Data Science

October 9, 2023

Why the Sampling Ecosystem Sets Up Honest Participants for Failure

This article discusses how the online sampling ecosystem favors professional respondents and bad actors. It advocates for a transformative shift towards focusing on rewarding quality survey behavior that aligns more closely with how honest people take surveys.

Why the Sampling Ecosystem Sets Up Honest Participants for Failure

The challenges with data quality in market research are not new, and it is clear that they won't simply disappear without proactive mitigation. With this in mind, the Global Data Quality partnership that coordinates data quality efforts across major international market research associations signals a serious commitment to finally addressing the crisis.

When it comes to improving data quality, the participant is key: we can't conduct good research without a high-quality sample. While there are small steps researchers can take to ensure we’re working with good-quality data, we remain at the mercy of the larger ecosystem when it comes to sampling and the intricate process that takes place behind the scenes.

Unfortunately, this system is designed to attract exactly the type of participant we wish to avoid: the professional respondents and bad actors who have turned survey participation into a career. Even if the system succeeds in recruiting a greater number of ordinary individuals into panels, they are on a path toward failure.

Rewarding quantity over quality

To examine the problem more closely, let’s take a look at a potential root cause: Earning-Per-Click (EPC).  

EPC measures how much sample suppliers in a marketplace, such as Lucid, or Cint, can expect to earn per survey. Naturally, suppliers want to maximize profits so they direct traffic to surveys with higher EPC which are easier to monetize, like surveys with a higher incidence rate, shorter interview length (LOI), and lower drop-off rate. Surveys that don't meet these criteria struggle as they are likely to be pushed down the suppliers’ priority list.  While increasing the cost-per-interview (CPI) can help, it's rarely enough to make up for poor survey metrics.

Likewise, individuals who qualify more often, complete surveys more efficiently, and promptly proceed to the next one, earn more in the long term. Coupled with abysmal incentives on a per-survey basis, this system has created the perfect environment for professional participants and bad actors to flourish. It has been easy for them to learn that quantity is rewarded over quality. This survey behavior is the polar opposite of how ordinary, infrequent survey-takers approach surveys.

Ordinary survey-takers are set up for failure

While individuals with questionable intent have figured out how to profit from surveys, people who only sporadically participate in surveys are at a real disadvantage. The way ordinary people take surveys goes against the EPC mantra, which makes surveys harder to monetize for sample suppliers.

How do ordinary people take surveys? Compared to online panelists, infrequent survey-takers haven’t been conditioned to bend the truth to qualify for surveys. They’re less likely to say “yes” to satisfier questions, or to click on a large number of attributes to maximize their chance of being selected for the survey. This results in a lower incidence rate, which is not attractive to sample suppliers.

They also read more carefully and do not benefit from familiarity with common survey questions, which would enable them to skim through a survey while still grasping the essence of the questions. In fact, it takes them approximately 20% longer to complete a survey, relative to an online panelist. Despite the fact that a longer LOI could imply that participants are more thoughtful and attentive, it hurts the EPC and ultimately impacts traffic to the survey.

Finally, ordinary people have less patience than online panelists. They are, indeed, twice as likely to drop off. One could argue that they drop off before the quality of their answers suffers, which is a good thing. However, high drop-off rates negatively impact the EPC and, consequently, the traffic sent to your survey.

Reimagining survey monetization to protect the good participants

Much of the discussion about improving data quality revolves around the participant experience and incentives. Though critical, the participant experience doesn't exist in a vacuum, so to improve data quality, the sampling ecosystem needs to start rewarding quality.

Offering higher incentives may help us attract higher quality participants, but it also runs the risk of attracting more fraudulent individuals, potentially resulting in increased costs without improving data quality. Without any change to our approach, we will continue to reward the same dishonest participants, all the while failing to retain the ordinary individuals we work so hard to recruit.

To see sustainable improvements in data quality, we need to design a system that supports the type of survey behavior we want to promote and reward: More thoughtful and honest responses, fewer serial survey-takers. 

Aligning survey metrics with human behavior 

The current emphasis on EPC metrics in survey research prioritizes quantity over quality, leading to issues with data integrity. Honest, ordinary survey participants, who provide more thoughtful and genuine responses, face disadvantages in this system.

This approach goes against the overall goal of improving data quality. If we aim to attract and retain better-quality participants, we need a fundamental shift in the sampling ecosystem, focusing on rewarding quality survey behavior that aligns more closely with how ordinary people take surveys.   

While such a radical change may seem daunting, it is crucial to developing a sustainable and thriving industry. 

surveyssample qualitydata collection

Comments

Comments are moderated to ensure respect towards the author and to prevent spam or self-promotion. Your comment may be edited, rejected, or approved based on these criteria. By commenting, you accept these terms and take responsibility for your contributions.

SW

Scott Worthge

March 19, 2025

All good points, Karine. I have to mention a few others, though. My framework is that after several decades of watching the online research industry get born and grow into what it is now (yes, I go back that far in MR), I see "data quality" as being subject to five major influences - source for inviting someone to a survey (including targeting, river and anything else to start the process), validation that you have the right person to allow them to start the survey, engagement during the survey (which includes attention-check measures as well as determination of how well the respondent is actually conducting themselves across open-ends as well as the usual statis choice questions), data evaluation after fielding is completed (including AI-tool use for respondent scoring and anything else you can employ to review your results) and replacement of those results not meeting known quality standards. At any step in this process your "data quality" can go to crap. You invite the wrong people, don't qualify them, don't engage them well, don't check the results to see the garbage, and don't have a plan to get better data for what you toss, there's your "bad data". A lot about how sample suppliers do what they do is in need of change and improvement - tech exists to help greatly here. We also need to critically evaluate how to assess the engagement process during the survey, which includes survey design and incentives. New tools and processes exist here as well for assessing engagement using AI-led scoring. Haven't found a good AI-backed survey-builder yet but I'm looking. On the back end more tech exists and works for identifying and tossing the crap and synthetic data modeling (NOT using SD personas) instead of refielding. LOTS of new tools available and I'm testing all of them to see what we all can do for improving end to end data collection. So, yes - the old EPC model, right? We can deal with the realities of its shortcomings in a lot of ways. You and I should get together - I'd love to show you the real-world testing I'm doing of new tools and processes to make all of this better. Nothing my company sells, nothing we can make money from - just new solutions emerging to deal with what we have to work with now.

Olivier de Gaudemar

Olivier de Gaudemar

March 17, 2025

The concept of a screening questionnaire has to go. A question is a question. The new players who "aim to attract and retain better-quality participants" will most likely come from outside the current ecosystem.

JD

JD Deitch

March 17, 2025

Too right, Frank.

Disclaimer

The views, opinions, data, and methodologies expressed above are those of the contributor(s) and do not necessarily reflect or represent the official policies, positions, or beliefs of Greenbook.

More from Karine Pepin

Truth To Be Told: Five Realities About Online Sample That Compromise Data Quality
Data Quality, Privacy, and Ethics

Truth To Be Told: Five Realities About Online Sample That Compromise Data Quality

Explore five key truths about sampling, uncovering fraud, low-quality respondents, and transparency issues that have eroded data quality over two deca...

From Deliverables to Research Assets: How Insights Teams Can Leverage Content Design Principles for Greater Influence
Research Methodologies

From Deliverables to Research Assets: How Insights Teams Can Leverage Content Design Principles for Greater Influence

Learn key principles of content design that enable researchers to distill insights into assets, fostering stakeholder influence and sustainable busine...

The Cost of Being Wrong: How Overconfidence in Ineffective AI Detection Tools Impacts the Research Ecosystem
The Prompt

The Cost of Being Wrong: How Overconfidence in Ineffective AI Detection Tools Impacts the Research Ecosystem

Discover the challenge of identifying AI-generated open-ended responses and the potential consequences for researchers and the market research industr...

Humans and AI: A Collaboration to Assess the Quality of Open-Ended Responses in Survey Research
The Prompt

Humans and AI: A Collaboration to Assess the Quality of Open-Ended Responses in Survey Research

Over the years, significant time and resources have been dedicated to improving data quality in survey research. While the quality of open-ended respo...

Sign Up for
Updates

Get content that matters, written by top insights industry experts, delivered right to your inbox.

67k+ subscribers