CSPF explained: why two 5-tick aircons are not equal

CSPF stands for Cooling Seasonal Performance Factor. It is a ratio. Take all the cooling a unit delivers over a full cooling season, then divide it by all the electricity it consumed producing that cooling. Both quantities are energy, so the units cancel and the answer is a bare number. Higher means more cooling wrung out of the same electricity.

The word seasonal is doing the work in that name. A season is not one condition. Outdoor temperature moves, the load in the room moves with it, and the machine spends different amounts of time at each point. CSPF accounts for that spread. It weights every operating condition by the hours the machine is expected to sit there, then totals the lot.

That weighting is what separates CSPF from a single-point rating. A single-point figure records one moment: one outdoor temperature, one indoor temperature, one compressor speed. It answers how the machine behaves under a defined test condition. CSPF answers how the machine behaves across a range of them, which is nearer to the question a buyer is actually asking.

The number carries no unit, which is why it can be read directly. A value of five means the machine moved five units of heat for every unit of electricity it drew, averaged over the whole profile. A ratio well above one is normal and not a contradiction. An aircon relocates heat rather than creating cold, and relocating heat costs far less energy than generating a temperature difference outright.

An aircon spends most of its running hours below full capacity. Full capacity is what the machine needs on the hottest afternoon, with the room still holding the heat of the day. Once the room reaches the set temperature, demand collapses. From there the unit is only replacing heat leaking back in, and that is a fraction of what it can produce.

A full-load-only figure therefore describes a condition the machine rarely occupies. Rank two units on that condition alone and the order can invert against real use. A machine tuned to look strong at maximum output can be ordinary at the low outputs it will hold through most of a night.

Weight the measurement toward the hours the machine actually accumulates and the ranking changes character. It stops describing what the unit can survive and starts describing what it will cost. That is the entire argument for seasonal weighting, and it is why the metric exists at all.

COP and CSPF express the same idea at different resolutions. Both set cooling delivered against electricity drawn. COP does it at one operating point. CSPF does it across a weighted spread of operating points. Neither is more correct; they answer different questions.

COP is the right figure when the question is capability. It states what a machine can do under a named condition, which is what a technician needs to check a system against its specification or to judge whether it is underperforming. A snapshot is exactly right for that job.

CSPF is the right figure when the question is consumption. Nobody holds an aircon at one condition for a year. Averaging across a load profile produces a number that tracks the electricity meter more closely than any single reading can.

The energy label does not print an efficiency ratio at all. It carries the tick rating, the brand, the model and its type, and the cooling capacity. It also carries an annual cost estimate, an annual consumption estimate, and the registration number. The ratio itself sits behind the ticks, inside the criteria that decide which band a model lands in.

For household split systems NEA writes those criteria in COP, and for inverter models in weighted COP. NEA defines weighted COP as 0.4 x COP at full load plus 0.6 x COP at half load. Read the weights before moving on. Sixty percent of the figure rests on the half-load reading, not the full-load one.

So the part-load argument is already inside the Singapore criteria. The band a model clears is not resting on a full-load number. It is resting on a two-point blend that leans toward part load, which is the same reasoning CSPF applies at higher resolution and across many more conditions.

A tick is a band, not a score. Every model clearing the threshold for a band shows the same number of ticks, whether it cleared by a hair or by a wide margin. Two units can sit at opposite ends of one band and still print an identical label.

That is the practical trap in comparing shortlisted models. The tick count filters, then it stops. It does not rank within itself. Once two candidates share a band, the label has said everything it can say, and any further separation has to come from the figure underneath.

Ask the retailer or contractor for the efficiency figure on each model, and ask which metric that figure is. A registered model has a value behind its band. A manufacturer publishing CSPF will carry it on the specification sheet. Either is usable for a comparison. Mixing the two is not.

Push back on a recommendation that still leans on tick count once the shortlist has narrowed to models sharing a band. At that point the tick has stopped discriminating between the options in front of you. A recommendation resting on it is resting on nothing, and the question to put back is simple: which model has the higher figure, and on which metric.

Weigh a small gap in that figure against the rest of the quote before letting it decide. A narrow efficiency difference inside one band moves a running bill less than how the system is sized, sited and installed. Ask for the figure because it is the only thing separating two tied models. Do not treat it as the largest number on the page.

Every seasonal figure is produced on a standard profile. The method fixes the outdoor conditions, the hours assigned to each, the indoor conditions, and the load the machine is asked to meet. That standardisation is what makes two models comparable. It is also what stops the number from predicting your bill.

A real flat departs from the profile in several ways at once. Runtime is the obvious one, because the profile assumes a usage pattern and yours will not match it. Room load is the quiet one: west-facing rooms, top floors and large glass frontages all raise the heat the machine has to remove. A lower setpoint holds the compressor at higher output for longer. People and appliances add heat the profile never budgeted for.

Installation and condition move the number further still. A crowded outdoor unit, a system sized wrong for the room, fouled coils, or a slow refrigerant loss all drag real performance below the rated figure. None of that is visible in any published value. A high seasonal figure sets the ceiling for a machine. Everything after the purchase decides how near the system gets to it.

The correct use of CSPF is ranking, not forecasting. It tells you which of two machines turns electricity into cooling more efficiently under identical treatment. It does not tell you what either will cost in a particular home. Any figure presented as a predicted saving is carrying assumptions that were never measured on your flat.