This article is not meant to be an authority on statistics, but simply an explanation of why we recommend buying an "extra decimal" whenever possible. Rounding errors are a seldom discussed, but ever-present factor in the scale industry, and something the consumer should be made more aware of.

All scales have a displayed resolution, whether 1 lb or 0.0001g. The displayed resolution is how "accurate" the scale is. It is essentially a compromise between reality and budget. Let me explain.

I will use human body weight as an example. Let's assume that I weigh 175 lbs... In reality, let's assume that technically speaking, I weigh 174.937678540283 lbs.

To illustrate how a scale (in general) would display how much I weigh, here is a list illustrating how scales decide between competing resolutions, all of which are "accurate" based on the scale's displayed resolution.

2 lb resolution: 174 lbs (closer to 174 than it is 176)
1 lb resolution: 175 (closer to 175 than 174)
0.1 lb resolution: 174.9 (closer to 174.9 than 174.8 or 175.0)
0.05 lb resolution: 174.95 (closer to .95 than .90)
0.02 lb resolution: 174.94 (closer to .94 than .92)
0.001 lb resolution:174.938 (closer to .938 than .937) *** This is the real illustration.

As you can see, (in general) a scale will display a value based on the REAL weight always being something other (more specific) than a resolution...

SO, if your application is something that requires the resolution of 0.001g (for example) in any situation, a perfectly calibrated balance will still display at worst a rounding error of 0.0004g - if in reality the sample weighs 112.3456g - the balance will display the resolution 0f 112.346g- a difference of 0.0004g

This 0.0004g does not appear on the scale, and will likely only influence your results when doing formula work, compounding, etc.

If your application requires accuracy to a certain spec, and budget allows, we recommend buying an extra decimal place.... in practice, this will allow you to know when you are at the desired volume of additives - without worrying about rounding errors in the single digit %.

Austin H.
General Manager

Author: Summit Measurement