A winning methodology.

Change Research has one of the strongest accuracy records in the industry. Here's the methodology behind that record, and what it means for your research.

Our Track Record

Accuracy you can verify

2022–24
3.3pt
Final margin, on avg
86%
Within 6 points
All-time
4,000+
Polls conducted
50
States polled (+ DC and PR)
Accuracy Reports by Year
Our Methodology

Dynamic Online Sampling: why it works

Change Research pioneered Dynamic Online Sampling to fix what's broken in traditional polling. Instead of relying on respondents picking up their landlines, we recruit respondents through a combination of targeted digital ads, text messaging, and selective panels — reaching voters where they actually are, including segments of the population often underrepresented in polling.

As responses come in, we use our Dynamic Online Sampling methodology to continuously adjust targeting across demographics including age, gender, race, and partisanship to ensure the sample accurately reflects the electorate or population. Any imbalances are rapidly detected and corrected while the survey is still fielding, not patched after the fact. The result is a sample that reflects the full population: Gen Z and seniors, rural and urban dwellers, members of every gender, race, creed, and political persuasion.

Traditional Phone Polling

The old way

  • Dependent on a shrinking pool of people who answer calls from unknown numbers
  • Can't reach small geographies and demographic subgroups
  • Corrections after fielding delay final results
  • High cost puts frequency out of reach for most campaigns and organizations
Dynamic Online Sampling

Change Research's way

  • Representative samples through targeted recruitment
  • Real-time adjustments while the survey is still fielding
  • Cost-effective, enabling more frequent polling
  • Reach small geographies and hard-to-reach populations
Accuracy counts

Our track record of calling it right

  • Presidential polling

    Cycle(s)
    2024
    Scope
    17 polls
    Accuracy
    Within 2.6 pts of the final margin*

  • U.S. Senate & congressional

    Cycle(s)
    2022 + 2024
    Scope
    38 polls
    Accuracy
    Within 3.6 pts of the final margin*

  • State legislative

    Cycle(s)
    2022 + 2024
    Scope
    78 polls
    Accuracy
    Within 3.0 pts of the final margin*

  • San Francisco mayoral

    Cycle(s)
    2024
    Scope
    Citywide
    Accuracy
    Within 2 pts of the final result

  • Chicago aldermanic & mayoral

    Cycle(s)
    2023
    Scope
    9 wards + citywide
    Accuracy
    Within 5.4 pts of the final margin*

  • #1 media pollster

    Cycle(s)
    2020
    Scope
    Battleground states
    Accuracy
    Our polling for CNBC was the most accurate of 17 media pollsters, within 4.5 pts of the final margin

* on average, across polls in the final 30 days.

We trust Change Research when we need to move quickly and when accurate results are critical. We need accurate data in-hand to make high-stakes investment decisions. In 2024, we were one of the chambers in the country to pick up Democratic seats, and Change helped us have the information that we needed that helped make that happen.
Claire Krafka
Claire Krafka
Executive Director, Ohio House Democratic Caucus
Frequently Asked Questions

Methodology & Accuracy

How does Dynamic Online Sampling work?

We recruit representative survey respondents through a combination of channels — including digital ads, text messages, and selective panels — rather than phone calls. As responses come in, Change Research adjusts targeting in real time to correct any imbalances while the survey is still in the field, rather than relying solely on post-survey weighting.

How do you ensure representative samples?

We use a multi-layered approach. Targeted recruitment across digital ads, text messaging, and selective panels ensures broad demographic reach from the start. Critically, we use Dynamic Online Sampling to monitor the sample as it fields and correct any imbalances in real time — so we’re not dependent on large post-survey adjustments to fix representation gaps.

After fielding, we weight responses to ensure that the survey sample is representative across demographics, geography, partisanship, and more. For likely voter populations, we use sophisticated turnout models to estimate the makeup of the electorate.

What's your accuracy track record?

In the last 30 days of the 2022 and 2024 elections, we predicted the winner’s final margin within 3.3 points on average across 141 partisan races spanning President, Governor, U.S. Senate, U.S. House, State Senate, and State House. We’ve conducted over 4,000 polls across all 50 states plus D.C. and Puerto Rico, consistently outperforming traditional phone polling in accuracy — a track record we publish and stand behind.

How does Dynamic Online Sampling polling compare to phone polling?

Dynamic Online Sampling delivers results faster, and as phone response rates have declined sharply, the most sophisticated online methods now produce more representative samples. There are also other methodological advantages: respondents answering online are less likely to give socially desirable answers that don’t reflect their actual views — particularly on sensitive issues like race or abortion. Together, these factors explain why our accuracy record speaks for itself.

Can I see the methodology behind my specific poll?

Yes. Every Change Research poll includes a detailed methodology section covering sample size, fielding dates, weighting methodology, and margin of error. Clients also receive access to full crosstabs and methodology documentation through the Change Research Data Portal.

See how our methodology uncovers breakthrough insights

Tell us about your race or issue, and we’ll walk you through it.

Get in touchSee our results