Methodology & Accuracy

Our unique approach—Dynamic Online Sampling—has become the new gold standard in opinion research. We combine the best elements of different methods.

More voices. More insight.

We recruit fresh participants online for each and every poll, so Embold Research polls are not affected by the dwindling response rates of phone polling. We use proprietary and patent-pending approaches to recruit participants, and we don’t rely on online panels of habitual survey takers. We reach millennials and seniors, rural and urban dwellers, and members of every gender, race, creed, and political persuasion. In 2020, our methodology was more accurate than any other.

Target the right respondents—fast

Targeted respondents will see a solicitation with a generic civic symbol of the region or district that prompts them to participate in a survey. Once a respondent clicks on the image, they are directed to a Change poll, which has the look and feel of a typical online survey with multiple choice and open-ended questions. We generally do not offer financial incentives to respondents for participation, unlike online panel providers, as incentives can affect the accuracy of the results.

Dynamically adjust

As respondents answer the survey, our technology looks at the reported demographics of the sample, and it adjusts our advertising targeting in real time. (For example, if it sees that we have a lot older respondents, it will adjust targeting to reach more younger people). This approach delivers a cross-section of respondents that reflects the diverse makeup of a specific community. Any imbalances in the survey sample can be quickly detected and corrected while the survey is still fielding, rather than relying solely on post-survey techniques for big corrections.

Apply best practices in weighting

Once all of the responses are in, our technology takes over, quickly and accurately performing automated post-stratification weighting to ensure that all segments of your population are properly represented in the poll.

Polling Accuracy

Embold Research is the nonpartisan, nonpolitical division of Change Research. We use the same accurate methodology and tech. Because electoral outcomes provide a uniquely objective benchmark of accuracy, the data that follows focuses on our methodology’s track record in political polling.

Dynamic Online Sampling

2020 Accuracy

Dynamic Online Sampling outperformed other polling methodologies in 2020

In 2020, our private and public polls were nearly 25% more accurate than other polls in battleground states—with an average absolute error of 3.5 points in the presidential race across all polls in Arizona, Florida, Michigan, North Carolina, Pennsylvania, & Wisconsin in the two weeks before November 3rd. Overall, live phone polling fell short: public phone polls were 23% less accurate than polls conducted online and 48% less accurate than Change Research’s polls.1

Change Research’s Presidential Model Was More Accurate Than 538 and The Economist

Change Research models incorporated individual-level data from hundreds of thousands of Change Research’s individual survey respondents and did not rely on aggregates from other pollsters.

1Based on an analysis conducted by Change Research which compared Change Research’s public and private polling in six key battleground states in the two weeks before Election Day with all publicly available polls on 538 conducted in the same geographies and time period.

Embold Research is the nonpartisan, nonpolitical division of Change Research. Our methodology and polling have been featured in: