Confidence Intervals in Survey Research

Confidence intervals are critical to survey research but can be one of the more challenging topics to explain to non-researchers. Technically speaking, a confidence interval is defined as an “interval estimate of a population parameter.” Here is what this means in the context of customer satisfaction surveys...

Confidence intervals are critical to survey research but can be one of the more challenging topics to explain to non-researchers.  Technically speaking, a confidence interval is defined as an “interval estimate of a population parameter.” Here is what this means in the context of customer satisfaction surveys:

When we conduct a survey for a client we typically receive a file that contains all of transactions for that period – this file is the population of transactions.  We then complete surveys on a portion of this population. The resulting pool of surveys is called a sample.

From this sample we can create a statistical estimate of what the actual customer satisfaction level is among the entire customer population.  This estimate comes in the form of a range of values, or an interval.

For example, we may calculate that 70% of customers are satisfied with the service they received. This statistic is referred to as the point estimate. Using this same example, we may tell a client that the actual satisfaction level of their customer population falls between 65% and 75%. This range is the confidence interval.

You are most likely familiar with the polls used in political races by media outlets. At the end of the news story you may see something like “this poll has a margin of error of +/- 5%”. This just means that the confidence interval is determined by adding and subtracting 5% from the point estimate. This same type of math applies to survey results.

It should be noted that margin of error is not always +/-5% and in fact varies sample to sample. Margin of error is dependent on the amount of survey sample available, the actual survey scores, and sometimes the size of the customer population.  A good survey supplier will develop a sampling plan to maintain a consistent margin of error, and will execute this plan through systematic quota control.

A margin of error is also always attached to what is called a confidence Level. A confidence level describes the level of certainty in an estimate.  A confidence level of 95% or greater is considered acceptable, with 95% widely used in survey research.  At a 95% confidence level, we would be assured that 95% of the time, our confidence interval will contain the actual population parameter.

The statistics involved here are very standard and can be found in hundreds of books and websites. Luckily for researchers, there are also several software programs that do the work for us (see below for literature and software suggestions).

Finally, confidence intervals can also be determined for other statistics besides proportions. Proportions, however, are fairly typical point estimates used in customer satisfaction research (ex. % Top Box, %unresolved issues, % First Call Resolution, etc.)

A recap of terms as it applies to customer satisfaction surveying:

·         Confidence Interval – An interval estimate of a population parameter.

·         Population – This is the entire pool of customers.

·         Sample – The portion of your customer population that are surveyed.

·         Population Parameter – A measure that includes your entire customer pool. This is thus not an estimate but an exhaustive, all-inclusive figure. Usually not known/retrievable.

·         Point Estimate – The statistic calculated from the survey (for example, %Satisfied).

·         % Margin of Error – A percentage that can be added and subtracted from the point estimate to determine the confidence interval.

·         Confidence Level – Describes the level of certainty in an estimate

 

Recommended Texts and Websites

·         Statistics For the Utterly Confused by Jaisingh  - A good, quick reference  for any researcher

·         http://isixsigma.com

Recommended Software

·         SPSS – Robust analytical software used in the sciences. We use it in most in-depth analyses

·         Minitab – An excellent tool for running project statistics and analysis quickly and effectively

·         Microsoft Excel (We have created a formula in Excel that calculates confidence intervals and are happy to supply to our customers upon request)

 

 

 

 

Comments

There have been no comments made on this article. Why not be the first and add your own comment using the form below.

Leave a comment

Please complete the form below to submit a comment on this article. A valid email address is required to submit a comment though it will not be displayed on the site.

HTML has been disabled but if you wish to add any hyperlinks or text formatting you can use any of the following codes: [B]bold text[/B], [I]italic text[/I], [U]underlined text[/U], [S]strike through text[/S], [URL]http://www.yourlink.com[/URL], [URL=http//www.yourlink.com]your text[/URL]