Work Sampling

Work Sampling is an Industrial Engineering tool for analyzing work. By analyzing the work through work sampling, we will be able to calculate the amount of work content in terms of percentage of available working time and also it can be used to evaluate the proportions of total time of work devoted to the various activities.

In other words, we will be able to say that a worker was active for x% of his working time and he was idle for (100-x)%. And we can say the proportion of activities of the work (ie 10% of his active time he was tightening the bolt, 15% of his active time he was lubricating, etc.).

By knowing the idle percentage of the operator, it will be the base of improvement, from where we should reduce it. Work sampling data can also used for analyzing the non value adding activities and its percentage of occurrence. It can also be used for analyzing whether manpower allocation is optimum or not.

I recommend you to apply work sampling only if the time study is not possible. Because time study is the most accurate method for analyzing work.

How work sampling is done (short description)

Work sampling analyst select the machine or operator to be studied and he collects all the data regarding the work. Determines the no of observations to be taken and time interval of the observation. Once these two is determined, he prepares the data collection sheet. This sheet contains time at which the observations are to be made and once the time for observation is arrived he records what the operator or machine is doing. And after the study he analyse the data.

So, for conducting a work sampling following is to be done,

1. Selection of machines or operators to be studied
2. Collection of data regarding the work to be studied
3. Calculation of no of observations to be made
4. Calculation of time interval between observations
5. Preparation of data collection sheet
6. Conduct work sampling study
7. Analyze data

Ok, lets go more deep in to this.

1. Selection of machines or operators to be studied

Define the workmen/equipment to be covered for the study clearly, that means which all equipment or machinery to be included or excluded in the study. Concerned approval from the management/ foreman/ section head is to be obtained before the study commences, for getting their support.

We may need to explain the technical aspects of work sampling and why we are doing that. The operator to be studied is also to be made aware of this study, he may have a lot to say to the work sampling analyst and which will be valuable. While selecting a worker he/ she should be the representative of the group and should be familiar with standardized work methods.

2. Collection of data regarding the work to be studied

The analyst must clearly know the different states of the operator’s activity which are to be observed. For example, the selected workmen may operate a diesel generator in the absence of power and may require to be absent from his usual work place for a brief moment. If the analyst is unaware of this arrangement, he/ she may mark the workmen as idle during this time.

So before conducting a work sampling a discussion with operator who does the job is very important, through which the work sampling analyst will get sufficient data regarding the work.

3. Calculation of no of observations to be made

The purpose of determining the no of observations to be taken is to make sure we get the desired accuracy level for our sampling study. This accuracy is determined by two factors, margin of error and confidence level. If we want high accuracy we will fix the margin of error and confidence level according to it, and calculate no of observations needed for that accuracy and vice versa. And if we need high accuracy results no of observations to be taken will be higher.

Before going to this subject we must know the following statistical terms related to sampling.

1. Sampling
2. Margin of Error
3. Standard error
4. Z score or critical value
5. Confidence level
6. Confidence interval

1. Sampling

Sampling is a process of randomly selecting or collecting data from a large population, for analyzing something related to the large population. The no of data collected will be comparatively very low to the population, which is the importance of sampling. A short sample taken from a large population gives an accurate depiction of the distribution of the population.

Eg: For knowing the working percentage of an operator during his working time, we make observations randomly and calculates the percentage of working (p(hat)) of the operator from the sample observations. This percentage of working of the operator will be approximately equal to that of actual working percentage (p) of the operator during his working time, if the sampling study is done statistically.

2. Margin of Error

Margin of error is calculated by the formula,

Margin of error = Z * Standard error

Z = Z score or critical value

Don’t worry you will be able to understand these terms by continuing reading this article.

3. Standard error

In our previous example, if we repeat sampling experiment, and we plot the p(hat) of each sampling study in x axis and its frequency of occurrence in y axis, we will get a graph as shown below.

This is called a normal distribution curve or a bell curve. The center value of the graph will be average of p(hat), which will be approximately equal to the actual percentage (p) of working of the operator during his working time, if sufficient no of experiments are done.

And now the spread of the graph around this value p, will depends on the standard error (σ_p) of the sampling proportion p(hat). Here is how the spread of p(hat) depends on standard error, 68.27% of the data will be accommodated within the one standard error from the either side of the p, 95.45% of the data will be accommodated within the two standard error from the either side of the p, 99.73% of the data will be accommodated within the three standard error from the either side of the p, 99.9937% of the data will be accommodated within the four standard error from the either side of the p, 99.99943% of the data will be accommodated within the five standard error from the either side of the p and 99.9999998% of the data will be accommodated within the six standard error from the either side of the p.

Formula for standard error is,

Where n is the no of observations.

If we are using p instead of p(hat) in above formula we will get the standard deviation.

4. Z score or critical value

It is the no of standard error from the value p to its either side, for a particular confidence level.

5. Confidence level

Suppose the confidence level of sampling study is 95%. That means if repeat the sampling study over and over again with same n no of observations, the probability of p coming in between the confidence interval is 95% of no of study and for 5% of no of study it will be outside the confidence interval.

6. Confidence interval

Suppose 0.10 is the margin of error (it can also be written as 10%) and p(hat) is 55% , 45% and 65% is the confidence interval, ie plus or minus 10% to p (hat).

Calculation of required no of observations for work sampling

Before starting calculating no of observation, we need an estimate value of p(hat), for that we initially conduct a pilot sampling study with n taken as we feel comfortable and also make sure to collect sufficient no of samples according to the population. In our case one shift can be considered as a population. And suppose we got 45% as p(hat) in our pilot sampling study.

If you are not confident about pilot study take estimate p(hat) as 0.5, which will give the maximum no of observation to be done for a particular margin of error and confidence level.

Now we can calculate the no of observations required for margin of error 10% and confidence level as 95%.

we apply this to the formula of margin of error

Margin of error = Z * Standard error

Here margin of error we taken is 10%, ie 0.1

Z=1.96 (If confidence level is 95%, Z score is 1.96. This can be obtained from any confidence level z score conversion table)

By calculating this, we will get the value for n as 95.07. So, 96 is the no of observations we should take for getting desired level of accuracy.

So, if we take sampling study with 96 no of samples as we calculated, we can say that, if we repeat this study over and over again, the actual working percentage of operator will be between 10%, to the either side of the p(hat) of a particular study, for 95% of no of the studies.

I am sorry, you may feel hard to understand this concept, that is statistics. I have tried my level best to share this knowledge in simple words. Most of them will understand the concept by reading the above paragraph. If you found any error in this article please give feedback as comment to this article. OK, now we can go to the next section.

Please note that the no of observations what we calculated here is the minimum no of observations to be made for getting desired accuracy level. There is no problem if we increase the no of observation beyond the calculated one, if the cost permits.

4. Calculation of time interval between observations

Ok, now we got the no of observations that to be taken in our sampling study. Next thing we want to know is the time interval between the observation. Normally we can call one shift as a population, since all shifts are repetitive, but output may vary. So, one sampling study to be completed in one shift.

We will came up with the following two situation while taking observations;

1. Analyst staying at one work station and making observations
2. Analyst making observations for more than one work station which are sufficient distance apart

Analyst observing one work station

If an analyst is staying at one work station, the minimum time interval between observations can be shortest as possible which ever is comfortable by the observer, say 3 minutes.

Analyst observing more than one work station

Instead of one operator suppose we may want to study the operators of a big factory or area. In that case the time study analyst will take sufficient time to move between the operators for completing one round. Suppose it takes 10 minutes to make one round for an analyst. Add sufficient amount of time to the analyst for taking rest to this time and in this case suppose rest time is 3 minutes. So, minimum time interval between each round should be 10+3 = 13 minutes.

In the above two cases, I do not mean that every time interval should be a constant, that also should vary to some extent, then only the sampling time will be truly random. In our case, let the time interval vary between minimum three and maximum of 6 minutes, for the case analyst observing one work station. Otherwise, if there is an activity repeating every three minutes in the work, the sampling study will give wrong information.

Now we have all data to list down observation timings

5. Preparation of data collection sheet

This is the final step before we conduct the work sampling study.

Design a chart for recording following information,

1. Predetermined times at which work sampling observations to be made
2. Remarks column for noting any information regarding each observation
3. Other relevant information such as Shift, Operator name, Output of shift, Date, etc.

First point I will explain, other points I think there is no need of further explanation.

Determination of time at which observations are to be made for work sampling

This is the step where we need to give importance to the word random. Predetermined times for taking work sampling observations should be random and time interval between observation time should vary, with minimum time interval as we calculated. Generate times at which observations to be made by keeping in mind these two points.

For example, in our work sampling study, the shift starts from 8:00 am and ends at 4:00 pm. And the analyst is observing only one operator. So, the observation times are as follows 08:03 am, 08:08 am, 08:12 am, 08:15 am, and so on. As you can see there is no repetitive pattern in these timings, but it is sticking to the time interval we have discussed earlier.

I have created a tool for generating random times for work sampling. This tool can be used free of cost by visiting this page.

After determining the times, make sure the no of observations is more than the required no of observation. If it is less than the minimum the no of observation we calculated, do the sampling study for the next shift also. Please note that do not stop sampling study at half or 3/4th of the shift even if the required no of observation is reached. Because activities of worker may not be the repetitive throughout the shift. I recommend you to do 2 or 3 sampling study for getting good results, if no of shifts required for work sampling study we calculated is one.

6. Conduct work sampling study

Now with the prepared data collection sheet and all preparations we have done, make the observations about the work at the specified times. And mark what the operator or equipment is doing during the observation. Also don’t forget to write notes about each observation wherever necessary, this will be useful in analyzing the sampling data.

While selecting the shifts for work sampling, make sure that those shifts are representing whole shifts considering material availability, output, working condition, etc. If the output of the shift is varying try to select the shifts which produces output which meets the normal customer demand.

7. Analyze data of work sampling

List down all activities observed during the observation and its percentage of occurrence. This will be explained in the following paragraphs.

Suppose during our study, the following activities and its no of occurrence were noted,

Following table is the percentage of occurrence of an activity, its calculated by no of occurrence of an activity / total no of observations.

From these data we will be able to calculate percentage of idle and working

Percentage of working p(hat) = 75/96 = 78.12 %

Percentage of idle (1-p(hat)) = 21.88 %

Another application of work sampling (Calculation of optimum manpower required)

In another example suppose there are 10 operators allocated to look after a processing plant. All the operators are assigned to somewhat similar tasks and there is no distinction between any of them.

After work sampling we find that the average standard utilization of the operators in the plant was found to be 70%. Therefore the optimum manpower required for the operation of the processing plant will be (70/100 x 10) = 7 Operators, for same output of plant, which was at the time of study.

Very important : Wherever we publish or use the results of the analysis don’t forget to mention the output of the shift during the study. For example, lathe operator was active 78.12 % of time and was idle 21.88 % of time while he produces 17 products.

General points about work sampling

1. In case of time study the accuracy level is high as the operation is divided into fine activities/ elements and a detailed picture is arrived at, whereas work sampling fails to give a detailed analysis.
2. The management and workers may fail to understand the basics of work sampling as they do with time study, because some more statistics is involved in work sampling than time study.
3. I recommend you to do two to three sampling study and consider all these data together in analyzing the data.

Thank you for reading. I hope you got sufficient information regarding the work sampling. If you have any thoughts regarding this article you can share it in comment section.

I am Thankful to Mr. Vivek N Unni for contributing to this article.

Know more about our time study consulting service.

If you know any subject that can be related to manufacturing industry or industrial engineering, you can earn some income by becoming article contributor of this website. For knowing more about it, please visit Join us page.

You don’t need to have any experience in article writing, just knowledge on the subject is needed.

Also you can know more about our team of article contributors by visiting the about us page.