THE PURPOSE AND APPLICATIONS OF SAMPLING
During root cause analysis, it is often necessary to collect data about the
problem and its possible causes. Collecting data might take a long time, be
costly, or require a lot of effort. Sampling is a way of economizing the data collection
process.
The main purpose of sampling is to draw conclusions about a larger
group based on a smaller sample, as long as you are aware of the sample’s
limitations.
Applications in root cause analysis include:
• Effectively collecting data about problems or causes
• Gaining a better understanding of the situation
Types of Sampling
Sampling is a collective term that encompasses several approaches to the
cost- and time-efficient collection of data. Some of the most common types
of sampling include:
• Random sampling. Random numbers are used to determine which
units will be drawn from a larger population. Random numbers can
be found in special tables, by using a computer to generate them, or
simply by throwing dice. An example is pulling out numbers 4, 11, 19,
21, 34, and so on, for testing for defects.
• Systematic sampling. A means to overcome the fact that random
sampling can at times be difficult or even impossible. In systematic
sampling, measures are made at fixed intervals of time, numbers,
length, and so on. For example, every 20 minutes, the number of
customers waiting in line are counted.
• Stratified sampling. A necessary tool when you know that there are
differences between categories within the entire population. In such
situations, data is purposely collected from each of the categories
so that the samples represent the categories in the right proportions
to one another. If a company has seven salespeople, customer
satisfaction scores can be collected from the customers of each,
relative to the number of customers each serves.
• Cluster sampling. An adequate approach when the population is
known to be stable and without much variation. In this case, a group
of the units is taken to represent the whole population—for example,
the entire batch of parts produced during an hour may represent an
entire week’s production.
THE STEPS IN USING SAMPLING
Unlike most of the other tools presented here, sampling is not one unified
tool where the steps can easily be outlined. Rather, sampling is used to support
other tools. Some important issues to keep in mind:
• Assess the nature of the population to be sampled to decide on
a suitable type of sampling approach (take into account the
homogeneity of the population, any clustering of data, and so on).
• Collect the sample of data according to the chosen sampling
approach.
• By calculating simple figures such as averages, means, and so on,
you can test whether the sample is a reasonable representation of
the population.
An Example of the Use of Sampling
Customers of a manufacturer of ballpoint pens used for promotions frequently
complained because many of the pens didn’t work or stopped working
after a relatively short time. The pens’ recipients transferred this same
level of quality to their view of the companies giving them away.
The pens were manufactured in seven different manufacturing lines. To
determine if the quality varied among these lines, the company wanted
to undertake a random quality control assessment of the finished products.
However, to ensure that the same data were collected from all seven lines,
the company needed to introduce a systematic sampling procedure: for one
week, they collected the first three pens produced each hour from each of
the seven lines. Each pen was tested immediately and again after having
been used for one minute.
The data showed clear differences between the manufacturing lines.
Five lines produced pens of adequate quality, while two others generated
close to 95 percent of the defective product. After these stunning results, the
next step was to start looking for the causes of these high defect rates, which
is a different story.
CHECKLIST FOR SAMPLING
❑ Before sampling, assess the sample population to
ensure that a suitable sampling method is being used.
❑ During the sampling, collect data in accordance with
the chosen sampling analysis approach.
❑ After a period of sampling, ensure that the sample
does, in fact, represent an accurate picture of the
entire population.
