Accuracy by employee: measuring without blaming

A warehouse can show 98 percent record accuracy overall, yet that figure says nothing about how the errors are spread. More often than not, most of them come from two or three people, one shift or one operation. The point of measuring by employee is not to punish but to see where training or a process change is needed. Punish the person who finds an error and next time errors will be hidden, and the accuracy figure loses its meaning.

What to count

For each employee take not the number of errors but the error rate: lines with an error ÷ all lines the person handled. A line is one item position in a document. The data comes from your system: who entered the document, which operation it was (receiving, picking, dispatch) and whether a later check, such as a count or a customer complaint, found a discrepancy.

For example, one month in receiving:

• Employee A: 1 200 lines, 18 errors, or 1.5 percent.
• Employee B: 900 lines, 9 errors, 1.0 percent.
• Employee C: 300 lines, 12 errors, 4.0 percent.
• Employee D: 600 lines, 21 errors, 3.5 percent.

Altogether that is 3 000 lines and 60 errors, an average of 2.0 percent. By the number of errors, D (21) and A (18) look the worst and C (12) looks middling. By rate, A is better than average and C is the worst. The same data gives two different conclusions, which is why you should work with the rate alone.

Alongside the rate, record the value of the errors in so'm. A rate by lines counts a slip on expensive electronics and one on cheap stationery as the same. Two measures together, rate and value, show more precisely whom and which operation to focus on first.

Fair comparison

Compare only people who work under the same conditions:

• type of operation: do not compare receiving, picking and counting with one another;
• range: someone handling look-alike goods that differ only in size and colour will slip more often than someone handling uniform packs;
• shift and load: errors rise on the night shift or in the pre-holiday rush;
• length of service: count a new employee's first four weeks separately.

Sample size matters too. The 300 lines of employee C are too few for a firm verdict. At the average of 2.0 percent, 300 lines would be expected to give 6 errors, and he made 12: twice as many, but one month of data is not enough to pass judgement. Collect at least two or three months and several hundred lines per person, then watch a rolling window of the last three months.

Separate the types of error

The same 3.5 percent can have different causes. Split the errors by type: wrong quantity, a different item entered (a mix-up), a document entered late, the wrong warehouse chosen. Suppose 15 of the 21 errors of employee D are documents entered late. That is not carelessness: most likely there is no terminal at the receiving point, and data is first written on paper and then copied into the computer. So the problem lies not with the person but with the process. If the errors of employee C are mostly entries of the wrong item, he needs practice in telling similar article codes apart and a layout that puts them farther from each other on the racks.

How to use the result

In the first month just measure and draw no conclusions. From the second month show the team the overall result and the most frequent type of error, without names. Keep one-to-one conversations for a figure that stays high for several months: ask about the cause first, then offer training or a process change. Use the working methods of those with good results to teach everyone else.

Do not tie bonuses to accuracy alone. People may reduce not errors but the number of reports about them, so choose control counts at random and compare the records with what is on the shelf yourself.

Conclusion

Overall accuracy does not tell you where errors cluster, while a breakdown by employee and operation does. Measure the rate, under equal conditions, over a sufficient period and without hunting for culprits. Through cycle counting and discrepancy analysis, CountLab breaks errors down by operation, zone and shift, and the management report helps you decide what to change: the training or the process.