Counts can be exact, if the situation is perfect, but this is rarely the case. There are three main categories for counting scale error. Now let’s discuss each category and how we deal with those errors:

**1. Operator Error**

There are several categories of operator error. Some are obvious such as forgetting to “tare” for a container or miscounting the sample, or sampling one part to get an APW and then placing a different type of part on the scale.

Some are less obvious such as allowing a draft to blow on the scale, not leveling the scale thus allowing the scale to teeter, leaning on a rickety table while using the scale, allowing some part of the scale to touch a wall or a power cord. Every spring we get calls from customers that their counting scale is inaccurate and needs service.

Sometimes there is a problem with the scale; other times someone opened a window near the scale and that made the scale unstable. We correct these problems by training the operator and the department heads how to recognize these problems and correct them.

Good scale design helps also. The better scales from companies like Setra and GSE can be programmed to prompt the operator what to do next, so he/she doesn’t forget to tare a container or enter a new ID. Many have settings for stability so that the scale will report that the conditions are not stable enough to compute an accurate APW, for example.

**2. Piece Weight Error**

Many people think that because their pieces all look the same and perform the same, that they weigh the same. Even test weights that are designed to weigh the same don’t weigh exactly the same-there is a tolerance on them too! If you use a one piece sample and that one piece is 3% heavier than the average piece in the population you take it from, you will get a 3% counting error all else being perfect. So you need to take a sample that is large enough that the sample to sample variations tend to cancel.

Actually, tests that Setra systems has conducted in conjunction with APICS concluded that in the average stockroom you need a 50 piece sample to be reasonably certain that your counts are within 1% of the true count. Many people have the erroneous idea that sample to sample variation depends upon the weight of the piece-that somehow smaller pieces are less uniform than larger ones. These people are misdiagnosing errors attributable to our third category below.

We diminish these problems by providing scales that can be programmed to force the operator to use a certain minimum sample size. So if the scale is programmed to require 20 pieces and the operator uses 15, the scale will force the operator to add 5 additional. This feature is found on medium and upper end scales, but should be on all of them. Even if the department head understands the importance of this, there is no way to be sure that the operators are using a sufficient sample, unless the scale can be programmed to force him/her to do so.

**3. Scale Related Error**

Every scale ever made has a finite resolution. This means that at some point, as the weight gets smaller and smaller, the scale simply cannot report weight accurately. Here is an example to bring home this point: We have a scale that computes average piece weight with a resolution of 1 g. and we have a 10 piece sample of parts. Each part weighs exactly 1.06 g.

Let’s assume that the pieces are absolutely uniform and that the scale has flawless weighing accuracy. The scale should weigh those pieces at 1.06 g. * 10 = 10.6 g., but because the scale has 1 g. resolution it will settle on 11 g. If it settles on 11 g., then the error would be (11-10.6)/10.6 = .0377, or 3.7%. That’s with a perfect scale and perfectly uniform pieces and each piece being heavier than the resolution of the scale! In many cases an error close to 4% is just too high.

We eliminate this problem by providing a scale that has a settable accuracy parameter. Thus if we cannot tolerate scale errors over 2%, we set the scale at 98% and it will prompt the operator for more pieces if it cannot achieve 98% accuracy with the sample it has. Many scales offer this feature, but show the error/accuracy to the operator and let him/her determine what, if anything to do about.

Most department managers would rather make the decisions themselves and have the scale enforce it for them, than to have the operator make the final decision. After all, the operator may have been told to always use at least a 25 piece sample and a 98% accuracy, but once he/she finds out that the scale will work with 10 pieces and any accuracy level, or when he is on vacation and someone else is doing his job, these instructions may not be followed.

Another way to help diminish this error is by using a more sensitive scale so that the operator will not be prompted to add lots of pieces. Often to gain sensitivity we give up capacity. This leads to having multiple weighing platforms tied into a common controller, as shown in the picture below. The operator can sample on a low capacity very sensitive scale, then press a button and count on a larger capacity unit after the APW is accurately computed on the sensitive scale.

In this picture, Rob samples tiny pieces on the analytical balance, but if he has more 0.4 lb. of them, he can switch to the 6 lb. scale or the 50 lb. scale retaining the accurate APW determined on the analytical balance. This system utilizes a GSE 675 counting scale, an analytical balance, a remote platform, a keyboard, a printer, a scanner. They all are seamlessly integrated. See our counting systems section for more about this application.