The Interactive Marketing Research Organization (IMRO, at the time a division of the Marketing Research Association) developed this seminal document at the onset of online panel research in 2006 in order to inform the survey, opinion and marketing research profession of issues surrounding panel management. This document then became the foundation for later efforts in understanding online panel quality.

.................

Preamble

Over the past decade, the growth of online forms of survey research data collection has eclipsed most industry expectations. In this time of tremendous growth and technological change, the proliferation of techniques and business models has challenged many of the traditional notions of research ethics and proprieties. As the industry moves to a new era of stabilization, many previously tenuous problems and issues are now coming into sharp focus.

This set of guidelines is the culmination of many voices from within the online research world and represents the best thinking as to the questions that individual practitioners and panel providers must be able to address to their clients’ satisfaction, along with recommended policy points, where appropriate.

Panel Purpose and Services

Is the panel used solely for market research? If not, what are its other functions?

The basis for contact in a panel must be clearly established and documented. All panelist information and data must only be used under the terms for which the panelist previously opted-in. At a minimum, panelists must have agreed to contact for ‘business purposes’, while agreement for ‘marketing research purposes’ is preferred. Even if an established business relationship (EBR) exists or is created, one should never contact panelists for sales purposes if they were recruited solely for marketing research.

Panel producers must be careful not to mix other EBR relationships with panel uses.  For example, a panel producer should not contact individuals using email information from third-party (client) product registration cards for marketing research purposes—unless they gave their express permission for an entity other than the client to contact them for that purpose.

2.   Can panel members be directed to specific sites for the survey questionnaire to be undertaken, or do you require all surveys to be hosted by your organization?

It is advised that all proprietary information collected for panelists must remain private and must not be shared with other companies without the explicit consent of the panelist. Therefore, when online questionnaires are administered by a company other than the one with which the panelist signed up, panelists must first be contacted by the original panel organization (i.e., email invitations must be sent by this organization). However, once invited by the host, panelists can be informed and then routed to the survey of another organization. HOWEVER, it is strongly advised that any organization to which the panelist is re-routed be held to the  same privacy and quality standards as the original panel host. At a minimum, the original panel company should review the subject matter and questionnaire specifications (length, tasks required) to be sure they are of high quality and appropriate for the panelists.

Points for Consideration:

Single site surveys benefit from better control and management and generally yield a more consistent respondent experience. Multi-site surveys, however, can sometimes benefit from technology available only on certain platforms and richer respondent experiences. As one passes data for linking across sites in order to link the respondent record, each entity should ensure the data being transmitted are protected.

3.   If you provide hosting services, what is your simultaneous respondent capacity and average latency?

Simultaneous respondent capacity refers to the number of individuals who can be engaged in survey taking at any particular moment. Capacity can be affected (degraded) by many things, including the length and complexity of the surveys being run at any one time. It is usually expressed as the maximum number of people who can be exchanging data packets with the server per second during some average period of time.

Points for Consideration:

Even if your server has a relatively low respondent capacity (e.g. fewer than 500 server transactions per minute or about 8.3 completed transactions per second) you can control for low capacity by extending the invitation period and not releasing huge numbers of invitations all at once.

Latency is the delay a panelist will experience as questions refresh when taking an interactive online survey—and a major source of survey abandonment. This does not apply to HTML form-based batch surveys that are submitted all at once upon completion. Some of these delays are caused by the users’ browsers and/or Internet connection (e.g. dial-up versus broad-band). Complexity and length of an interview, inclusion of complex activities and whether a dedicated versus a shared server is used also affect capacity and latency-effect experienced by users. Latency is typically reported as a mean lag time over some average period of time and usually requires some external traffic monitoring system to report this characteristic accurately.

Latency is technically calculated as the time it takes from the moment the respondent clicks the submit button to the moment the next page in the survey loads in the respondent’s browser. (i.e., browser->web server->database->web server->browser)

Some normative latency rates (which will change over time) for transaction-based sites are as follows:

  • Poor . . . . . . . . . . . . . . . 5+ seconds
  • Fair . . . . . . . . . . . . . . . .  4-5 seconds
  • Acceptable . . . . . . . . . . 2-4  seconds
  • Good . . . . . . . . . . . . . . . 1-2  seconds
  • Excellent. . . . . . . . . . . . 0-1 second

Recruiting and Panel Intake

4. From where are the respondents sourced and how are they recruited?

There are many ethical ways respondents are recruited for marketing research studies. Here are a few more common approaches:

  • Panelists are pre-recruited via advertising to join a panel specifically for marketing research purposes. Traditional online advertising methods on various high-traffic Web sites, including the use of banners, boxes, buttons, pop- ups, pop-unders and pop-ons, are all commonly used technological methods for getting someone to click on the recruitment ad and join a panel or take a survey.  Other forms of off-line advertising (e.g. newspaper or magazine ads, invitations packed in product packaging, printed materials handed out at retail or events, direct mail, etc.) are also commonly used to reach certain demographic or targeted groups.
  • Panelists are intercepted via a special Web redirect or pop-up to take a survey because they have visited a Web site (or specific sub-site). Some reduction in self-selection bias can be achieved using an intercept algorithm which only displays the invitation to one in every “n” people who hit a specific page.
  • Companies with an established business relationship (EBR) may invite their customers, who may have a reasonable expectation for contact, to take surveys and/or join panels—as long as they identify the source of the relationship.

IMRO also has published some very clear guidelines regarding unethical recruitment techniques.

Surreptitious Sampling can include but is not limited to the following:

  • Collection of respondent emails from Web sites, portals, Usenet or other bulletin board postings without specifically notifying individuals that they are being "recruited" for research purposes;
  • The use of Spambots, Spiders, Sniffers or other "agents" that collect personal information without the respondents' explicit awareness;
  • The purchase of bulk email addresses from sources that have not provided verifiable documentation that the individuals on the list have opted for contact for the purposes of research; and/or
  • The use of client customer lists that have been assembled without the express consent of the individual for future contact via e-mail.

Unsolicited/Unethical Recruitment (Spamming) can include but is not limited to the following:

  • Unsolicited email recruitment of potential respondents;
  • Misleading or off-topic newsgroup postings designed to "trick" a potential respondent into participating in research;
  • Junk mail sent in bulk to recruit for studies or panels;
  • Scamming, which refers to the practice of recruiting under false pretenses (e.g. recruitment for research that is in reality a sales or contribution solicitation pitch);
  • Spoofing, which refers to the practice of putting in a false or missing return email address; and/or
  • Chain or "buddy" letters aimed at recruiting respondents' friends, relatives or colleagues for studies or panels.

Points for Consideration:

SPAM invitations, which are usually the result of buying email lists and doing broadcast, unfocused requests, tend to have very low response rates (e.g., less than 1/10th  of a percent) and are also notorious for poor data quality—purportedly the result of purposeful sabotage in vengeance for the SPAM intrusion. This very low response rate is obviously too low (and skewed) for serious research purposes.

5.   Have members clearly opted-in? If so, was this double opt-in? To what exactly have they been asked to opt-in?

The concept of opting in for participation in online research or panels is central to MRA/IMRO standards. There are several opt-in conditions that are recognized by various organizations.

The most simple is the single opt-in condition, wherein someone clicks on an advertisement or link and proceeds directly to a survey or registration form.  MRA/IMRO recommends one further step for empanelment: “double opt-in”, which is to gather information (profiling data) followed by a second confirmation of the willingness to join. This double opt-in is important to cement the knowledge of joining and secure documentation for further contact under the prior opt-in condition of CAN SPAM.

Along with the opt-in condition, the process and procedure for opting-out at any time should be clearly stated.  In addition, the MRA/IMRO Code of Ethics makes the following disclosures a requirement:

In all cases, the purpose of research conducted by IMRO members will be clearly and accurately stated along with the limitations of the use of personal information gathered. Individuals will have the right to be removed from potential research respondent lists and given a clear and simple way of communicating such a decision. Members will not willingly and knowingly mislead respondents as to the parameters of the research process including length of survey, incentive, use of data, etc.

The panel owner should record and keep secure the details of each respondent’s enrollment. The dates, method and source of invitation to join (such as sampling, website registration) should be kept as a part of the panel information.  Panel providers should periodically confirm opt-in status, especially for those panelists who have not responded to recent invitations to participate.

Points for Consideration:

Some MRA/IMRO members have moved to an even more rigorous process for confirming and documenting the opt-in status. This is known as a “triple opt-in” and entails sending a follow-up email to the registered address requesting confirmation of the registry. This move has been taken to guard against prank sign-ups, and to avoid accusations of spam from the legitimate email owner. This practice also helps to cull out potential cheaters and ensures ample documentation if people do forget their registration.

6.   Do you participate in or sponsor aggregator sites in which respondents can see multiple survey opportunities in one place?

Some sites permit respondents to select from multiple surveys for which they may qualify. Although MRA/IMRO has not taken a specific position on this form of recruitment, it does raise sampling questions which further compound potential criticisms of self-selection in online research.  Self-selected samples are almost never representative of the population and therefore not projectable, although in some instances they can give valuable information and insights to the researcher among certain targeted audiences.

Permitting respondents to select which interview they wish to participate in, may improve response rates, however one must carefully understand the nuances of a self-selected sample (as opposed to “structured recruit” based on pre-assembled profiling data of a panel).

Points for Consideration:

One practice that MRA/IMRO has rejected is the publication of “qualifying hints” in order to beat screeners. This information is sometimes published on sites which offer links to “open surveys” (e.g. not controlled by invitation).  Sites that offer these links and cheating tips will often charge to send people to legitimate panel recruitment sites.  MRA/IMRO advises members to attempt to de-list from these types of aggregator sites if possible.

7.   Do you require review of questionnaires hosted by others?

Panel companies should review all questionnaires to be put before their panelists to be sure they are in compliance with that company’s privacy policy (e.g., the questionnaire should not ask email address, etc.) and standards (usability, competence, civility, etc.).  In order to preserve the relationship and goodwill with their panelists, panel providers should “test” third party programming to be sure the servers have adequate capacity and the survey is programmed properly.

Points for Consideration:

Many research firms attempting to do their own programming have slow servers and cannot handle the load of responses directed to their survey from a panel company.  Many panelists’ complaints revolve around surveys not loading, slow refresh rates, dropped connections, etc.  Universal testing is recommended because each survey is different—one which has many look-ups, calculated variables or error-test verifications may be much more server intensive than other surveys.  It is recommended that panel providers ask to test their non-hosted surveys and decline to send sample to slow or unreliable servers.

Disclosure and Privacy

8. Do you inform the respondent who is conducting the research (either agency or end-user) always, sometimes or never? Under what circumstances would the respondent be informed?

The identification of the sponsor of the research is not specifically required by CAN SPAM, but it is inferred, and should be considered a best practice element.  Without identifying the sponsor, the contactee is unable to determine whether the contact is legitimate or not. Therefore, even if one or both of the CAN SPAM pre-cursor elements are in place (e.g. either an EBR or an opt-in condition) the recipient cannot be expected to recognize the condition of these elements without the identification of the sender.

Points for Consideration:

  1. Some controversy has been generated as to whether companies can give their client lists to some other entity to contact potential panelists and respondents without revealing the source of the information.  IMRO has traditionally recommended against this procedure (others forbid it), because without establishing an EBR or opt-in condition, the invitation will appear to be spam to the recipient.  While some argue that this is not technically in conflict with CAN-SPAM, the potential for damaging the image of research is clear: by appearing to be spam and casting doubt on all future invitations.
  2. Spam filters or blackhole lists are also triggered by emails appearing to be spam, and researchers should be aware of how these systems work.  First, they operate outside regulatory agency purview, so that even if you are in complete compliance with CAN-SPAM, the filters can still “catch you.”  It is VERY difficult and time consuming to get de-listed, so it is very important not to run afoul in the first place.  Spam filters are also networked; meaning that when one picks you up, they share the information with both client- side servers and with other filters—making removal much harder the longer you wait.
  3. No “spoofing” is another CAN-SPAM requirement that is also used by spam filter and black hole listing agencies. Spoofing is using what “appears” to be a legitimate return email address (sender) but that in fact either re-directs or goes nowhere. It is therefore very important that if you are claiming a relationship via a client, that you not attempt to send emails spoofing their email system.  When a spam filter operator accuses you of SPAM, this is one of the first things they look for.
  4. Another technical issue that can indicate a problem to a spam filter monitor is when the DNS address your servers are broadcasting does not match the return DNS route.

Researchers should let their IT department know that any mismatches, re-routing or similar distortions can add “evidence” that your server is being used to pump out spam.

9. Do you inform the respondent who to contact in case they have questions, wish to amend responses or opt-out always, sometimes or never? Under what circumstances would the respondent be informed?

In order to comply with CAN SPAM, all email invitations must include a physical address, a working telephone number and a clear procedure by which people can opt-out from further contact.  Having this information has also been shown to add to people’s comfort level that the invitation is legitimate.  If this doesn’t appear, or doesn’t work, then sending IP address will receive more scrutiny by spam filter systems.

The only known legal requirement for allowing respondents to amend answers is if you are collecting financial data that could affect their credit ratings. Since this is rarely done as a part of survey research, MRA/IMRO has not taken a position on this issue.

10. Is there a Privacy Policy in place? If so, what does it state?

A published privacy policy, in compliance with industry guidelines and Federal and State law, must be easily available to prospective and current panelists and to clients. This information should be accessible both on the company or panel Web site and contained within or linked to within the invitation. Committing to a widely known and respected national or international set of standards (such as adopting the MRA/IMRO Code of Ethics) will help legitimize privacy, confidentiality and ethical standards.

11. With which regional, national and local laws with respect to privacy, data protection and research with children, e.g. EU Safe Harbor, and COPPA in the US, is the panel compliant? How is compliance audited?

Part of the MRA/IMRO Code of Ethics states: “IMRO members pledge to comply with the local, national and regional laws

and regulations regarding the use of all modes of data collection including both interactive and traditional channels of communication in the countries where research is being performed.” This includes not just privacy, but also spam, solicitation, and other regulations as well.

Panels that operate outside of the U.S. but do not control survey hosting and data aggregation may offer or require their clients to offer compliance with Safe Harbor guidelines. Being a registered Safe Harbor company, indicates the researcher’s willingness to comply with the U.S. Department of Commerce program bridging U.S. and European Union Data Privacy Protection Act statutes.

Internet-based research collecting data from children must comply with the Children’s Online Privacy Protection Act (COPPA), which forbids the collection of personally identifiable data from children under the age of 13, without explicit parental approval.  Most panels do not accept registries in any panel of persons below this age. In order to avoid complications with compliance, many researchers send invitations only to parents, who must evaluate the invitation to participate and who then, in turn, must approve their child’s participation.

Those who work in healthcare and pharmaceuticals research must comply with the Health Insurance Portability and Accountability Act (HIPAA) by conforming to secure data collection and storage standards and only reporting data in the aggregate.

Data Security

12. What measures do you take to ensure panelist identifiable information is secure?

Minimum measures that should be taken include password protection for the panelist database and adequate firewall and server protection against those outside the firm who may wish to gain access to servers. IMRO’s code of ethics states:

Rights of Confidentiality

By default and design, confidentiality shall be granted for all information collected from customers and individuals and will be used for the clearly stated and intended purposes only. All personal data will be secured against access by third parties and/or unauthorized individuals or organizations.

13. If survey hosting services are included, what measures do you take to ensure that responses, stimuli and client-confidential information are secure?

Minimum security measures that should be taken include firewall protection, security check-ups by third party services, and having a database server for data collected that is separate from the Internet survey server, which reduces the threat from hacking.

Points for Consideration:

Survey hosts may consider options for content protection (e.g. disabling image capture for images) where appropriate.

Panel Management

14. What is your panel size? What is the percentage of active members and how is “active” defined?

The size of a panel should be based on the actual number of unique individuals who have opted-in for research.  Multiple  qualification counting (e.g. classifying the same person for multiple sub-panels) should not be used to inflate the total size of the panel.  Estimated reach counting (e.g. basing panel

size on shared email addresses or common access to a single computer) should not be used to calculate the total size of the panel.

Points for Consideration:

If “river-type sampling” is employed (placing ads or invitations on high traffic Web sites or publications), the number of people who might possibly be reached should be referred to as “recruitment reach” or “recruitment pool” rather than stated as “panel size.”

The proportion of “active members” is defined as: The average daily number of active panelists (engaged, responding panelists who are eligible to be surveyed given category or participation exclusion restrictions). This is a meaningful measure of true panel size.  Including inactive or semi-active panelists may give a false or misleading impression of panel size.  Bad addresses, and inactive members should be deleted from the panel count on a frequent basis (at least once every three months).

Points for Consideration:

The absolute size of a panel is not a meaningful proxy for aspects of the panel that are typically important to buyers, such as ability to target subgroups, response rate or data quality.

15. Are panel members who have recently participated in a survey on the same subject (“category exclusion”) automatically excluded from a new sample, excluded only by request or never excluded? If they are automatically excluded, for how long or using what rules?

Response rates and data quality may suffer if panelists are offered repeated opportunities to complete interviews on the same topic. Panelists may show signs of conditioning (repeated interviewing influencing subsequent opinions).

These concerns can be mitigated by ensuring the panelists are provided an appropriate number of invitations and including a mix of topics.

Points for Consideration:

Providing a mix of different subject matter opportunities, and/or by creating topical sub-panels so that panelists are explicitly recruited as subject matter experts can help response rates and prevent either over use or serial disqualifications.

16. What restrictions, if any, are placed on how often individual members can be contacted for market research in a given time period (“participation limits”)?

Response rates and data quality may suffer if panelists are “over-surveyed.” Likewise, a panel’s reputation for providing an abundance of survey opportunities could attract panelists whose primary motivation is financial incentives. These concerns can be mitigated by ensuring that panelists are provided an appropriate number of invitations and a clear path to voluntary exclusion.

MRA/IMRO recommends that panelists be offered the opportunity to suggest how many times they wish to participate in surveys during the registration process and that they are invited no more frequently than that limit.  In addition, to avoid conditioning, we recommend that panelists take part in not more than two surveys every month.

Conversely, it has been shown that panelists and members who are not surveyed frequently enough tend to lose contact with the panel and, if contacted after a protracted period of time are 1) less likely to respond and 2) more likely to report the contact as being spam. As a general statement of best practices, panelists should have some form of contact (not necessarily a survey invitation) at least once every three months.

17. During the past 6 months, what was the average number of active panelists who had not reached participation limits or category exclusion available for surveys?

Panel providers should be prepared to provide information on the response rates to surveys and to state how many panelists are “available” for interviewing (e.g. have not exceeded their target participation rate.) Because this number will largely depend on the target population and category of interviews being sought, this number should be provided on an ad hoc basis, given the current panel pool status at any one particular time. Ideally, some indication of when additional sample is likely to be available (after the target participation time black-out period is over) should be available.

18. Is there routine mandatory attrition from the panel, and if so on what basis (time or level of participation)? What is the panel’s annual “churn rate,” that is, the percentage of all panelists who are voluntarily or involuntarily attrited and replaced each year?

Certain, highly specialized panels, can benefit from periodic mandatory attrition. The preferred way of managing attrition is to slowly decrease the number of invitations over time rather than “disinviting” the member. Announcing that a member is no longer being invited to participate in surveys produces hard feelings and can create a generally negative attitude toward research in general and should be avoided if possible.

In more general panels, natural attrition of about 25% to 30% per year tends to occur. This natural attrition is also referred to as the “churn rate.” Most of these members do not “quit,” per se, but rather become “inactive” over time.  Inactive members (those who do not respond to invites) should be tracked and eventually delisted from the active panel.  Once again, delisting should be a non-confrontational action that does not provoke ill will.

Points for Consideration:

A number of different protocols for identifying “inactive members” have been suggested.  One is to track the number of invitations that have failed to produce any response and after a certain number, (perhaps four to six), the panelist should be contacted to determine whether they wish to continue to receive invitations. Another is to track the period of time during which they have participated (perhaps nine months to one year), and then follow up with a similar inquiry.

19. How and how frequently do you replenish the panel?

All panels need to be replenished over time.  Panel providers should be prepared to discuss how and when panel membership numbers are supplemented and whether there is a continuous or periodic recruitment effort.  Best practices dictate that in order to maintain integrity, panels must be replenished at least once a year.

20. What is the incentive structure and fulfillment process?

Almost all online panel work involves some form of incentive or member acknowledgement for participation. The ability to offer a variety of incentive structures (e.g., drawings for prizes vs. per-panelist cash incentives vs. point accumulation) may be critical to satisfy the needs of different clients and panelist types.  In general, incentive rates follow recruitment difficulty: easy to find respondents generally are offered less to participate; very difficult or non-responsive groups require more. Panel providers should be prepared to provide information on the amount of incentive being given; specifically as a component broken out from the overall cost per interview (CPI.)

Points for Consideration:

Incentive structures that have no physical fulfillment (e.g., gift certificates delivered via email) are time- and cost-efficient but require that panel providers find other means for physical address verification. Some MRA/IMRO members recommend that every fourth or fifth incentive offered be done in the form of a check or some other vehicle that requires a physical mailing address.

Screening and Quota Management

21. What panel profiling information is kept on members? How and how often are profiles updated? In what other ways can users be profiled (e.g., responses to ad hoc surveys)?

MRA/IMRO suggests the following information would be helpful in classifying and identifying respondents and avoiding duplication in consumer samples: Gender, Level of Education, Household Size, Region, Location (postal code + house number), Birthdate (so that age can be calculated at any point in time), Children in Household, Working Status, Hours of Internet Usage, and Type of Internet Access.  Business-to- business information should include: Individual’s Job Function/ Title, Type of Business, Level and Type of Decision-making Authority, Size of Business (Number of Employees or Annual Revenues), and Geographic Scope of Operations.

Such information should be refreshed on a regular basis (every 6-12 months), and panelists should be urged to provide updated information on a continuous basis.  Non-participating panel members should be contacted and asked to verify their desire to participate (those who have not responded to six consecutive survey invitations). If non-participation continues (for more than nine month to one year), the panelist should be delisted. Verification of data by re-asking some of the questions in a regular survey can also improve data quality by dropping those who answer inconsistently. Also, some background information can be added to the profiling information from the regular surveys.

Points for Consideration:

If the panel owner is adding information about the panelist from ongoing survey work, it is considered unethical to use proprietary survey question information to pre-qualify a panelist for some other client’s research work.  Once again referring to MRA/IMRO’s code of ethics:

Rights of Confidentiality

By default and design, confidentiality shall be granted for all information collected from customers and individuals and will be used for the clearly stated and intended purposes only. All personal data will be secured against access by third parties and/or unauthorized individuals or organizations.

22. Are screening questions included in the survey (i.e., qualification does not rely solely on panel profiling information) in some, all or no surveys?

Few surveys rely only upon profiling data to ensure qualification for participation. Re-screening for important, distinctive, or rapidly changing qualification questions should occur to ensure the correct sample composition. This can also be used as an additional internal check against previously collected profiling data.

23. What quota management or targeting system is in place to prevent serial screening disqualification of respondents and over-surveying with certain group(s)?

Once a sampling plan for a given study has been set, panel profiling information can be used to target sample invitations to reach those with a high likelihood of qualifying.  If the response rates for a quota group are known, a minimum number of  invitations should be sent, in order not to turn away otherwise qualified respondents due to over-quota conditions.

Points for Consideration:

Research has shown that the single greatest “turn off” to online panelists is being terminated multiple times during screening. This effect is multiplied when they apparently qualify, but are then informed that the number of surveys needed is complete.  In order to protect the good will of respondents for all future research, panelists who are turned away due to filled quotas should be thanked for their interest and willingness to participate.  MRA/IMRO suggests some sort of consolation gift or non-qualification drawing for those who have attempted participation, but have, for whatever reason, terminated.

Sample and Data Quality Management

24. What processes for validation of identity are in place at the points of: panel intake; panel profile update; surveying?

Physical address verification at panel intake and during incentive fulfillment is a common and relatively fail-safe method for ensuring that panelists have not created multiple panel profiles in order to harvest invitations and incentives. Other forms of validation, such as profiling phone calls at intake and/ or on a regular basis, can validate the panelist’s demographic and professional characteristics.

At the point of panel intake, the panel owner should, at a minimum, verify that the physical address provided by a respondent is a valid postal address.  In addition, through survey questions and through periodic updates of the panel profile, panel companies should check for consistency in responses concerning demographics and consumer behavior/ preferences. The best methods for validating the identity of a respondent are those where the respondent is dependent on a unique address for redemption of incentive – such as standardized home address. A respondent can have multiple email, PayPal or “home” addresses, but won’t benefit, if the check or prize can only be sent to a valid home address.

Points for Consideration:

Some MRA/IMRO members recommend that every fourth or fifth incentive offered be done in the form of a check or some other vehicle that requires a physical mailing address.

Panel providers should also use IP and browser setting tracking to insure that multiple submissions from the exact same machine aren’t allowed within a certain period of time (usually the length of field for a survey.)

Finally, internal database scans for similar member identification and password combinations can detect multiple registry attempts.

25. How are responses validated so that only those panelists invited to a particular survey may respond and each panelist may only complete (terminate or complete) a survey once? How do you guard against duplicate memberships within the panel?

Since it is difficult to prevent multiple entries by a single respondent in an open survey environment (e.g. one that is simply posted on the Web as an open invitation), MRA/IMRO recommends that panel companies restrict respondents to a single point of entry (such as an invitation with a unique URL code string or a password protected entry page) to create a trail of data artifacts to block out any attempted duplicate entries.

It is also recommended that incentives be structured so that there is no monetary benefit to spoofing the system (i.e., a policy which says “if we catch you cheating you don’t get paid.”) Unique IDs should be set up in advance and sent with invitation to respondent (the panel company should avoid using consecutive IDs, which allow respondents to possibly enter a survey multiple times, blocking out other legitimate participants). Demographic questions should be included in the survey and compared with the panel profile.

Finally, the panel database should be scanned periodically for “identical record matches” on a variety of data points, indicating that multiple submissions for membership may be artificially generated. At a minimum, the panel provider should be checking for duplicate email addresses, should standardize addresses and check for duplicate physical addresses, should be checking for multiple survey responses coming from one IP address, and should scan for duplicate password/address- feature combinations.

26. What was the average response rate across all surveys during the past 6 months, and how is response rate calculated? How are deviations from planned response rate handled (i.e., sending more fresh invitations, reminder invitations)?

Response rate is based on the people who have accepted the invitation to the survey and started to complete the survey. Even if they are disqualified during screening, the attempt qualifies as a response.

Panel companies should have processes in place to monitor response rates. If the expected response rate is not being achieved, the panel company should offer the client multiple strategies and should discuss the impact on the research for compensating for any under-performance due to response rates. For example, the panel company can employ features such as reminders and engaging fresh sample or even top- up sample providers, but the use of these options should be discussed with the client in light of their respective impact on results.

Completion rate (as opposed to response rate) is calculated as the proportion of those who have started, qualified and then completed the survey. If they pause after they start then they go into the “incomplete” disposition and are not counted towards the completion rate.  If those in the “incomplete” range are contacted with a reminder, most panel systems will allow them to pick up where they left off. Panel providers should track and report the number of “incomplete conversions” as a part of the completion rate statistics. At the close of field, those who paused (and became listed as “incomplete”) and who did not return to complete (convert), should be listed as mid-terminates.

Points for Consideration:

  1. Response rate is dependent on a number of variables including incidence and survey salience elements including: relevance of topic, burden of survey, incentive, and affinity with the sponsor of the survey. A panel which has cleaned out non-responders and bad addresses, offers a higher incentive and/or fields a shorter survey will attract a higher response rate.  Of course, panels can’t always fix low response rates caused by unknown or unavoidable situations (such as fielding on Superbowl Weekend).
  2. Occasionally, response rates are impacted by client overestimations of incidence rates (the true proportion of qualifying individuals within a specified population). If incidence rates are not known, it is recommended that a short survey, with small incentive, be offered to check incidence and to pre-qualify respondents, prior to the actual launch of the survey.  If the incidence is overestimated after the launch of the study, and the panel provider needs to send the survey out to many more panelists, it can have a deleterious impact on panel longevity and future respondent cooperation as unnecessarily large numbers of panelists are simply screened out of the survey—a key factor relating to panel dropout rates.

27. Do you place restrictions on survey content, design/ appearance or length, and if so what restrictions? Do you require review of questionnaires hosted by others?

The design, appearance and length of a survey will impact the response rate. Online panel companies should do their best to “guide” clients regarding best practices in usability, questionnaire length based on the topic, target audience, survey design/appearance and it’s appropriateness for online methodology. Long surveys ( 0 to   0 minutes) can be accomplished, but generally require large incentives and lead to panelist fatigue and shorter panelist life.

Points for Consideration:

Some MRA/IMRO members have experimented with animations, Flash technology, graphic content and other elements to increase the level of engagement. While these may temporarily help with completion rates, if they are overused or are more time consuming, the opposite effect may occur.  In all cases, keeping a “clean” interface—one devoid of unnecessary graphic elements, animations, instructions, images, etc.—will help from distracting the user from concentrating on the survey questions themselves.

28. Is any allowance over total quota routinely made in order to allow for cases to be deleted in data cleaning? Are clients charged for this quota?

A panel provider should commit to deliver the number of completes for which it has contracted.  If the correct real- time screening protocols are in place to catch 1) attempted multiple survey submissions, 2) to detect pattern replication (indicating satisficing behavior: doing just enough to get paid for the incentive) and   ) to detect logical inconsistencies within the data; then no over sampling should be required or charged.  If there are no such protocols in place, then the panel provider may have to either oversample (at their own expense) or provide replacement interviews in order to meet their commitment.

If there are “outliers” within the data, which may legitimately occur within any sampling method, these should be removed after the fact at the discretion of the client working with the analysis. As long as the total number of rejected interviews does not exceed 1% to   % of the total sample, and if all the screening criteria are met, the panel provider should not be held accountable for interviews that are not used at the discretion of the analyst.

29. If other sample sources are used to fill a study (other panels, lists, etc.) what process is in place to safeguard against duplicate responses across the panels?

Sometimes panel companies are unable to achieve the desired sample size using their own membership, so they need to use other partner panel or sources.  Panel providers should disclose their role as a broker if additional sample is needed.  In addition, clients should inquire as to how they plan to ensure that there are no duplicate respondents across the two or more sample sources, including what de-duplication procedures are in place.

30.  What processes are there to guard against satisficing responses (respondents not concentrating/caring in their responses)? What measures are taken for panelists who are identified as frequent satisficers across multiple surveys?

Overall, the number of online surveys that are discarded because the answers are not consistent or contain a high proportion of “outlying” answers should be no more than 1% to 2%. This is about half as much as you would expect to find in a traditional mail-based survey.

In general there are four types of “cheating or satisficing behaviors” that should be screened for in real time.  Note: Although it is possible to do this screening post hoc, this will create a liability for the panel provider who has contracted to produce a certain number of “clean interviews.”

  1. Flatlining, Straightlining or Replicated Answer Patterns: This behavior is demonstrated by people rushing through the survey giving little or no thought to the responses. Typically the survey should contain pattern recognition algorithms to detect a series of the same answers (e.g. all “  s” on a matrix question, or a replicated zig-zag pattern). Caution should be applied in tossing interviews for pattern replication unless there is little or no chance that the pattern can not represent a legitimate opinion.
  2. Rushed Answers: Respondents who take a survey at much faster than normal rate are probably not paying close enough attention to answers. A mean time to complete should be calculated during pre-testing. Anyone who completes at less than half this time should be considered suspect. Note: when tracking speed in real time, be sure to use the mode rather than the mean in terms of time standards. This is because respondent may be interrupted and finally complete hours later radically increasing the mean time to complete.
  3. Illogical and Unreasoning Answers: Another problem related to flatlining is related to not taking the time to read questions correctly. But because randomly answering questions can escape a pattern recognition program, additional tests are advisable. The degree to which people are paying attention can be tracked by performing tests on answers that should logically be very different.  If, for instance, someone rates something as “too expensive,” they should not also rate it as being “too cheap.” This type of answer matching must be done on an ad hoc basis and instigated by the survey designer.
  4. Automated Survey Responses: Some potential survey cheaters use automated “keystroke replicators” or “field populators,” which attempt to recreate many duplicate surveys using a single survey as a template.  Panel providers should ensure that technological blocks are in place to detect and defeat these mechanisms prior to admission to the survey itself.

Points for Consideration:

Unlike phone and in-person interviewing, online research is self- administered.  Whenever the survey is administered without a proctor, some level of error is more likely to occur.  However, it is more likely that people will be lazy (straight-lining answers, for example) rather than being outright dishonest.  Pattern recognition and real-time error checking can be employed to avoid this type of bad survey behavior.

With regard to “honesty,” there is a good deal of literature that deals with varying degrees of veracity and candor depending on the level of anonymity in the survey environment. There have been studies done that suggest that many groups tend to be MORE honest (particularly with regard to socially sensitive questions) when the survey is administered in the relative anonymity of a computer-mediated environment.

Online attitudes can also be affected by poor question structure, off-topic questions or questionnaires that run too long.  In any of these cases, respondents may, without fear of offending a live interviewer, give flippant or irreverent answers.  It is, therefore, very important to make sure your surveys are intelligent, usable and on-target to avoid cynical responses or mid-terminated interviews.