Our method, in plain sight

Where the data comes from

• Specs and images come from Open Icecat, matched to each unit by its GTIN or model number, so the cooling power, noise, dimensions and energy rating are the manufacturer's own published figures.
• Prices and stock come live from the retailer's own product feed through CJ Affiliate, refreshed on a schedule, so the price you see next to Buy is the price the retailer is charging, not a number we set.
• Price history and deals. We record each unit's price at every retailer once a day. That lets us tell you honestly when a price is at the lowest we have ever tracked, always dated and named to one seller, never a market average or a forecast. A unit reaches the deals page only when its current in-stock price is that tracked low AND it is a real markdown of at least 5% and $20 off a higher price we genuinely observed, never off a manufacturer list price.
• Efficiency for US window and portable units is matched to the ENERGY STAR certified-products database by barcode, so for a certified unit its CEER and annual energy use are the official certified figures, not ours. For the few units the manufacturer publishes a CEER outside that database, we show the manufacturer's own rating, still a real published figure, never one we invented.
• Running cost is the one number we estimate rather than read off a source, and we say so: for those ENERGY STAR units we price their published annual energy use at the average US residential electricity rate (EIA) to get a standardized dollars-per-year for comparison. We keep it national on purpose (pricing it by your state would read backwards, since the hottest states have the cheapest power), show it only where the data exists, and never convert a mini-split, which is rated on the different SEER2 scale, to a dollar figure.
• If a maker does not publish a figure, it stays blank. We mark it "Not published" rather than estimate it. A number you see is a number a real source states.

How we rank and sort

• The default view leads with units you can actually buy in stock, then the ones with the most complete specs, because a decision needs the numbers to be there.
• Tell us your room and every unit shows whether it fits, sized from the maker's stated coverage or, where none is published, estimated from cooling power and clearly marked as an estimate.
• When you pick your region, we also size for its climate: a hot or humid place needs more cooling per square metre than a mild one, so we scale the need using published ASHRAE design temperatures and Manual-J regional practice. We always show both your real room size and the figure we size against, and when your climate is unknown we make no adjustment.
• A whole-home system gets no room verdict. A multi-zone or central ducted system splits its cooling across several rooms or through ductwork, so matching its total capacity to one room would overstate what that room actually gets. We label it a whole-home system and leave the sizing to the installer, who sizes each room.
• A blank cannot satisfy a noise limit. Set one and units whose maker publishes no decibel figure drop out, with the count shown, because we cannot claim a unit is quiet enough when nobody has said how quiet it is. A missing price is treated differently and kept in view: it is a gap in what we know, not an unverified claim about the unit.
• You can re-sort by price, by quietest, by cooling power, or by efficiency where it is rated. Units with no efficiency rating are shown last, never given an invented score.
• No one can pay for placement. Ranking is the same for every unit.

How we make money

When you click Buy and purchase from a retailer, compare.ac may earn an affiliate commission. It comes out of the retailer's margin, so it never changes the price you pay.

It also never changes how we rank or which units we show. We would rather send you to the right unit once than the wrong unit repeatedly, so the incentive and the advice point the same way.