WORK SAMPLING

Work Sampling Definition

z Work sampling is a technique used to investigate the
proportion of time spent on various activities in a workplace.

z The results of work sampling are effective

z for determining machine and personnel utilization
utilization,
z allowances applicable to the job,
z and production standards.

z Take a large number of observations at random intervals.

z Example:
z 1000 observations over 7 weeks
z In 700 observations machine is running
z In 300 observations machine is idle or not running,
z Machine works 70% of time, idle 30%.
z The accuracy of the data determined by work
sampling
p g depends
p on
z the number of observations
z and the period over which the random observations
are taken.

z Unless the sample size is sufficiently high, and

th sampling
the li period
i d represents
t ttypical
i l
conditions, inaccurate results may occur.
Ad
Advantages over Time
Ti Studies
S di

z Does not require continuous observation by an

anal st o
analyst over
er a long period of time.
time
z Clerical time is diminished.
z Total work hours expended by the analyst are usually
much fewer.
z Operator is not subject to long-period stopwatch
observation.
z Crew operations can be studied by a single analyst.
z Work sampling is a method of finding the
percentage occurrence of a certain activity by
statistical sampling and random observations.

z Work sampling, is the process of making sufficient

random observations of an operator’s activities to
determine the relative amount of time the operator
spends on the various activities associated with the
job.
z The major goal of work sampling is to determine
h
how llong, or h
how muchh off th
the work
kdday, iis spentt
on specific types of work.

z Work sampling
p g may y identify
y the fact that certain
operators spend a large portion of their time
waiting
g for work,, or p
performing gppaperwork
p tasks,,
or even performing activities that are not
included in their jjob descriptions.
p
z One off the
O th basic
b i foundations
f d ti off statistical
t ti ti l sampling
li
theory is the concept that the larger the sample size, the
results will be better or more accurate
accurate.

z In work sampling
sampling, a sufficient number of observations
must be made to be sure that the results accurately
summarize the work p performed. There are statistical
formulas to help determine how many observations
should be made.

z The number of observations that an analyst must make

off a particular
ti l jobj b also
l d depends
d on h
how much
h titime iis
devoted to a particular task.
The Theory of Work Sampling
z The theory of work sampling is based on the
fundamental law of probability.
z To make things
g easier,, it is more convenient to speak
p of a 95 per
p cent
confidence level than of a 95.45 per cent confidence level.
z To achieve this we can change our calculations and obtain:
z 95 per cent confidence level or 95 per cent of the area under the curve = 1.96 σp
z 99 per cent confidence level or 99 per cent of the area under the curve = 2.58 σp
z 99.9 per cent confidence level or 99.9 per cent of the area under the curve = 3.3 σp

z In this case we can say that if we take a large sample at random we can be
confident that in 95 percent of the cases our observations will fall within ±
1.96 σp
Work Sampling
p g
z Probability of such binary event happening m
outt off n times,
ti is
i determined
d t i db by th
the binomial
bi i l
distribution:
P(m) = n! pm (1-p)m-n
m! (n-m)!
( )
n = # obs, m = # events, p = prob of event, q=1-p
Prob of 0,1,2,3,4 heads in 4 tosses
n! pm (1-p)n-m / m! (n-m)!

# heads # tails Binomial expansion P(x)

0 4 0.0625

1 3 0.250

2 2 0.375

3 1 0.250

4 0 0 0625
0.0625
 Normal Approximation
#H #T P(x)
( )
0 4 0.0625
1 3 0 250
0.250
2 2 0.375
3 1 0 250
0.250
4 0 0.0625

As n → ∞,
∞ binomial → normal
normal,
mean = p, std. dev. s = √pq/n
z Determination of sample size

z As well as defining
g the confidence level for our
observations we have to decide on the margin of
error that we can allow for these observations.

z Let us look at our example about the productive

time and the idle time of the machines in a
factory.
factory
zStatistical method. The formula used in this method is:
 Work Sampling Acceptance
z The analyst must sell its use and reliability to all members of
th organization
the i ti whoh will ill b
be affected
ff t d bby th
the results.
lt

z This can be done by conducting several short sessions with

representation of the various interested parties.

z Both unions and workers favor work sampling techniques,

once the procedure is fully explained, since work sampling is
completely
l l iimpersonal,l ddoes not utilize
ili a stopwatch,
h and
d iis
based on accepted mathematical and statistical methods.

z The principle of work sampling: adequate sample size to

ensure statistical significance.
g
 Work Sampling Study Plans
z Detailed planning must be done before actual work
sampling study observations are made.

z Once the preliminary estimates have been made, the

analyst can determine the desired accuracy of the results
results.
This can best be expressed as a tolerance, or limit of
error, within a stated confidence level.

z Next, the analyst must estimate the number of

Next
observations to be made and determine the frequency of
observations.
observations
CONDUCTING A STUDY

z It is recommended that a uniform procedure should be followed to

perform a work sampling study is to

1. Establish the Purpose

z First, the objective of the study should be established. Work
sampling can be used to determine an overall perspective on the
work done.

2 Identify the Subjects

2.
z Second, the people performing the task must be identified, i.e.
general office work is being studied with the objective of determining
overall productivity.
p y

3. Identify the Measure of Output

z The third step in making the study is the identification of the
measure of the output produced or the types of activities performed
on the jobs being studied. This step is especially important if the
objective of the study is to measure productivity with the intent of
setting
tti a standard.
t d d
4. Establish a Time Period
z Fourth, the time period during which the study will be
conducted must be established.
established Starting and stopping
points for the study must be defined as well.

5. Define the Activities

z This step involves defining the activities that are
performed by the people under study
study. For example,
example the
definition used in a machine utilization study, including
only the categories of working, idle, and idle-mechanical
breakdown.
breakdown
6. Determine the Number of Observations Needed

z After the work elements are defined, the number of observations for the
desired accuracy at the desired confidence level must be determined.
The sample size is dependent on the percentage of time believed to be
spent on the major work element
element.
z If a reasonable guess cannot be made, then a trial study of perhaps 20
to 40 observations should be made to get an estimate of this portion.
ese initial
These t a obse
observations
at o s sshould
ou d be included
c uded with
t tthe
e rest
est o
of tthe
e
observations taken during the rest of the work study.

7. Schedule the Observations

z Once the number of required observations has been determined, either

pp p
from appropriate statistical calculations or from tables,, the actual
observations must be scheduled. Typically, the analyst will assign an
equal number of observations each day during the course of the study.
z For example, if 800 observations are required and 20 work days are
established
t bli h d as an appropriate
i t observation
b ti ti time, 40 observations
b ti should
h ld
be recorded each day.
z A random number table can be used to establish the random times for
each observation
observation.
8. Inform the Personnel Involved
z Before
B f the
h study
d iis actually
ll performed,
f d the
h personnell iinvolved
l d should
h ld bbe
informed about the objective of the study and the methodology that will be
employed. As in any productivity measurement study, this part of the
procedure is very important.
z Workers and their supervisors might think that they personally are being
measured rather than the work they are doing.
9 R
9. Record
d th
the R
Raw D
Data
t

z The next and perhaps the easiest part of any work sampling study is the
actual recording of the raw data
data. Although this recording can be
performed by anyone, it is desirable that a trained analyst be employed.
z It is also very important that the observations be made at exactly the
same location everyy time. Failure to be reliable in this manner mayy bias
the results.
10. Summarize the Data

z After the data have been collected, they must be summarized.

Determining the observations
needed
z To determine the number of observations
needed, the analyst must know the desired
accuracy of the results.

z The more observations, the more valid the

final answer.
 # Observations Needed
z 95% fall in interval p ± ℓ (limit of error)
z Where ℓ =1.96s = 1.96 √pq/n
z ℓ2 = 3.84 pq/n → n = 3.84pq/ℓ2
z Ex 1: p=0.08. How many observations needed for 2%
error?

z Ex 2: Only collect n=500. What is error limitation?

z There is a direct trade-off between the error or

accuracy off the
th study
t d and d the
th number
b off observations
b ti
collected.
Problem - Correlated Observations
z If several workers observed simultaneously
z Individual readings are not independent
z Need to correct standard deviation (s) and
z interval p + ℓ or p +1.96s
z s2 = ∑y(j)2/n(j) – np2
n(m-1)
z Where: n = total
Wh t t l##obs,
b m = # grouped d obs
b
z Where: n(j) = #workers at jth obs,
z y(j) = #workers “idle” at jth observation
 Correlated Observations – Ex #3
Round #workers nj #idle, yj yj2/nj
1 10 6 36
3.6
2 10 2 0.4
3 10 3 09
0.9
4 10 1 0.1
5 10 2 04
0.4
Total 50 14 5.4

s2 = [∑yj2/nj–np2]/n(m-1) =
s=0 0.086
086 and p=14/50 = 0 0.28
28
95% confident that interval:
28 ± 1.96(8.6)
1 96(8 6) or 28 ± 16.4%
16 4% ((correct)
t)
Determining Observation
Frequency
z The frequency of the observation depends
depends, for the
most part,
ƒ on the
th number
b off observations
b ti required
i d
ƒ and the time available to develop the data.

z The number of analyst available and the nature of

the work being studied also influence the frequency
of the observations.
z After determining the number of observations
per day, the analyst must select the actual
time needed to record the observations.

z To obtain a representative sample, observations

are taken randomly at all times of the day.

z The longer
Th l the
th overallll study,
t d the
th better
b tt ththe
chance of observing average conditions.
Making random observations

z To ensure that our observations are in fact made at random, we can use a
random table such as the one in the table in the following page.

z Various types of random table exist, and these can be used in different
ways. In our case let us assume that we shall carry out our observations
d i ad
during day shift
hift off eight
i ht h
hours, ffrom 7 a.m. to
t 3 p.m. An
A eight-hour
i ht h d
day h
has
480 minutes. These may be divided into 48 ten-minute periods.

z We can start
W t t by
b choosing
h i any number b att random
d ffrom our table,
t bl ffor example
l
by closing our eyes and placing a pencil point somewhere on the table. Let
us assume that in this case we pick, by simple chance, the number 11 which
is in the second block
block, fourth column
column, fourth row (table 12).
12)

z We now choose any number between 1 and 10. Assume that we choose the
number 2; we now go down the column picking out every second reading
and noting it down, as shown below (if we had chosen the number 3, we
should pick out every third figure, and so on).

z 11 38 45 87 68 20 11 26 49 05
z Looking at these numbers, we find that we have to discard
87 68 and 49 because they are too high (since we have
87,
only 48 ten-minute periods, any number above 48 has to be
discarded).

z Similarly, the second 11 will also have to be discarded

since it is a number that has already
y been ppicked out. We
therefore have to continue with our readings to replace the
four numbers we have discarded. Using the same method,
that is choosingg every
y second number after the last one
(05), we now have 14 15 47 22

z These four numbers are within the desired range and have
not appeared before. Our final selection may now be
arranged numerically and the times of observation
th
throughout
h t the
th eight-hour
i ht h day
d worked k d out.
t Thus
Th our
smallest number (05) represents the fifth ten-minute period
after the work began at 7 a.m. Thus our first observation
will
ill be
b att 7.50
7 50 a.m., and
d so on (table
(t bl 13).
13)
Frequency and Timing of Observations

z Use random number table

z 4 digits: #1=day, #2 =hr, #3-4 = min.
z Di
Discardd outside
t id range
z Arrange in numerical order
z Or use DesignTools!
Designing the Work Sampling
Form
z A standard form is usually not acceptable,
since each work sampling study is unique
from the standpoint of the
z total observations needed
needed,
z the random times that observations are made,
z andd th
the iinformation
f ti b being
i sought.
ht
Recording Form - More
Using Control Charts
z The control chart techniques used in
statistical quality control work can readily be
applied to work sampling studies.
z In quality control work
work, the control chart
indicates whether or not the process is in
control.
t l
z In a similar manner, in work sampling, the
analyst considers points beyond -/+ 3 sigma
limits of p as beingg out of control.
Control charts
z First, choose limits: +/- 3 sigma
z Create control chart with upper and lower
control limits
z G h daily
Graph d il ((or periodic)
i di ) means
z Our ppurpose
p of work sampling
p g is to determine
work areas that might be improved.
z Control charts can show the progressive
improvement of work areas.
Observations and Data
Recording
z The analyst should walk to a point a given distance
from the facility, make observations, and record the
facts.

z If the operator or machine being studied is idle, the

analyst
y must determine the reason for the idleness,
confirming the reason with the line supervisor before
entering the data on the form. The analyst must
learn to take visual observations, and then make the
written entries after leaving the work area.
Applications of Work Sampling
z Calculation of Standard Time
z Finding
g Worker or Machine Utilization
z Application to Service Industry
z S lf Ob
Self Observation
ti
Machine Utilization
z To determine machine utilization in the same
manner as that used in establishing
allowances.
Allowances Determination
z The determination of time allowances must be
correct, if fair standards are to be developed.

z Prior to the introduction of work sampling, analysts

frequently determined allowances for personal
reasons and unavoidable delays by taking a
series of all-day studies on several operations and
then averaging the results.
z Through a work sampling study, analysts take a
great number of observations at different times
g
of the day and of different operators.
z They can then divide the total number of
legitimate non-work occurrences that involve
normal operators by the total number of working
observations. The result equals the percentage
allowance that should be given to the operator
for the class of work being studied.
 Standard Time Determination
z Work sampling can be very useful for
establishing time standards on both direct and
indirect labor operations. The technique is the
same as that used for determining allowances
allowances.
z The analyst must take a large number of
random
d observations.
b ti
 Calculating a Standard Time
for a Single Operation(Press operator for 8-hr shift)

Information Source Data

Working + idle time Time cards 480 min
# parts produced (#P) Inspection 420 units
records
Working time fraction (ni/n) Work sample 0.85
0 85
Average rating Work sample 110%
% Allowances Determined 15%

Observed time (OT) = Tx(ni/n) / P

Normal time (NT) = OTx R/100
St d d titime (ST) = NT(1+Allow)
Standard NT(1+All )
T: Total time, P total production time for the period studied
Calculating a Standard Time
(From Work Sampling)
z Observed time (OT) = Tx(ni/n) / #P

z Normal time (NT) = OTx R/100

z Standard time (ST) = NT(1+Allow)

Self Observation - 1
z Find “typical” day of college student
z % of 24-hour day for different activities:
z sleeping
p g ((record this directly)
y)
z classes (sample this and others)
z studying
z sports/leisure activities (TV, internet)
z eating
z transportation, going to-from classes
z miscellaneous
i ll
The following
g considerations should
be kept in mind:
1. Explain the work sampling method before using it.
2. Confine individual studies to similar g
groups
p of
machines or operations.
3. Use as large a sample size as is practicable.
4. Take individual observations at random times, so
that observations are recorded for all hours of the
day.
5. Take the observations over two weeks or more.
 Ad
Advantages
t off Work
W k Sampling
S li

z Can be used to measure activities that are

impractical to measure by direct observation.
z Multiple subjects can be included.
z Requires
q less time and lower cost than
continuous direct observation.
z Training
g requirements
q less than DTS or PMTS.
z Less tiresome and monotonous on observer than
continuous observation.
z Being a subject in work sampling is less
demanding g than being
g watched continuously
y for
a long time.
 Disadvantages
g and Limitations

z Not as accurate for setting time standards as other

work measurement techniques.
z Usually not practical to study a single subject.
z Work sampling provides less detailed information
about work elements than DTS or PMTS.
z Since work sampling deals with multiple subjects,
individual differences will be missed.
z Workers may be suspicious because they do not
understand the statistical basis of work sampling.
Work Sampling vs. Time Study
Advantages Disadvantages
1) Measure simultaneous 1) Expensive for 1 oper.
irregular events
2) Covers days, weeks 2) More difficult for short
3) Easily interrupted cycles,
y fine details
4) Less training 3) Not well understood
5) IInstant
t t observations
b ti 4) N
No record
d off method
th d
6) Less analyst fatigue 5) Ignoring principles ruins
7) Less operator tension study (random!)
( )
8)) No watch needed 6)) Manyy observations