JMIR Publications

ABSTRACT

Background:

The use of targeted advertisements on social media platforms (e.g., Facebook, Instagram) has become increasingly popular for recruiting participants for online survey research. Many of these surveys offer monetary incentives for survey completion in the form of gift cards; however, little is known about whether the incentive amount impacts the cost, speed, and quality of data collection.

Objective:

This experiment addresses this gap in the literature by examining how paid advertising campaigns on Instagram with ads offering no incentive, a $5 gift card, or a $15 gift card influences the cost, speed, and quality of completing a 10-minute online survey.

Methods:

Each campaign had the same budget to spend over four days and targeted 15- to 24-year-olds in three different designated market areas of similar size. Ads directed users to an eligibility screener and a 10-minute online survey if eligible. Ads for each condition ran for either the full allotted time or until there were 150 total survey completes.

Results:

The $15 incentive condition resulted in the quickest and least expensive data collection, requiring 17 hours and an ad spend of $338.64 to achieve 142 completes. The $5 condition took more than twice as long (39 hours) and cost $864.33 in ad spend to achieve 148 completes. The no incentive condition ran for 60 hours, cost $1,398.23 in ad spend, and achieved only 24 completes. The $15 and $5 incentive conditions had similar levels of fraudulent responders, whereas the no incentive condition had no fraudulent responders.

Conclusions:

Overall, we found that a higher incentive resulted in a quicker data collection, less money spent on ads, and higher response rates, despite some fraudulent cases that had to be dropped from the sample. When considering the cost of incentives, a $5 incentive appears to be more cost-effective than a $15 incentive; however, this does not consider other costs that might be associated with running a campaign for a longer period of time. A longer experiment is warranted to determine whether fraud increases over time differently across conditions.