Search Verdict

Reading a spec sheet: the six numbers that predict how a product feels

Most specifications describe a laboratory. A few describe your Tuesday. Here is how to tell them apart, in any category.

3 min read

The short version

  • Peak figures describe the best moment. Sustained figures describe the product.
  • Any number with "up to" in front of it is a ceiling, not an expectation.
  • Ask what the condition was — brightness, volume, temperature, duration.
  • A missing specification is a specification. Manufacturers publish what flatters them.
  • Tolerance and consistency matter more than the headline value in almost every category.

Specification sheets are written by marketing departments working from engineering data, and the gap between those two documents is where most bad purchases live. Nothing on a spec sheet is usually false. Plenty of it is irrelevant, and the relevant parts are frequently absent.

Here are the six questions that separate the two, in any category.

1. Is it a peak or a sustained figure?

A processor’s boost clock is a peak. A monitor’s peak brightness is measured on a small window, not the whole panel. A grinder’s throughput is measured with the hopper full.

Peaks describe the best moment a product has. Sustained figures describe what it does for the length of the task, and they are almost never published — which is why we measure clock speed at thirty minutes rather than at ten seconds, and why the Meridian 14 loses on the specification sheet and wins in use.

2. Under what condition?

“Eleven hours of battery life” is not a claim until you know the screen brightness, the workload, and whether the wireless was on.

Battery figures move by 40% or more between 120 nits and 300 nits. Headphone battery figures move by a third between 65 dB SPL and 80. Every number we publish carries its condition for that reason, and any number that does not carry one is a number measured under whatever condition produced it.

3. What is the tolerance?

This is the question almost nobody asks, and in several categories it matters more than the value itself.

An espresso machine set to 93 C that varies by 2 C between shots produces inconsistent coffee. One that sits at 91 C and varies by 0.4 C produces consistent coffee you can dial in. The second machine has the worse specification and the better behaviour.

The same logic applies to grind distribution, panel uniformity, GPS drift, and shoe sizing. Consistency is the thing you experience; the headline value is the thing you can adjust.

4. Is the missing number the important one?

Manufacturers publish specifications that flatter them. The absence of a figure is therefore information.

No published sustained clock speed means the sustained clock speed is not good. No published retention figure on a grinder means retention is high. No published firmware support window means there is not one. No stated frequency for a noise cancellation figure means it was measured where the product is strongest.

5. Is it measured or derived?

Some figures are measured on the product. Others are calculated from component specifications and never verified.

Rated stabilisation stops, CIPA battery ratings, and “equivalent” measurements of any kind are derived. They are useful for comparing products within one manufacturer’s range and close to useless across manufacturers, because everyone derives differently.

The R7’s CIPA rating is 580 frames. We got 1,100 in normal shooting. Neither number is wrong; they are answers to different questions, and only one of them is your question.

6. Does the specification describe a part or a product?

A monitor with a good panel can have a bad scaler. A laptop with a fast processor can have a cooling system that never lets it run. A chair with a mesh backrest that appears in every premium chair can have a frame that flexes.

Component specifications tell you what is possible. They do not tell you what the manufacturer did with it, and the difference between a $649 monitor and a $1,200 one with the same panel is entirely in that gap.

The short version

Ask for the duration, the condition, and the tolerance. Notice what is missing. Then find someone who measured it.

Filed under

Who tested it

Marcus Oyelaran

Lead reviewer, computing

Nine years of shipping laptops back to manufacturers with the battery report attached. Cares more about sustained clocks than peak ones.

Sofia Almeida

Test lead, kitchen

Pulled espresso professionally for six years before writing about the machines. Owns four thermometers and trusts two of them.

Keep reading

All guides

Computing

Why we stopped recommending laptops on release day

Three of the last five machines we tested shipped with a firmware bug that was fixed within ten weeks. Waiting one quarter is the cheapest upgrade there is.

Computing

The best laptops for everyday work

We ran nine machines under sustained load for thirty minutes each and measured what was left. The winner loses every ten-second benchmark on this page.

Updated
8.9 out of 10 — Strong buy

Computing

Kestrel Meridian 14 review: the laptop that stops apologising in the second hour

It loses the benchmark and wins the afternoon. Eight weeks with the machine that holds 94% of its clocks after half an hour under load.

Updated