<

Conducting a Survey

Published on
16,966 Points
5,666 Views
13 Endorsements
Last Modified:
Awarded
Editor's Choice
In doing marketing for any business it is sometimes difficult to know how to effectively reach your target audience.  What drives your consumer to buy your product or use your service?  What does your target audience respond to?  Does your target audience like your new product/feature?  If you don't know the answers to these questions, a survey may be the best way to find out.  But exactly how do you write a good survey that will give you the information that you need?  Well, the process of conducting a survey, analyzing the results, and drawing conclusions based on the results can be a very complex process.  Here are some simple steps to follow to get you started.    

1. Determine What You Need to Know

It is important to decide what information you need to receive from the survey before you start.  The first thing you want to do is determine what problem you are trying to solve and what information you need in order to solve it.  Once you have figured out what information you need to know, then determine what conclusions you would like to draw based on the data that you receive.  Then formulate a specific plan of attack focused on the end results, and use this plan to guide you in writing your survey.  It is also important to determine who you want to send your survey to.  Your target audience will influence the survey method, what kind of questions you ask, and the validity of the data you receive.  

Setting up a good foundation will allow for everything else to fall into place.  If you figure out what information you need to know before you begin to write questions, the questions will be focused on the information you need, which will lead to better results and conclusions.  

This blog entry explains this in further detail and gives great examples.  

2. Choose the Best Method

There are different types of surveys that you can use: mail surveys, online surveys, telephone surveys, and face-to-face interviews.  Consider which method works best for your target audience and the resources you have for conducting the survey.  If you choose to perform an online survey, you can use the following survey tools to help you create your survey: SurveyGizmo, Wufoo, SurveyMonkey, and Zoomerang.

3. Ask Good Questions

The most important thing when creating a survey is asking good questions that will give you the data that you need.  Focus your questions around the information that you need.  Questions are either focused on respondent's behaviors or their attitudes and beliefs.  The order of the questions in the survey is also very important.  The questions most relevant  to the information you want to know should be placed at the beginning of the survey.  Group related questions with each other, so questions focused on behaviors should be placed together and questions focused on attitudes and beliefs should be grouped together.  All demographic questions (age, residence, income, etc.) are best placed at the end of the survey.  

There are a few more basic things to keep in mind when writing your survey.  It is important to include a brief introduction to the survey stating the purpose and providing instructions.  A common mistake when writing survey questions is asking two questions in one.  For example, "Is the new exercise program motivating and challenging?"  It is better to separate this into two questions; one asking if the program is motivating and another one asking if it is challenging.  Keep the formatting and sentence structure consistent among all questions asked, and thank your participants at the end.

Check out Qualities of a Good Question for examples of what to do and what not to do when writing survey questions.

4. Gather Enough Data

You cannot draw any conclusions from your survey results if you do not have enough data.  It is usually not feasible to reach the entire population size*, so targeting a sample that represents the population is a more reasonable approach to selecting who to send your survey to.  You must distribute more surveys than you need responses in order to account for those who will not respond.  Not all of the people that you send a survey to will fill it out, so you must plan accordingly.  The following table shows the number of completed surveys that you need in order for the results to be statistically significant.  

Number of Surveys Needed
The best way to ensure that you receive enough usable surveys is to figure out the number of completed, usable questionnaires needed in the final sample, and then work backward.  So assuming a population size of 5000, you can conclude from the table above, that you need 880 completed surveys in order for your results to be significant.  With the following calculations you can conclude that you will need a starting sample of about 2716.  Assume that 90% of the people you send a survey to will actually receive it.  Of that 90%, assume that on average about 40% will respond, and that 90% of the returned surveys are complete and usable.  (880 / 0.9 / 0.4 / 0.9 = 2716)

5. Analyze Results and Draw Conclusions

After you conduct the survey and receive the responses, analyze the results and draw conclusions based on your analysis.  Computer programs such as SPSS  (recently renamed PASW) and Minitab are used for statistical analysis.  You can import the survey data into these programs which run the statistical analysis for you.  Once you have analyzed the data, it is important to look at the results for each question, one at a time and draw conclusions based on each question.  These conclusions will guide you to make the best marketing decisions for your company.

For more information about statistical analysis, check out Statistical Help.

The next time you are in need of more information regarding an important marketing decision, conducting a survey may be your answer.  Just remember to focus on the end goal, choose the best method, ask good questions, and draw conclusions from statistical analysis.  Surveys allow you to understand your target audience and guide you to make educated marketing decisions reinforced with statistical evidence.

*Population size is the group of people with the characteristics that you wish to understand.
**Resource: "How to Conduct Your Own Survey" by Pricscilla Salant and Don A. Dillman  
13
Comment
Author:Katie248
8 Comments
 
LVL 18

Expert Comment

by:WaterStreet
Nice article.  In my opinion, it seems to give all the basics for a user to go all the way to survey completion and into statistical analysis without further advice, in most cases.  Voted as helpful.
0
 
LVL 54

Expert Comment

by:b0lsc0tt
Very interesting article.  I wish I remembered more from my statistics classes and needed to use the information more often.  Actually I might not really mean the latter since that class made me remember there were somethings in math I didn't find natural, easy, and enjoyable.

Great points on the importance and usefullness of surveys though.  I couldn't agree more and was amazed how hard it can be too make a good survey (with good results/conclusions).  It really is a key in business though and I thought your summary and info good no matter what the readers interest or understanding is.

I am a little amazed by the table though.  The percent of usable, complete surveys sure does take a nose dive at a certain point.  Is that really true to that extent?  What type of survey or are those numbers pretty consistent whether phone, email, etc?  Is the table from the resource you mentioned at the end or another source?  I don't disagree with the general "idea" it presents but found the actual figures a little surprising (i.e. 1 more survey returned for 100 times the surveys sent in the last 2 figures).

Thanks for the contribution and time to put it together.  I almost miss my crazy (but extremely smart and interesting) Stat professor. :)

bol
0
 
LVL 21

Expert Comment

by:alainbryden
b0lsc0tt, the chart column labels may be misleading. The second column is not an indication of how many surveys you can expect to get back, it is how many surveys you *need* to get back, to ensure your results have a 95% confidence interval.

Great article Katie. You really stuck to the essentials unlike so many novels that get carried away with making statistical validation seem more complicated than it is.

Alain
0
Upgrade your Question Security!

Your question, your audience. Choose who sees your identity—and your question—with question security.

 
LVL 3

Author Comment

by:Katie248
b0l,

I was also a little surprised by the table, which is from the source i sited at the bottom, the first time I saw it.  And I could be mistaken here, but I think what it is showing is that the larger the population, the more representative your sample size is.  So with a large population, you reach a point where the sample size is enough, and you will be able to draw the same conclusions with that sample size than if you received anymore data from more people.  I hope that makes a little more sense.  

Alain,  

I agree, the statistical analysis side can seem very complex (and in some cases it is, especially for someone who does not have a lot of experience in stats).  But I wanted to touch on the basics here.  A more in depth article could be written just about the statistical analysis of the data results.  

Thanks to both of you for your interest in my article,
Katie
0
 
LVL 93

Expert Comment

by:Patrick Matthews
bol,

The result in the table, with the flattening out of the number of completed surveys required, looks odd, but is supported by the math.

This is basically a binomial distribution, with p = 0.5.  (The greatest number of surveys needed comes when we expect the results to be evenly divided, thus using p = 0.5 is a conservative practice.)  In an infinite population, the number of surveys needed such that the margin of error is +/- 3% would be:

n = z^2 * p * (1 - p) / E^2

Since z for 95% confidence is about 1.96, this reduces to:

n = (1.96 ^ 2) * 0.5 * 0.5 / (0.03 ^ 2) = 1067 and change

However, whenever the sample includes a significant proportion of the total population, then to get a better fix on the margin of error you need to make the "finite population correction" (see http://en.wikipedia.org/wiki/Margin_of_error).  (You can always use the finite population correction to adjust the variance, but for situtations where the sampling proportion is less than 5% of the population, the adjustment is so small that it can usually be ignored.)

When the total population is only 100 and you survey 92 of them, the possibility that the sample result and true population value are not the same is extremely low, and thus with only 92 surveys completed you can still report a margin of error of 3%.

In Katie's table you can see the curve start to flatten out at around 25,000 or so, which makes sense, because the number of surveys required for that entry is roughly 4% of the total population.

By the time you get to the really big numbers on the table, the populations are large enough that, for the purposes of the formula, they may as well be infinite.

On a separate note, it appears to me that the label 'Standard Deviation' in the table is incorrect.  It should be 'Margin of Error', 'Confidence Interval', or the like.  The standard deviation is just one of the inputs that determine the margin of error/confidence interval.  [This is a replacement paragraph, as per poster email, inserted by A101, 5:37 PM EDT]

But otherwise, a well-written article, and one that forced me to dust some of the cobwebs off of my stats memory.

Patrick
0
 
LVL 18

Expert Comment

by:WaterStreet
Patrick,

Wow! Thanks for the effort you put into that.

The EE Page Editors are available to make minor corrections and updates after an article is published.  The intention is to make quality articles available and to promote them.  If the author requests a specific correction then either I or another Page Editor can make that update.


WaterStreet,
EE Page Editor
0
 
LVL 6

Expert Comment

by:Jenn Prentice
Congrats Katie!!!
0
 
LVL 18

Expert Comment

by:Ravi Agrawal
Niced one,

Ravi.
0

Featured Post

Free Tool: Path Explorer

An intuitive utility to help find the CSS path to UI elements on a webpage. These paths are used frequently in a variety of front-end development and QA automation tasks.

One of a set of tools we're offering as a way of saying thank you for being a part of the community.

Join & Write a Comment

An overview of how to create reports in Adobe Analytics (formerly Omniture Site Catalyst) using pageNames, events, eVars and props. This video will show you how to install the Omniture Debugger tool so can see (and test) what is being passed int…
Use Wufoo, an online form creation tool, to make powerful forms. Learn how to choose which pages of your form are visible to your users based on their inputs. The page rules feature provides you with an opportunity to create if:then statements for y…

Keep in touch with Experts Exchange

Tech news and trends delivered to your inbox every month