A real poll of real voters.

What every poll includes

• Questionnaire design and review, with the electorate, sample and fielding plan agreed before launch.
• A sample drawn from the registered-voter file, with available vote history and phone matching.
• Text-to-web fielding, with a text-plus-phone option scoped in your quote.
• Weighted toplines, crosstabs, a written analysis memo and a 45-minute briefing. Small or unreliable cells are flagged or withheld.
• A methodology disclosure and the analysis files you keep.

Up to 15 minutes or 20 questions, demographics included. Plan on 5 to 7 business days after questionnaire approval; we confirm feasibility, coverage and timing in your quote.

Methods you can inspect

Registered-voter primary. For new general-election polls, the registered-voter estimate is the primary result, calibrated by bounded raking to the poll’s modeled electorate composition. The likely-voter estimate is a separately labeled sensitivity: its model did not pass pre-registered validation and is not presented as a proven correction. Primary-election polls have a separately specified electorate and analysis plan; historical releases retain their original methods.

AAPOR disclosure. Release disclosures state the sponsor and conductor, sample frame, population, collection mode, field dates, question wording, weighting, effective sample size, design effect and limitations. The public appendices include AAPOR outcome rates and disclose missing records. This is disclosure under AAPOR’s standards, not a claim of AAPOR certification.

Shadow checks beside the headline. Pre-registered diagnostic comparisons examine data defects, recalled vote, education, post-ballot partial responses, likely-voter modeling and undecided allocation where applicable. Their status stays visible; an experimental result does not silently replace the primary estimate. The two historical primary releases mark these general-election shadow checks not applicable.

Inspect a release, its disclosure and methodology appendix →

Our record, without a victory lap

Our public Delaware release contains 1,160 completed interviews; the Rhode Island post-election release contains 1,412. These are the samples in those frozen releases, not a promise about your poll.

As checked October 5, 2026, the public accuracy record has 1 scored race. The Delaware U.S. Senate Democratic primary poll missed the certified top-two lead by 9.2 percentage points, toward Jeff Appelhans. One race is not enough for the record’s cross-pollster ranking, and we are not claiming a validated accuracy advantage.

Rhode Island does not count in the prospective record: its initial public result appeared after polls opened, and the later release is labeled a post-election reanalysis. Retrospective analysis never becomes a pre-election prediction.

The live record compares published top-two leads with certified results, alongside other public polls of the same race where available, using one stated measure.

See the live public accuracy record → · Delaware release → · Certified Delaware result → · Rhode Island post-hoc release →