How to Perform a Longitudinal Study: Tracking Your Performance Over Time

A survey is like a snapshot: From one survey, you can only draw conclusions about a single time, place and group of people. And that’s often all you need to know.

But sometimes, you also want to understand how the people you surveyed are changing. In that case, one single survey (or snapshot) is not enough – so we need to repeat surveys in order to track and understand trends over time.

If you want to see how people are changing, there are two ways you can do this. The first is benchmarking, which means that you are asking different groups of people the same question over time to see how views change. The second way, and the focus of this post, is a longitudinal survey.

When you run a longitudinal study or survey, you’re essentially following the same group of respondents over a long period of time – for weeks, months or even years.

This differs from a cross-sectional survey, which is a fancy way of saying that each of your survey respondents only completes the survey once, but you might conduct the survey multiple times to collect some benchmarking data. (That’s your snapshot survey.)

How repeating surveys will help you make smarter choices

While you may not be in the market for running a decades-long study in the near future, you can benefit from repeating surveys and tracking changes in your respondents’ attitudes and behaviours over time. (By the way, when you survey the same people time and time again, you’re running what’s also called a panel survey.)

For example, let’s suppose that you’re an online marketer who wants to know how your readers will react to a new email newsletter design.

Instead of just sending your readers a survey after you’ve changed your design, send them a survey asking them what they like (and don’t like) about the current design (i.e., create a concept test). You can even use their feedback to inform your newest design!

Then send them a follow-up survey after you’ve sent them the newly-designed newsletter. Because you’re surveying the same people, you can compare their attitudes and opinions about the first design against how they react to the second design and smaller changes will be statistically significant. If you decided to do two cross-sectional surveys with different groups of people, you would need to see a larger change in order to see a significant difference.

If you make more changes to the design based on your readers’ feedback, you can continue to refine your design over time and ensure that satisfaction ratings don’t dip below the initial satisfaction ratings for the first design.

An alternative to the longitudinal panel survey

Repeating surveys with the same panel works well when you’re tracking changes in your respondents’ attitudes and behaviours, but sometimes you’re unable to survey the exact same people time and time again.

In this case, even if you’re sending your newsletter out to the same people, you may not be able to capture the same opinions. People may unsubscribe from your newsletter and newer readers may come along.

That’s when you conduct something called a rotating panel survey. All you need to do is gradually rotate a portion of your initial sample out of the panel survey and supplement them with new readers.

This way, your survey will provide a good estimate of the opinions of your entire readership, old or new, while at the same time capturing the changing opinions of the same group of people.

3 tips for creating surveys for a longitudinal study

Here are three things you need to bear in mind when creating your longitudinal study:

1. It’s better to keep the questions identical for all of the surveys. Research has shown that changing the way a question is asked can result in substantially different answers, even from the same people. So to examine trends with the most accuracy, you should try your best to ask the same questions in every survey you repeat.

2. When planning a panel survey, remember that not everyone who responded to the first survey will respond to the second survey. So if you want to do a panel survey where you survey the same group of people three times and want at least 1,000 respondents for your third survey, you’ll need to survey more than 1,000 respondents in your first survey to account for people who probably won’t respond to your second or third survey.

3. You need to decide how frequently you want to run your survey. If you conduct surveys too frequently, you could waste your resources and time, since not enough time may have passed for any change to occur. However, if you conduct surveys too infrequently, more people may drop out of your panel survey because they’re fed up of taking too many surveys, which means that you’ll have fewer respondents in your later surveys.

When you think about it, the applications for a longitudinal study are endless. You can see whether your new ad actually influenced people purchasing your product or follow up with product purchasers to see whether they’ve enjoyed using your product.

Even though one dataset can shed light on a single occurrence, the context that comes from repeating surveys over time will help you make informed decisions and improvements.

Authored by Mingnan Liu, Survey Scientist, SurveyMonkey.