Best Practices for Setting up and Launching Online Studies

Introduction

This chapter offers best practices for setting up and launching an online research study.

We will learn how to approach many essential decisions such as how to name and describe a project to attract appropriate participants while avoiding selection bias, how to determine fair and effective compensation, how to use platform tools for targeting specific participants, and how to apply demographic quotas. We will also learn about strategies for improving representativeness, including how to demographically match the sample to the US census. We will examine how to best configure a study to capture necessary participant identifiers for tasks like longitudinal tracking or bonus payments, while still protecting anonymity. Finally, we will learn about best practices for launching a study, including when to launch, how to monitor data collection, how to address participant feedback, and how to solve any technical issues that arise.

After learning how to set up a project, we will examine more complex research designs and the strategies required for their successful execution. We will focus on best practices for conducting longitudinal research, with a particular emphasis on techniques to maximize retention across multiple waves of data collection. We will learn about strategies for clear communication, such as crafting effective reminder messages, and how to structure incentives for sustained engagement.

By the end of this chapter, you will be equipped with the practical knowledge needed to set up and manage online data collection efficiently. Whether you are embarking on your first online study or seeking to refine your approach to more complex projects, the strategies outlined here will help you conduct research with greater confidence and success.

Chapter Outline

Setting up an online study takes just a few minutes, but conducting quality research requires careful planning. This planning often involves knowing how to describe projects to participants, how to set appropriate compensation, target specific samples, and control sample composition. For longitudinal or more complex studies, it is often necessary to combine all these aspects to get the most out of participants and maintain data quality.

In the sections that follow, we will learn about strategies for successfully collecting online data.

Best Practices for Setting Up an Online Project

The basic process of setting up an online study requires connecting a survey to a participant recruitment site, describing the project, selecting the eligibility criteria, launching the study, and managing the incoming data. While each site varies in its user interface and capabilities, the basic principles of gathering data from paid sources of participant recruitment are largely the same. Here, we will learn about the best practices of setting up an online study with the CloudResearch Connect platform (https://connect.cloudresearch.com/).

If you have already created a Connect account (perhaps to participate in studies in Chapter 2 or to collect data in any of the previous chapters) you can log into the researcher portal using your existing credentials. If you are new to Connect, visit https://bit.ly/3DnsQHl, and select Connect for Researchers. Click "Create an Account" and follow the prompts.

After logging in, you will see the dashboard with all your current projects (Figure 14.1). Simply click the blue "Create Project" button in the top right corner to begin setting up a new study.

The five steps required to set up a Connect study are displayed in Figure 14.2. We will learn about the best practices associated with each step.

How to Name a Study

The first step of any project is to give it a project name. Connect provides two places to name a project (Figure 14.3). The first one is a public name that participants will see on their dashboard when browsing studies (Figure 14.4). The second one is an internal name.

It is generally best to use generic wording for study titles like "Survey about attitudes" or "Research study in psychology" to avoid selection bias. For example, if a study is about people's attitudes toward gun ownership, avoid titles like "A study about attitudes toward guns." People with strong attitudes toward gun policy, on both sides of the debate, might be more (or less) drawn toward participating, creating selection bias.

An exception to this rule is if the study has special requirements. These requirements should then be included in the study's title and description, as appears in Figure 14.5. Notice how the study has a very generic name (Academic Study), but before the name the researchers listed several requirements such as access to ChatGPT and Google Chrome. Failing to include this information up front risks confusing participants and creating significant dropout, as people will accept the study and only then realize they don't meet the requirements.

How and Where to Describe Your Project

After naming a project, Connect provides two places to describe it (Figure 14.3). The first place is a brief project description box. Here, researchers can briefly tell participants what the study is about (e.g., answer some questions about personality). This information will appear on the dashboard as they browse studies.

The second place to describe a project is in the instructions to participants box. This section offers expanded information and appears when participants click "View" before accepting the project (Figure 14.7).

For simple surveys, it is common to provide minimal instructions (or perhaps no instructions). However, when a project requires participants to download software, participate in video calls, view sensitive content, complete longitudinal follow-up surveys, or do anything more than answer survey questions, it is best to provide clear details within the instructions box (in addition to a brief mention of this in the title). Ideally, this information will be presented in a way that is easy for participants to immediately grasp, such as with bullet points or a numbered list.

It is important for study descriptions and special instructions to be presented simply because participants do not read these details very carefully. They are much more likely to read the title than the study description. This is also true for the informed consent page. Therefore, it is best to include only the most critical information in the study description field and to present it in bullet point form.

For example, if the study is very long (over 45 minutes), you could write: "This study requires at least 45 minutes to complete. Please only accept if you have 45 minutes right now." This description is short, contains crucial information, and discourages rushing while reducing dropouts.

For longitudinal studies, clearly stating that follow-up sessions are required—along with information about how long they will last and how much participants can earn—allows people to self-select into the study based on their willingness to return, improving retention (e.g., Hall et al., 2020). For other kinds of studies, a clear project description can be equally effective.

How to Anonymously Follow up with Participants

In most online studies, data is gathered through a survey platform like Qualtrics, SurveyMonkey, or Engage. The dataset, on its own, contains no personally identifiable information about the people who participated. However, there are many common scenarios where the researcher will need to link specific survey responses back to individual participants without compromising their anonymity. Here are some of these scenarios.

Longitudinal Research: Longitudinal studies require tracking the same participants across multiple waves of data collection. This is impossible without a reliable way to link each participant's data from one wave to the next. Specifically, a study with two waves of data collection will have two datasets, and the rows of those datasets (corresponding to specific participants) need to be matched up. The only way to do that is with participant IDs.

Awarding Bonuses: In some studies, participants receive a bonus payment for high performance. To deliver this bonus, the researcher needs to know which row in the dataset corresponds to the specific participant receiving the bonus.

Managing Data Quality: If some participants fail attention checks or provide data that indicates fraud (see Chapters 10-11), the researcher might want to reject their submission, flag them in Connect, and prevent them from participating in future studies. Each of these actions requires knowing the participant ID.

Selective Recruitment for Future Studies: Sometimes, a subset of participants might meet specific criteria that make them ideal for a follow-up study. To invite only these people, the researcher needs to be able to identify them.

Follow-up Questions or Clarifications: In some cases, the researcher might need to follow up with a specific participant to clarify a response or address an issue they reported. A unique identifier makes this possible while maintaining anonymity.

In all these instances, the ability to link study data to a unique participant identifier is required. CloudResearch Connect is designed to facilitate this with anonymous participant IDs. These IDs are unique alphanumeric strings assigned to each participant by the platform. They allow researchers to manage participation, track participants across studies, and distribute payments without needing to collect personal information like names or email addresses.

The primary method for linking participant's anonymous IDs to survey data is by embedding the participant's Connect ID into the data file. There are two ways to do this. The simplest way is to ask participants to paste their Connect ID into the survey. This will create a column of participant IDs that can be copied and pasted to Connect for longitudinal follow-up, bonusing, and other forms of participant management. However, this method is prone to errors. Some respondents will not copy their ID accurately or fail to comply with the request for some other reason.

The more certain way to add participant IDs to the datafile is to configure the study on Connect to automatically pass each participant's unique ID to the survey when they click the study link. The option for doing this appears just after you provide a link to the survey in the study set up. Connect provides instructions for capturing IDs with several different survey platforms (Figure 14.6). Figure 13.13 shows what this embedded data looks like in Qualtrics.

How to Figure out Fair Payments

An important part of any study is setting the participant payment. Fair payment starts with understanding platform norms. On Connect, most researchers pay around $10 per hour, and participants consider $11 per hour fair payment (Moss, 2024).

Although the average payment is around $10 per hour, there is a range to what researchers can pay. Often, the amount that is appropriate depends on the project and what participants are asked to do.

The minimum payment researchers can offer on Connect is $7.50 per hour and the recommended payment is around $8.50 per hour for basic surveys and simple tasks. This means for a typical 15-minute study, participants earn between $1.85 and $2.15.

As projects become more complicated, however, payment should increase. When participants are asked to devote extended attention or effort, engage in creative thinking, solve complex problems, use special skills or knowledge, return for multiple rounds of data collection, or engage in something like a daily diary study, they should be paid more than the minimum rate. In a mock jury trial, for example, participants may need to devote an hour to closely reading both sides of a lawsuit and then render a verdict. For these projects, participants often earn $15-20 per hour or more.

When determining what is a fair payment, consider what participants are being asked to do and how close you can get to the payment norms of the platform (about 14 to 16 cents per minute in 2025). Finally, it is a good idea to conduct a pilot study to accurately measure the actual completion time of a study.

How to Find the Participants you Need

Many research studies aim to draw samples from the general population. However, other studies have more specific recruitment goals. These might include people with certain political leanings, like Democrats or Republicans, those who have specific conditions like symptoms of depression or past suicidal ideation, or people within defined age brackets or socioeconomic status groups. There are several methods for selectively recruiting such specific groups of respondents.

After setting payment, there are options for demographic targeting. While many research studies aim to draw samples from the general population, others have specific recruitment goals. These might include people with certain political leanings, like Democrats or Republicans, people with specific conditions like symptoms of depression or past suicidal ideation, or people within defined age brackets or socioeconomic groups. There are many ways to selectively recruit such participants.

The first and often most straightforward method is to use pre-existing qualifications offered by the platform. On Connect, for example, CloudResearch maintains a profile for each participant based on hundreds of questions. People answer these questions when they join the platform and then on an ongoing basis. These qualifications cover a wide range of demographic, behavioral, and attitudinal variables. If the research goal is to sample only Republican participants, for example, a researcher can select this political affiliation when setting up the study (Figure 14.7). Participants who have previously identified as Republican would then be eligible to participate in the study. Others would not.

Sometimes, researchers need to recruit a group of participants for which a qualification does not exist. In such cases, there are generally two options. One option is to request the qualification. On Connect, researchers can click the "Request a Demographic" button (visible in Figure 14.7) and then submit the form shown in Figure 14.8. Connect adds these requests to the participant profile page and usually makes a sample available within a few days.

A second approach is to add screening questions to the survey itself. When using this approach, researchers often ask qualification questions early in the survey. For example, a marketing survey looking to gauge people's interest in a new product derived from avocados might target people who eat avocados several times a week (Figure 14.9). Qualified participants (e.g., people who indicate they "Eat avocados at least twice a week") would continue to the main part of the study while everyone else would be directed out of the survey.

There are many advantages to within survey screening. First, it is more flexible than other options. Instead of using the platform's qualifications, within survey screening can combine people's answers to multiple questions that are designed by the researcher, better ensuring access to the target sample. Second, within survey screening often ensures greater access to very hard to find groups of participants. When qualifications are added to Connect, thousands of participants answer within a few days. But many more participants will never answer. Screening within the survey can reach people who may not complete system qualifications. Finally, within survey screening offers the most current information about participants. Some data that participants provide changes or grows "stale" over time. For example, imagine someone who works from home but later takes a job that requires working in an office. They are unlikely to update the information in an online research platform. Thus, within survey screening offers access to more current information about participants, improving sample targeting.

For all the advantages, however, screening participants in the survey requires some advanced survey set up that draws upon the question logic and branching options introduced in Chapter 13. In most cases, setting up this kind of screening will require consulting a knowledge base or help guide from the survey platform, but the process often works like this.

First, researchers add a branch to the study's survey flow as shown in Figure 14.10. Next, the researchers use question logic to determine who is ineligible based on answers to the screening questions. Beneath the branch, the researchers add an end of survey element and create a custom end of study message for unqualified participants while allowing everyone else to continue to the main part of the study. Within the end of survey element, it is necessary to override the default option and direct participants to a specific URL that is provided by the participant recruitment platform (Figure 14.11).

On Connect, the redirect URL for within survey screening is embedded within a feature that is also called branches. Branches within Connect allow researchers to create a custom study ending for different groups of participants and to track who belongs to which group. Tracking people is important because Connect requires that all participants be paid for the time they spend in the survey. This requirement prevents participants from continually attempting survey screeners without ever earning anything for their time. To pay people a small amount for a survey screener, Connect allows partial payments within the branching system (Figure 14.12). Researchers can write a message to people who are screened out, estimate the time of the screener, and set a partial payment of 25 cents or more. Everyone who is screened out of the study will be marked as a partial completion. Overall, this method provides a way to quickly qualify respondents based on specific criteria, when those criteria are not available as standard platform qualifications.

Creating Quotas

In addition to demographic targeting, quotas offer fine-grained control over the composition of an online sample. For example, while one study might target participants from a specific political party, another study might recruit people from several different political parties and have them appear in specific proportions within the sample. This is where quotas are useful. Quotas allow us to define specific "bins" or subgroups based on one or more characteristics and then recruit participants until a predetermined number is met for each bin.

For example, imagine we are conducting a study where we want an equal number of Republicans, Democrats, and Independents. Setting quotas for these bins will make sure that the sample is not skewed towards one political party or another.

On Connect, quotas can be set for any variable used in demographic targeting. After turning the quota option on, researchers can define each "bin." Defining a bin means entering either the number of participants or percentage of the sample that should meet each criteria. In Figure 14.13, each bin is set to recruit 100 participants, which is a third of the total sample. Connect will stop recruiting participants in each category once that target is reached.

How to Increase Representativeness

Quotas are particularly valuable when the goal is to increase sample representativeness. Connect, offers a template matched to U.S. Census for age, gender, race, and ethnicity. By selecting "Apply Census Matched Template," these quotas are automatically added to the project. Figure 14.14 shows the quotas for age. As displayed, 26% of the sample is set to be between the age of 30 and 44, which is the same as the U.S. population. Researchers can add additional quotas to the template such as geographic region, education and many others.

When is the Best Time to Launch a Study

Once the study is set up, it is ready to launch. But when is the best time?

One advantage of online research is its lack of time constraints. Unlike laboratory studies, researchers can launch projects at any hour. Likewise, participants can complete studies at their convenience. Most activity on Connect and other researcher-centric platforms occurs between 8:00 am and 6:00 pm (see Litman and Robinson, 2020), but participants take studies twenty-four hours a day, seven days a week. The participants who are online late at night may, however, be significantly different than those who participate during the day (e.g., Arechar et al., 2017; Casey et al., 2017; Fordsham et al., 2019).

Participants who are active late at night report lower levels of conscientiousness and higher levels of anxiety, depression, procrastination, internet compulsion, disruptive sleep behaviors, disordered eating, and neuroticism compared to those who are active during typical daytime hours (Fordsham et al., 2019).

If there are theoretical reasons to believe that the variables mentioned above are correlated with your research topic, and may thus potentially introduce some level of bias, it is best to carefully control your launch time.

If there are theoretical reasons to believe that the variables mentioned above are correlated with your research topic, it is best to control the launch time. Doing so can help mitigate bias.

Once a study is ready to launch, there is an option to schedule a specific launch time using the calendar wizard. If we select the desired date and time, the project will then automatically go live at that time and appear on the Dashboard. Once launched, a project becomes visible to eligible participants based on the targeting criteria.

How and When to Communicate with Participants

The Conversations center in Connect operates like an email system while protecting participants' anonymity (Figure 14.15). Researchers can message individual participants or groups by entering Connect IDs into the recipient's box. This system of communication helps when addressing technical issues, explaining a rejection, awarding a bonus, or sending reminders about follow-up waves in longitudinal studies. Participants can also contact researchers to ask questions or explain issues when completing a study.

It may go without saying, but communications should always be polite and professional. Polite communication shows respect for participants' time while injecting some humanity into a largely impersonal transaction.

Best Practices for After a Project is Launched

When a project is launched, Connect creates a list of qualified participants and sends them notifications. These notifications speed up data collection when compared to simply publishing the study without alerts (Litman & Robinson, 2020). Additionally, email notifications pull in less active participants who might otherwise miss the study, as they are less likely to be monitoring their dashboard. Because less active participants receive higher priority in the notification system, studies typically fill with more casual participants than would otherwise be the case (see Robinson et al., 2019).

Regardless of what happens behind-the-scenes, data will begin streaming in once a project is launched. Therefore, it is important to know what to do when a project is live.

What To Do While a Project is Live

While a project is live, the dashboard provides real-time monitoring. It shows how many people are active in the project, the number of submissions pending approval, and the bounce rate (Figure 14.16).

The bounce rate is particularly valuable. It refers to the percentage of people who start the study but then return it. The typical bounce rate on Connect is close to zero or in the low single digits. A bounce rate exceeding 20% typically signals an issue that requires attention. If the bounce rate surpasses 50%, Connect automatically pauses the project to prevent a negative participant experience.

Several factors can cause a high bounce rate, such as a project description that fails to set accurate expectations, a technical problem such as a broken link, or a survey element that participants find confusing or objectionable.

During the study, participants can provide feedback that might indicate the source of the issue. Connect provides two feedback channels. The first is post-completion reviews. After finishing a study, participants can rate their experience, including the accuracy of the time estimate, the fairness of the compensation, and the researcher's conduct. They can also provide open-ended comments that may alert the researcher to any problems.

Second, Connect offers technical error reports. Through this channel, participants can report specific problems with question phrasing, answer options, programming mistakes, downloads, broken links, or media playback issues (Figure 14.17).

Monitoring these feedback channels, especially during the first few minutes after launch, can help identify and address problems before they compromise an entire dataset. If participants report issues that need correction, the researcher can pause the project on the dashboard. Pausing will prevent new participants from accepting the study while allowing those already engaged to complete it. Depending on the issue, the researcher may need to make changes to the survey or Connect settings before resuming data collection.

If there are no problems when you launch a project, all you need to do is sit back and watch the data roll in.

General Best Practices

Pilot Testing

Pilot testing involves opening a study to a small number of participants as a trial run before collecting a complete dataset. Although many research projects operate on tight timelines, pilot testing is worthwhile because it allows researchers to verify that data is being properly recorded, catch programming errors or problems with study materials, check that links and survey logic are functioning correctly, accurately estimate completion time, and determine appropriate compensation.

Connect makes pilot testing simple. When launching a project, researchers are asked whether they want to conduct a pilot or launch the full study. A pilot launch with even a few participants can reveal issues that might have been overlooked. Launching a pilot with ten or twenty participants is a great start. After the pilot is complete, the project can smoothly transition to the full launch.

Fixing Study Errors

When a mistake is revealed during the pilot phase or while the study is live, the first action in correcting it is to pause the study. On Connect, the pause button is located by clicking on the "..." menu and selecting "Pause." Pausing a study prevents new participants from starting it, but anyone currently active will be given the chance to complete their session.

Once the study is paused, the researcher can correct the error. This usually means changing something in the Qualtrics survey or editing the payment, study length, or task description on Connect.

If the changes occur in Qualtrics, simply make the changes, save them within the survey platform, and then resume data collection on Connect. When analyzing the data, you will want to separate people who experienced the error from those who did not, but otherwise the data collection can continue within the same project. If parts of the study need to change on Connect, however, you will need to close the study and clone the project.

Cloning a project copies all the study details into a new project where they can be edited. In addition, cloning ensures that each participant from the original project is ineligible for the second project, avoiding duplicate participation.

After editing a project, you can conduct a new pilot or conduct a full launch.

Conducting Successful Longitudinal Studies

Online research can be used to conduct a variety of complex projects, including video interviews, interactive experiments with multiple respondents interacting with each other, at-home product testing, and many others. Of all the complex types of designs, longitudinal research is perhaps the most common.

Longitudinal studies track the same participants over multiple sessions. We have encountered longitudinal designs multiple times throughout this book. In Chapter 6, we analyzed a longitudinal dataset with five hundred participants who completed anxiety, depression, trauma, and other measures on two occasions over one year apart. As we discussed in-depth in that chapter, longitudinal studies allow researchers to build a stronger case for causal inference by establishing temporal precedence. Longitudinal research also allows researchers to examine developmental changes, the trajectory of attitudes or behaviors over time, and the long-term impact of events or interventions. Many research questions in behavioral sciences are inherently longitudinal. For example, assessing the test-retest reliability of a measure requires having the same respondents participate on at least two separate occasions.

Longitudinal studies are extremely common online. This is because online platforms make it much easier to recruit, track, and retain participants across multiple sessions. However, the success of any longitudinal project hinges critically on participant retention.

Losing participants between waves of data collection can reduce statistical power, introduce bias if those who drop out differ systematically from those who remain, and ultimately undermine the study's validity.

Chandler et al., (2021) examined 1,200 online longitudinal studies spanning numerous behavioral disciplines and including over 36,000 participants. They provided recommendations for best practices when launching longitudinal studies online. By applying these and other recommendations (see Hall et al., 2020) researchers can significantly improve retention and the overall quality of longitudinal data.

The Immense Potential of Crowdsourcing Platforms for Longitudinal Studies

Generally speaking, longitudinal retention in online studies is very high compared to more traditional sources (see Hall et al., 2020). The average retention on MTurk across 1,200 separate studies was close to 70%, with much higher retention over short intervals. Similar and even higher retention rates can be expected on Connect and other researcher-centric platforms. However, retention across longitudinal sessions is significantly lower on market research panels. Typical retention across one week is only around 30% and falls even more with a longer gap between intervals.

For most longitudinal studies, researcher-centric platforms are preferable to other alternatives, such as market research panels, volunteer panels, and university subject pools. Indeed, beyond relatively simple two wave longitudinal studies, crowdsourcing platforms allow for longitudinal designs that are not feasible anywhere else. For example, in one study conducted by IARPA, two thousand CloudResearch participants engaged in 2-hour political forecasting tasks every week for over one year (Moss, 2022). Eighty-five percent of these respondents completed close to 90% of the sessions over that period. And, as discussed in Chapter 9, OpenAI ran a study in which 1,000 respondents interacted with ChatGPT daily for a period of thirty days. Finally in another study, the authors of this book conducted a ten-year follow-up in which 30% of people came back ten years after the initial session. As all these studies demonstrate, the possibilities of using online platforms for longitudinal recruitment are immense and likely provide opportunities for longitudinal follow up beyond any other participant recruitment source.

Study Title and Description

The groundwork for retention in a longitudinal study begins with how the study is initially presented. As emphasized in the previous section on setting up studies, it is important to clearly state the longitudinal nature of the project in both the study title and the detailed description. Participants should be informed upfront about the expected number of waves, the approximate timing or frequency of each follow-up, the estimated time commitment per wave, and any specific requirements for ongoing participation. This will allow people to self-select into the study based on their willingness and ability to commit to the entire project, which can significantly improve long-term retention rates.

Compensation and Incentives

Compensation is important for retaining participants. Offering people above-average payment increases retention in online longitudinal studies (Hall et al., 2020). Additionally, increasing the incentive for each successive wave of data collection or offering a bonus for finishing all waves of the study also improve retention.

In one study, researchers recruited 600 geographically diverse participants for seven sessions that spanned one year (Hall et al., 2020). The payment increased incrementally across sessions from 50 cents to $1, $2, $3, $4, $5, $6, to $7 with an additional $5 bonus for completing the last session. Their retention rate was 73% for the second session, but after that, retention remained very high (around 95%) throughout the rest of the study.

Communication and Reminders

To successfully conduct longitudinal research, it is important to communicate with participants. When researchers remind participants about upcoming study sessions and encourage participation, they often see less attrition than researchers who do not send these messages (Chandler et al., 2021). Often, polite, timely reminders about participating in a study are enough to increase participation and minimize attrition. These reminders also help participants stay engaged and aware of upcoming sessions.

Reputation Qualifications

When maximizing retention is crucial, participant qualifications can be helpful. As discussed earlier, platforms like Connect allow researchers to target participants based on various characteristics. Selecting participants who have a track record of active participation on the platform can increase the likelihood that people will return for future waves. In fact, highly active participants are more than 60% less likely to attrit compared to less active participants (Hall et al., 2020).

Tracking IDs

Effective tracking of participant IDs is also important for longitudinal research. As described above, embedding anonymous participant IDs into the dataset allows researchers to match data from the same participant across multiple waves. This is essential not only for data analysis but also for managing communications, sending targeted reminders to those who have not yet completed a current wave, and for accurately distributing any wave-specific payments or completion bonuses.

Facilitating Large-Scale and Intensive-Longitudinal Designs

Behavioral scientists are increasingly interested in large-scale longitudinal research. This may mean tracking the same participants over an extended period or asking people to complete intensive-longitudinal research, such as experience sampling and daily diary studies (Bolger and Laurenceau, 2013).

Managing these complex studies presents significant logistical hurdles. Imagine a daily diary study requiring three data collection points each day for a month. This would mean setting up and managing 81 separate online studies. This includes figuring out who is eligible for each wave (depending on prior participation), sending reminders, tracking responses for each specific wave, and implementing potentially complex incentive structures based on performance or completion of multiple stages.

The Connect platform offers a feature called "waves" that is specifically designed to automate the setup and management of such studies. Researchers can use a setup wizard to configure the entire study timeline at once. This includes specifying the number and frequency of all data collection waves, defining when each wave should launch, determining which participants are eligible for each stage, and setting the compensation structure, including potential bonuses for completing a certain number of waves. By using these tools, researchers can conduct more complex studies investigating how social and psychological phenomena change and unfold over time without the hassle of manually administering the study.

Summary

This chapter reviewed practical strategies for effectively setting up and launching online research studies. We examined best practices for naming and describing a project to minimize selection bias, the importance of clearly communicating any special study requirements, and methods for anonymously tracking participant IDs. We also covered how to determine fair compensation based on platform norms and study complexity, and how to use platform tools for targeting specific participant groups and applying demographic quotas, such as those matched to the U.S. Census, to enhance sample representativeness. Then, we also discussed considerations for the optimal timing of study launches and the importance of real-time monitoring of data collection, including how to interpret participant feedback and address technical issues promptly.

We placed a strong emphasis on conducting successful longitudinal studies, focusing on strategies for increasing participant retention. Key strategies discussed included setting clear expectations about the study's duration from the outset, staggering the compensation and incentives to encourage retention, selecting participants who are more active in the platform, and the role of consistent communication, such as sending reminders for follow-up waves. We also explored how platform features, like Connect "Waves," are streamlining the management of large-scale and intensive-longitudinal designs.

Throughout this chapter, a central theme has been that successful online research requires thoughtful and participant-centered practices. The practices apply regardless of the specific online platform used or the precise nature of the research questions being asked. They are also the focus of the next chapter on research ethics and the participant experience.