How we calculate

Data sources

Our figures come from two public sources only:

• EPREL, the European Product Registry for Energy Labelling, run by the European Commission. Suppliers must register every model with an EU energy label before selling it in the EU, with the values shown on the label. We read the public part of the register through its official interface.
• Eurostat, the statistical office of the European Union, dataset nrg_pc_204: electricity prices for household consumers, published for each half-year.

We add no values from brochures, shops or our own measurements. When a value is missing from EPREL, the page shows a dash or leaves the line out; we do not estimate it.

We rebuild the database regularly from both sources. Each rebuild reads every product group in full and recomputes all statistics, rankings and alternatives from scratch. We plan to rebuild the EPREL data weekly. Eurostat publishes new prices twice a year; the period of the price in use is shown next to it, for example “Eurostat 2026 H1”.

What we take from EPREL

For each registration we store the model identifier, the supplier or trademark, the EPREL registration number, the energy class, the dates on which the model was placed on and, where applicable, taken off the market, and the label values of its product group: for example yearly energy use, volumes and noise for fridges, or energy per 100 cycles, capacity, water use and spin noise for washing machines. Where the supplier declares them, we also take the guarantee period and the minimum years of spare parts or software updates.

Records we leave out

Registrations that EPREL marks as blocked or not published are skipped. So are registrations without a registration number, model identifier or supplier name.

Personal data

The site does not need personal data, so it is removed before anything is stored: supplier contact details are dropped, and where EPREL lists the supplier as a natural person, such as a sole trader, so are the name in the organisation record and the trademark-owner details.

Variants of the same model

Manufacturers often register the same appliance several times, for instance once per colour or retailer. When registrations of the same brand and category have identical label values, we merge them into one product page that lists every model number. The page is named after the variant published first, and counts as on sale if at least one variant is.

Incomplete records

Each category has a few values without which a fair comparison is impossible, such as yearly consumption and total volume for fridges. A product that lacks one of them, or whose type cannot be determined, still gets a page, but is left out of group statistics, rankings and alternatives.

On sale or not

A model counts as on sale when EPREL has no end-of-market date for it, or that date lies in the future. EPREL does not say whether a particular shop stocks it. We also record which fields of a registration changed between two imports; product pages show when a model was first published, placed on the market and, where applicable, withdrawn.

How a yearly cost is computed

The formula is the same in every category: yearly cost = yearly energy use in kWh × electricity price in € per kWh.

Some labels state a yearly figure directly. Others give energy per cycle, per 100 cycles or per hour, or a power in watts; there we turn the label value into a yearly figure with a usage assumption. These assumptions are ours, not part of the label, and in most categories you can change them.

For TVs and monitors we use the SDR value, as the label’s energy class does; in HDR mode many screens draw considerably more, and product pages show both. For fixed air conditioners, rankings, alternatives and the yearly cost use seasonal cooling only, because every unit cools but not every unit heats: adding heating would make a cooling-only unit look cheaper than a reversible one. Where a unit can heat, product pages show the heating consumption on top, for the average climate and, where declared, for warmer and colder climates.

Gas ovens are labelled with gas energy converted to kWh, to which an electricity price does not apply; they get no yearly cost and no rank. Ovens with several cavities are compared on their first cavity. Charging a phone or tablet costs roughly one euro a year, so there we compare durability instead.

Ten-year figures are ten times the yearly cost at today’s price, without inflation or discounting. The total cost of ownership calculator adds a purchase price you enter.

Electricity prices

Prices come from Eurostat dataset nrg_pc_204 for household consumers using 2,500 to 4,999 kWh a year, including all taxes and levies, in euros per kWh. For each country we take the most recent half-year published. The list contains the EU-27 average, every member state and the other European countries Eurostat covers, such as Norway and Serbia, where a price is published. English pages start with the EU average. All prices are on the electricity prices page.

Your own tariff and usage

The bar at the top of product, ranking and category pages lets you choose another country or enter your own price per kWh, and, where a category has a usage assumption, change the cycles a year, hours a day or uses a year. Costs are then recalculated in your browser. Energy use scales in proportion to usage: 110 washes instead of 220 halve the yearly kWh and cost. These settings are saved in your browser only and are not sent to us; see the privacy notice.

Why we compare only within the same type and size

A large fridge-freezer uses more electricity than a small fridge without a freezer, and a 65-inch TV more than a 24-inch monitor; comparing them would mostly measure size. So every product belongs to a comparison group of type and size band, and every statistic, rank and “better than” figure refers to that group. The type is derived from EPREL values, for example fridges from their compartments and dishwashers from their dimensions.

Each band includes its lower limit and excludes its upper limit: a 7.5 kg washing machine is in the 7–8 kg band, an 8 kg machine in the 8–9 kg band. Models without a declared size form a group of their own.

Rankings and typical values

Within each group, models are ranked by yearly energy use at the default usage, lowest first. A change in usage scales every model by the same factor, so the order stays the same with your own usage. Phones and tablets are ranked by durability score, highest first. The energy class is shown, but the ranking uses the underlying consumption, because two models in the same class can differ noticeably.

On a product page, “better than 80 %” means that 80 % of the models in its group use more energy (for phones and tablets: have a lower score). Models with the same value share a rank. Every model in the group with complete data counts, including models no longer on sale.

A “typical” value is the median of the group: half of the models are below it, half above. Category and ranking pages also show how consumption is spread and the share of each energy class. Ranking pages show only models on sale by default; a switch adds the others. Short lists such as “quietest” show the best five models on sale in each group with at least three such models.

No manufacturer, retailer or anyone else can pay for a position. Rankings come from label values only.

Cheaper-to-run alternatives

Product pages list up to five models that cost less to run. A model is listed only if all of the following apply:

• it is on sale;
• it has the same type and is in the same size band;
• its size is close, within the tolerance in the table below;
• it uses at least 10 % less energy a year;
• its label data is complete.

Alternatives are sorted by yearly energy use, lowest first, with the saving in kWh and in euros at your chosen price. They are selected by consumption alone, not by brand, shop price or any commercial relationship; check features, dimensions and price yourself. Phones, tablets and gas ovens have no yearly cost and therefore no alternatives.

The durability score for phones and tablets

The EU label for phones and tablets states how durable a device is. We combine five of its values into a score from 0 to 100:

• battery endurance per full charge;
• battery lifespan: charge cycles until capacity falls to 80 %;
• number of falls survived without defect in the drop test;
• repairability index;
• minimum years of operating system updates.

For each value we determine the model’s percentile rank among all registered devices of the same type (smartphones and tablets separately): 0 for the lowest value, 100 for the highest; for all five, higher is better. The score is the average of these ranks, rounded to a whole number. A model needs at least four of the five values; if one is missing, the average uses the other four.

The score is our own summary, not an EU rating, and weights the five values equally. If battery life matters most to you, compare that value directly. Drop-resistance class, dust and water protection and a user-replaceable battery are shown on product pages but are not part of the score.

Which pages we offer to search engines

There is a page for almost every registered model, but not every page is useful on its own. We only offer a product page to search engines when it carries enough real data: complete label values, enough comparable models in its group, and links from other pages of the site. The other pages stay online and reachable through search, rankings and links, but are marked so that search engines leave them out. Washer-dryers and range hoods are not offered to search engines yet.

What our figures cannot capture

• Real usage. Label values are measured under standard test conditions. A fridge in a warm kitchen or a TV at full brightness uses more. Our usage assumptions are averages; your household may differ either way.
• Your price and future prices. Eurostat figures are averages for a consumption band and a past half-year. Your contract may be cheaper or dearer, and prices change. Enter your own tariff for a closer figure.
• Purchase price. We do not show shop prices. A model that is cheaper to run can cost more to buy, so compare both.
• Other costs. Water, detergent, gas, repairs and maintenance are not in the yearly cost. Water use is shown separately where the label states it.
• Accuracy of declarations. Label values are declared by the supplier. Market surveillance authorities check samples; we cannot verify individual values.
• Features and quality. Washing or drying results, picture quality and reliability are not ranked here.

Corrections

We reproduce label values as the supplier registered them in EPREL. If a value looks wrong, there are two possible causes:

• The EPREL record is wrong. Only the supplier can correct it; the next rebuild then shows the new value. We do not change values on request while the EPREL record stays the same.
• We made a mistake, for example in reading a field, assigning a type or calculating. We correct the code, and every affected page is recalculated.

Please write to [email protected] with the model number, the EPREL registration number and what is wrong (see the contact page). Changes to this methodology and to our data coverage are listed in the changelog.

Questions