How we rate scales

A coffee scale exists to read 0.1g. Anything coarser cannot dose espresso, which is the whole reason to buy one rather than use a kitchen scale.

What we can actually read: Resolution and maximum capacity are published by everyone, and the two pull against each other: high capacity usually costs resolution.

Scored directly from published figures.

The 2 rules below score this criterion out of 10, and they add up to 10 exactly. That result is then worth 26 of the category's 100 points.

• Resolution6 of 10

  0.1 g or less

  6 pts

  over 0.1 g, up to 0.5 g

  4 pts

  over 0.5 g, up to 1 g

  2 pts

  over 1 g

  1 pt

• Max capacity4 of 10

  500 g or less

  2 pts

  over 500 g, up to 2000 g

  3.5 pts

  over 2000 g

  4 pts

10 of 10

The lag between pouring and the number moving. On espresso, a scale half a second behind is a scale you are guessing with.

What we can actually read: Almost nobody publishes a response time. One scale on this site states one. Where it is missing the criterion says so rather than inventing a figure.

Scored directly from published figures.

Every rule below is struck in this category, because nobody publishes the inputs. Nothing here is scored from a specification. That result is then worth 20 of the category's 100 points.

• Response timenot scored here

  Fewer than one maker in four publishes this here, so its 10 points leave the sum rather than being guessed at.

  0.5 s or less

  10 pts

  over 0.5 s, up to 1 s

  8 pts

  over 1 s, up to 2 s

  5 pts

  over 2 s, up to 3 s

  3 pts

  over 3 s

  1 pt

10 of 10

Scales live under a portafilter in a wet environment. Water resistance is the difference between a three-year scale and a six-month one.

What we can actually read: Water resistance rating, surface material and warranty are published, though waterproof is claimed far more often than an IP rating is given, and a claim without a rating is treated as a claim.

Scored directly from published figures.

1 of the 3 rules below are struck through in this category, because too few makers publish the input. The 2 that remain score this criterion out of 10 between them: a struck rule leaves the sum entirely rather than counting as a zero. That result is then worth 18 of the category's 100 points.

• Water resistancenot scored here

  Fewer than one maker in four publishes this here, so its 4 points leave the sum rather than being guessed at.

  ipx7, ip67

  4 pts

  ipx6

  3.5 pts

  ipx5

  3 pts

  ipx4

  2.5 pts

  ipx3

  2 pts

  ipx1

  1 pt

  none

  0.5 pts

• Surface3 of 10

  stainless steel, glass

  3 pts

  silicone, aluminium

  2.5 pts

  plastic, abs

  1.5 pts

• Warranty3 of 10

  none

  0 pts

  up to 1 yr

  1.2 pts

  over 1 yr, up to 2 yr

  2.2 pts

  over 2 yr

  3 pts

10 of 10

The things that turn a scale into a brewing instrument: a timer that starts itself, a flow rate readout, brewing modes.

What we can actually read: All published. Flow rate and auto timer start are the two that separate a coffee scale from a kitchen one.

Scored directly from published figures.

The 6 rules below score this criterion out of 10, and they add up to 10 exactly. That result is then worth 14 of the category's 100 points.

• Timer2 of 10

  present

  2 pts

• Auto timer start2 of 10

  present

  2 pts

• Flow rate display2 of 10

  present

  2 pts

• Brewing modes2 of 10

  4 or more

  2 pts

• Auto tare1 of 10

  present

  1 pt

• App connectivity1 of 10

  present

  1 pt

10 of 10

Reading it in a dim kitchen and keeping it charged.

What we can actually read: Display type, backlight, power source and battery life are published, though battery life is claimed more often than measured.

Scored from figures that predict it, because the thing itself is not published.

3 of the 5 rules below are struck through in this category, because too few makers publish the input. The 2 that remain score this criterion out of 10 between them: a struck rule leaves the sum entirely rather than counting as a zero. That result is then worth 12 of the category's 100 points.

• Display3 of 10

  oled, e-ink

  3 pts

  lcd, led

  2.5 pts

• Backlitnot scored here

  Fewer than one maker in four publishes this here, so its 2 points leave the sum rather than being guessed at.

  present

  2 pts

• Power2.5 of 10

  rechargeable, usb

  2.5 pts

  aaa, aa, battery

  1.5 pts

• Battery lifenot scored here

  Fewer than one maker in four publishes this here, so its 1.5 points leave the sum rather than being guessed at.

  20 hr or less

  0.5 pts

  over 20 hr, up to 60 hr

  1 pt

  over 60 hr

  1.5 pts

• USB-Cnot scored here

  Fewer than one maker in four publishes this here, so its 1 points leave the sum rather than being guessed at.

  present

  1 pt

10 of 10

Points per pound against the category median, which is unusually tight here: $10 to $99 for what is often the same sensor.

What we can actually read: Needs a published price here and on at least three rivals.

Computed from price, not from a rule list

Value is the only criterion here with no thresholds of its own, because it cannot be read off one product. It asks what else the same money buys in this category, so it needs the category. That result is then worth 10 of the category's 100 points.

Points already earned on everything else

out of 10

Multiplied by the category median over this price

median ÷ price

Capped at

10

No price published, or fewer than 3 priced rivals

5 of 10

Half marks is not a verdict, it is a refusal to guess: with nothing to compare against, a product is neither good value nor bad, and scoring it either way would invent a finding. It corrects itself as the category fills.