By the Wavefield Research team · Published Nov 5, 2026
A panel survey interviews the same group of respondents repeatedly over time, in waves, so you can see who changed rather than just whether the total moved. The same phrase also gets used for any survey fielded to an online survey panel. This guide covers both meanings, and the discipline the first one demands.
The design is the most powerful instrument in survey research and the most expensive to run honestly. The pages ranking for this term define it and stop. The useful part is the arithmetic underneath: how fast a panel shrinks, how answering changes the answerers, and what the institutions that run panels for a living do about both.
Half the confusion around this topic is that the research industry uses nearly identical words for two unrelated ideas. A panel survey is a design: the same respondents, re-interviewed across waves, answers linked person by person. A survey panel is a sample source: a pool of people who agreed to take surveys, from which any study, including a one-off, can draw respondents. Vendor pages blur the two because panel companies sell the second thing under the prestige of the first.
The test is linkage. If wave two can be joined to wave one at the level of the individual respondent, you are running a panel survey. If each study draws whoever is available from the pool, you are running one-off studies on a survey panel, which is a perfectly good thing to do and a different thing. How those pools are built, priced, and vetted is its own aisle, covered in market research recruitment.
A cross-sectional survey photographs the population once. Repeating it with fresh samples re-photographs the crowd. A panel films named individuals, and three kinds of question yield only to the film:
A tracker built on fresh samples can report awareness up four points. A panel survey can report that the gain came from lapsed buyers returning rather than new buyers arriving, because the same people answered both waves. When the churn is the question, only a panel sees it.
Because measurements are dated per person, a panel can show that the price complaint came before the cancellation, not just that complainers cancel more. That is still short of proof, but temporal order is the one ingredient of a causal story a one-shot survey can never supply.
By wave four you know each panelist's history, so a two-minute questionnaire can do the analytical work of a twenty-minute one. Long-running panels answer new questions against years of stored profile, which is why institutions keep them alive for decades.
If none of your questions are on that list, stop here and save the money. A repeated cross-section, the same instrument re-fielded to fresh samples, tracks a population at a fraction of the cost and with none of the problems in the next section. That is what most commercial trackers actually are, whatever the sales deck calls them.
None of these are failures of execution. They are structural costs of the design, present in every panel study ever run, and the reason a panel has to earn its keep against the questions above:
Some share of panelists drops out every wave, and the losses multiply. Keep 80% each wave, a respectable rate, and by wave five you are interviewing 41% of the people you started with. Worse, the leavers are not random: the busy and the disengaged go first, which quietly reshapes who the panel represents.
People who answer the same questions repeatedly stop being ordinary respondents. They notice the brand more, form opinions they would not otherwise hold, and learn which answers end the survey faster. The measurement instrument changes the thing it measures, and no weighting scheme can undo that.
The people who stay for twenty waves are disproportionately the ones who enjoy surveys or want the incentives. A seasoned panel drifts toward practiced respondents, which is also why fraud checks matter more over time, not less.
A one-off survey pays for respondents once. A panel pays to recruit, pays incentives every wave, pays to replace leavers, and pays staff to manage retention between waves. The meter runs whether or not you fielded anything this quarter.
The attrition arithmetic deserves one more line, because it is the part that surprises people: retention multiplies, so 90% per wave leaves 66% of the panel after four waves, and 70% per wave leaves 24%. Small differences in per-wave retention become the difference between a panel and a remnant.
The best public documentation of panel management is Pew Research Center's American Trends Panel, a probability-based panel of roughly 10,000 U.S. adults running since 2014. Pew recruits new panelists every year, stating the reason plainly: some panelists stop taking surveys, so new people must replace them to maintain the panel's size. Panelists who skip the annual profile survey are no longer invited at all, and Pew occasionally retires panelists from overrepresented groups rather than weighting them ever more heavily. Weighting, in turn, explicitly corrects for panel attrition. Every mechanism maps to a tax from the previous section, which is the point: a well-run panel is a machine for managing decay.
Scaled down to a commercial program, the same playbook becomes four rules:
A trend is only real if the instrument held still. Freeze the core questions at wave one and version anything you change, because a reworded question starts a new trend line no matter what the chart implies.
Serious panels recruit replacements on a schedule rather than letting the panel age in place. New joiners also provide a control group: if veterans and fresh recruits answer the same question differently, you are looking at conditioning.
A wave-five finding from 41% of the original panel is a different claim than a wave-two finding from 85%. Put the retention rate beside the trend line so readers can weigh one against the other.
Attrition skews composition, so each wave gets weighted back to the population targets, and the report states the effective base after weighting. This repairs the shares; it cannot repair what conditioning did to the answers themselves.
The mechanics behind two of those rules live in their own guides: how raking pulls each wave back to population targets in survey weighting, and what a given wave's sample can resolve in the sample size calculator. Inside companies, the commonest panel-shaped instrument is the employee pulse survey, which inherits every rule above, anonymity floors included.
Most teams searching this term do not need a longitudinal design. They need respondents this week, and the survey panel aisle is how you buy them. The aisle splits in two, and Princeton's Survey Research Center guide to online panel services draws the line cleanly: recruited panels, built so members have a known probability of selection, and opt-in panels, whose members volunteer through self-selection. Its one-sentence economics holds for the whole industry: opt-in samples are less expensive and less representative than samples from recruited panels.
Commercial research runs overwhelmingly on the opt-in side, managed with quotas and weighting, and that is the aisle Wavefield operates in: vetted opt-in panel sample from $3.50 per quality-checked complete, with a fraud sweep that flags suspect responses for review rather than silently deleting them. The honest trade is stated in the price. A probability panel like Pew's costs institutional money and earns institutional defensibility; an opt-in quota sample costs a few dollars a complete and answers most commercial questions well, provided the report says which one it used.
A panel survey is a longitudinal design that interviews the same group of respondents repeatedly over time, in waves, with the wording held constant. Because each person's answers are linked across waves, it can measure individual-level change: who switched, who lapsed, what preceded what. The same phrase is also used loosely for any survey fielded to an online survey panel, which is a sampling arrangement rather than a design.
Pew Research Center's American Trends Panel is a working example: roughly 10,000 U.S. adults, recruited by random address-based sampling, who take one or two surveys a month, with new panelists added every year to replace those who leave. Commercial panel survey examples include consumer purchase panels and customer communities re-surveyed quarterly. Many brand trackers that look like panels are actually repeated cross-sections, fresh samples each wave.
Three structural ones. Attrition: panelists drop out every wave, the losses compound, and the leavers differ from the stayers, which skews later waves. Conditioning: answering the same questions repeatedly changes how people answer, and sometimes how they behave. Cost: recruitment, per-wave incentives, replacement recruiting, and retention management run continuously. Most commercial questions do not need individual-level change, which is why repeated cross-sections dominate practice.
A cross-sectional survey measures a fresh sample once: a snapshot. A panel survey re-interviews the same people across waves: a film of named individuals. The practical middle option is the repeated cross-section, which re-fields the same questionnaire to fresh samples and tracks the population without attrition or conditioning, at the price of never seeing which individuals moved.
The loss of panelists between waves, through fatigue, relocation, or lost interest. It compounds: at 80% wave-over-wave retention, under half the original panel remains by wave five. Because leavers differ systematically from stayers, attrition biases later waves as well as shrinking them. Panels manage it with incentives, refreshment cohorts of new recruits, and weighting each wave back to population targets.
No, and the vocabulary trips people constantly. A survey panel is a pool of people who agreed to take surveys: a sample source you can buy from for any study, including a one-off. A panel survey is a research design that follows the same respondents across waves. You can run a panel survey on a survey panel, but most surveys fielded to panels are one-off studies with no second wave.
Related: cross-sectional surveys · brand tracking software · market research recruitment · survey response rates
A repeated cross-section tracker gives you the trend line without the attrition: $299 to stand up, $99 a wave, panel sample from $3.50 per quality-checked complete.
Start a study