Here are some implementation and design tips to consider when construction web-based surveys:
In an ideal world with unlimited resources, researchers should make first contact with potential respondents through regular, hard copy snail mail. Since trust is a huge issue for surveys, one would ideally have a formal letter on official organization letterhead that is hand-signed by the primary-investigator. If possible also have a second letter from a credible outside source, perhaps one known to the potential respondents, vouching for your trustworthiness and credibility as a researcher. These measures can go a long way to laying a trusting foundation between researchers and respondents.
However, the world is not always ideal and we oftentimes must make initial contact with participants via email. This tipsheet provides some tips on how to make these emails more effective:
Personalize them with respondent name
No html, graphics, or specialized text formatting – just plain text
- Fancy graphics will often not appear automatically
- Unless you can make your logo appear automatically
Include hyperlinks links to the survey plus a cut and pasted URL if possible as many email programs will automatically disable hyperlinks
Always make initial contact from your .edu account – no Gmail, Yahoo, etc.
- Official email accounts are credible; free accounts are not
Link to your official .edu webspace – If you’re not online, you don’t exist.
Always disclose how you got their email and selected them
Say how long the survey will take
NO ATTACHMENTS! Respondents do NOT trust attachments because of viruses
In your email where you deploy the survey, the last statement before the link should be an informed consent, something like:
- By clicking the “next” button, you are consenting to participation in this survey and indicating that you have read and understood the directions above. You may skip any question you choose not to answer. Again, your participation is greatly appreciated.
For follow ups, use the same email but vary the contact time and the appeal made
Be explicit about survey directions, especially technical instructions
If you are surveying cross-nationally, match the contact email to the country, if possible. Just like you may not trust emails from Nigeria, Nigerians may not trust emails from the US.
Avoid talking about yourself too much. Even though you’re the “So and So Endowed Chair of Whatever at Big Name University,” your respondents have probably never heard of you, couldn’t care less, and aren’t that impressed. Your invitation is about thanking your respondent for giving you their time, not for promoting yourself.
Web surveys are primarily visual, so look and clarity are paramount
PRETEST thoroughly; gather feedback on look and the ease of using the interface
- Check the look on multiple laptops and desktops, and on different screen settings
You want a simple but attractive look
Be consistent in the design across the survey
Data quality is more important than being an artist
- Simple and easy surveys yield better data quality than fancy surveys with lots of bells and whistles
Remember, a significant number of older people still don’t use mice – Is your survey accessible to the pure keyboard folks?
Assume older technology – dial up, older computers, older browsers
Be creative, but follow some guidelines…
Don’t clutter pages – elegance!
Remember how the eye travels – from 10 o’clock and then clockwise around the page; design around that travel path with the most important elements where the eye starts
- Use appropriate size
- Your response options should be at least one font smaller than your questions
- Don’t mix fonts
- Avoid fancy fonts – Times New Roman (especially) and Arial work best
- Bold, italicize, or underline key words and phrases, but only do one at a time
- Italicizing or bolding instructions is common
- Bolding questions is also common, but don’t bold both directions and questions
- Shading backgrounds is good, especially on general instructions or in matrices (white and then some other color alternating). Light gray is best, but some pastels may work.
- White is boring, but it works and is trust-inducing. Lighter shades of blue are also excellent colors.
- Consider white, black, and gray your free colors. Choose at most one other.
- Again, pastels are best. Avoid colors that are too dark or too light, plus really bright colors.
- Always check your color scheme on different monitors and adjust accordingly.
- Remember, your goal is data quality, not to be the Picasso of web surveys
Scrolling and pagination
- Left-right scrolling should be avoided at all costs
- Limited up-down scrolling is ok
- Respondents would rather see more pages than have to scroll down too much
- It’s ok to put multiple elements on a page, even desirable; one question per page makes the survey too long, but balance this with scrolling
- And beware of ugly white space
Progress bars are a wash – they engage some, but turn off others
Never put columns of questions on a page – ONE COLUMN ONLY!
If instructions apply to questions on multiple pages, repeat instructions
Disable the back button to avoid optimizing behavior
If DK or unsure are options, show them
Match text boxes to the content expected
- Match on length particularly – longer boxes when more text entry is expected, shorter boxes for less text
- If numbers are desired, use validation!
- Leave a little space between the scale and the question
- Pure horizontal or vertical layouts for response scales are best
- Never break up a scale into two columns
- Be as consistent as possible
- If scales are horizontal, there should be adequate but equal space between responses
- Drop boxes are ok, but not desirable – getting into them is just another little burden on respondents
- Be professional
- Use them to your advantage to explain, emphasize, or illustrate concepts, but only if it is essential – USE THEM SPARINGLY!
- Having your logo up top in the banner is ok, and this is a “free” graphic
Avoid pop ups. If you need a link so that some material is accessible, have it appear as a new tab in the same browser. In Qualtrics, you will need to create these as separate surveys, generate a stable URL for the survey/pop up, and then insert a hyperlink that connects to that stable URL.
End of survey messages
- Develop your own
- If you filter someone out for ineligibility, come up with a separate message that says this and thanks them for their effort.
Author: Patrick R. Miller, DISM Survey Research Associate