it meant to or not.
Use that as a consistency check before trusting a page of figures. Work the arithmetic on the numbers you were handed and see whether the printed answer comes back. Where it does not, the three values were not lifted from the same row, and the usual culprit is a rated capacity paired with a minimum draw from a modulating range.
The check costs nothing and catches a real class of error. Specification tables carry several columns of capacity and several of draw, and a figure copied into a quotation loses the column heading that made it meaningful. Ask which operating row each number came off before treating the set as one machine.
One condition, and what the reading leaves out
The last clause of the definition is where the limits live. At full load cooling capacity means the compressor was held at maximum output for the duration of the measurement. Everything the figure claims belongs to that state and to no other.
The air on both sides was held still as well. A test fixes the indoor condition and the outdoor condition, then keeps them there while the readings are taken. NEA names an international test method for each machine type, and a different one covers a multi-split than covers a single split. Two figures produced under different methods are two different measurements wearing the same label.
So the omissions are large and predictable. The reading knows nothing about the hours a machine spends throttled down, nothing about outdoor air hotter than the test condition, and nothing about the flat the system ends up in. What a rated figure gives up between the chamber and the room has its own guide, and the same reasoning applies to this ratio.
The measurement is not dishonest for any of that. Holding the condition still is the only way two machines can be ranked at all, and a shared condition is what makes any published figure worth reading. The error sits downstream of the test, in reading a controlled comparison as a description of a year.
| Part of the definition | What it pins down | What it therefore excludes |
|---|---|---|
| Part of the definitionTotal cooling capacity | What it pins downHeat removed while the test conditions hold | What it therefore excludesOutput at any other indoor or outdoor condition |
| Part of the definitionEffective power input | What it pins downElectricity drawn by the whole machine | What it therefore excludesNothing, and that is the point of the wording |
| Part of the definitionAt full load | What it pins downThe compressor pinned at maximum output | What it therefore excludesEvery hour the machine spends throttled down |
| Part of the definitionUnder a named test method | What it pins downConditions that two models both had to face | What it therefore excludesComparability with a figure from another method |
A ratio is comparable only inside its own method
Efficiency figures travel badly between documents, and the label attached to one rarely says which method produced it. A brochure from another market can carry one value. A regional review can carry a second. The local registration can carry a third. All three can differ, and none of them has to be wrong.
Treat a number with no metric named beside it as unusable rather than as roughly right. The instinct is to assume the larger figure belongs to the better machine. That assumption holds only when both figures came out of the same method, and it fails quietly the rest of the time, because nothing on the page announces the mismatch.
Why one reading was not left to stand alone
A single full-output reading describes a fixed-speed machine reasonably and a modulating one poorly. A compressor that changes speed spends most of its life away from the condition the test used, so the local criteria stopped resting on that reading alone for those models.
For a machine that modulates, NEA blends two readings into a weighted COP. More of that blend rests on the reading taken at half output than on the one taken at full output. The reasoning behind the weighting, and the seasonal metric that carries it further, belong to the guide on part-load efficiency rather than here.
The two scales are not interchangeable, and NEA has published the conversion that proves it. In a November 2024 presentation on the labelling and standards regime, the agency gave CSPF as 1.1917 multiplied by COP, plus 0.3111. It then set each current split-type minimum beside its seasonal equivalent, and every equivalent lands more than a full point above the figure it was converted from.
Read those pairs as a warning about arithmetic rather than as a lookup table. The same machine scores higher on the seasonal scale than on the single-point one, by construction. A specification sheet quoting a seasonal figure beside a local requirement written in COP has placed two different scales side by side, and the model looks stronger than the comparison actually shows.
Note what that conversion is and is not. A straight line fitted across a population estimates where a whole market sits, which is what a regulator needs when setting a level. It cannot translate one particular machine, because two units posting the same single-point figure can behave differently once their part-load behaviour is measured. Running the formula on a model from a showroom floor produces a number nobody tested.
The direction of travel is toward more sampling points, not fewer. Three-phase commercial equipment is graded on an integrated figure that blends four COP readings taken at quarter, half, three-quarter and full output, with most of the weight on the three-quarter reading. EER and SEER appear on imported sheets as the same idea expressed on a different unit basis.
The local floor is written in this ratio
This is not a spec-sheet curiosity, because the efficiency minimum for household aircon is stated in the same quantity. NEA writes the requirement as a full-load figure where the compressor runs at one speed, and as a weighted one where the machine modulates, with a standby power ceiling beside it. The MEPS floor guide carries the current values and which tick bands still hold products.
That places the ratio upstream of everything else printed about a machine. A model that misses the number cannot lawfully be supplied here, so no label is ever printed for it and no showroom ever displays it. Every efficiency figure a buyer sees belongs to a machine that already cleared this measure.
It is also the one term in the whole transaction that cannot be renegotiated. Capacity can be revised, the outdoor position can be argued about, the pipe route can be improved and the price can move. The ratio was fixed in a factory and tested before the machine reached the country. Whatever a buyer settles on, that figure is what the system will carry for its whole service life.
Push back on any efficiency claim that arrives without a metric attached to it. Ask which measure the number is, and at what load it was taken. A seller quoting a large figure from an overseas sheet against a local threshold is not lying and is still not answering, because the two numbers were never on the same scale.
Why a regulator writes a floor as a ratio
A ratio has one property that makes it usable as law: it is free of scale. A bedroom unit and a system serving four rooms both produce a figure between roughly three and six, because each was measured against its own output. One threshold therefore governs an entire class of machine, whatever size the machines are.
Write the minimum in kilowatts instead and it falls apart immediately. A ceiling on electrical draw would ban large systems and wave through small inefficient ones. A floor on cooling output would say nothing about consumption at all. Neither quantity means anything on its own, which is the same reason neither belongs in a running-cost claim by itself.
The same property is what lets the figure travel where nothing else on a sheet can. Capacity cannot be compared across a bedroom unit and a whole-flat system, and draw cannot either. The ratio holds two machines of different sizes against a single yardstick, which is exactly why the tick bands could be built on it.
One boundary survives all of that, and it is the boundary this page keeps returning to. The figure travels across brands, capacities and formats. It does not travel across metrics. A number is only usable next to another number produced by the same method, and the scale-free property does nothing to rescue a comparison between two different measures.
What to do with the figure you are handed
The ratio ranks machines and refuses to forecast bills. Given two models treated identically, the higher figure buys its cooling with less electricity, and that ordering holds. What neither figure knows is how many hours the machine will run, what setpoint it will be asked to hold, or what the room will throw at it.
So the useful question at the counter is narrow. Establish that every figure on the shortlist is the same measure taken the same way, then read the gap between them. A gap inside a single tick band is real but small, and it moves a bill less than sizing, condenser position and the standard of the install.
The table below sorts the forms an efficiency claim arrives in. The middle column is what the number is genuinely a ratio of. The right-hand column is what to settle before letting it decide anything.
Where a figure cannot be produced at all, that is information too. A registered model has a tested value sitting behind its band, and a retailer who cannot reach it has described the shortlist rather than narrowed it.
- Ask for the tested figure and the metric name together. A number without its metric cannot be compared with anything.
- Ask at what load the figure was taken. Full output and a blended value answer different questions about the same machine.
- Keep one measure across the whole shortlist. Mixing a single-point value with a seasonal one ranks the scales, not the machines.
- Read a ratio as a ranking tool rather than a cost forecast. Runtime and conditions decide a running cost estimate, and no published figure has seen either.
- Treat a strong ratio on the wrong capacity as no advantage. Efficiency is measured against a machine's own output, so it says nothing about fit.
| How the claim reaches you | What the number is a ratio of | What to settle before ranking on it |
|---|---|---|
| How the claim reaches youA single figure on a brochure | What the number is a ratio ofCooling at full output over what that output cost | What to settle before ranking on itWhether the rival model's figure was taken the same way |
| How the claim reaches youA weighted figure on a registered model | What the number is a ratio ofTwo readings blended, leaning on the lower output | What to settle before ranking on itThat every model quoted uses the identical blend |
| How the claim reaches youA seasonal figure from an overseas sheet | What the number is a ratio ofA spread of conditions averaged on a foreign profile | What to settle before ranking on itThat nobody is reading it against a local threshold |
| How the claim reaches youIt meets the national standard | What the number is a ratio ofNothing specific to this model at all | What to settle before ranking on itHow far above the minimum the tested figure sits |
| How the claim reaches youThis one will cut your bill | What the number is a ratio ofNothing measured anywhere | What to settle before ranking on itThe hours, setpoint and room behind the promise |
Ready to get started?
Tell us what’s going on. Symptoms, setup, photos, anything we should know. We’ll assess and come back with the right next step.