Collect
Recruit deliberately, ask neutrally, and record the conditions under which every response was given.
Methodology standarddemosight-bias-control-1.0 · Effective 2026-08-29
The proof behind the pulse
No magic
numbers.
A result is only as credible as the chain of evidence behind it. Demosight makes that chain inspectable—from collection through interpretation.
Correct
Detect low-quality activity, disclose exclusions, and calibrate the achieved sample to defensible benchmarks.
Understand
Estimate uncertainty, test subgroup differences, separate association from causation, and track change consistently.
Explain
Translate evidence without false precision, publish limitations, and link every conclusion back to the underlying method.
Population boundary
Place, residence, and identity are not interchangeable.
Adults currently living in the United States, with U.S. geography and benchmark definitions. It is never blended into Philippine estimates.
Adults currently living in the Philippines. Only this frame can support a Philippine resident-population estimate.
Current land-based, sea-based, or between-contract overseas workers, with destinations and work context defined in advance.
Long-term residents, citizens, students, families, and later generations abroad under an explicit eligibility definition.
Frames may be compared only when the study predeclares the populations and discloses how each was recruited, benchmarked, weighted, and limited.
Permanent operating rule
Designed to reduce bias. Never marketed as bias-free.
No survey method can guarantee zero bias. Demosight instead makes each known source of error visible, measurable where possible, and capable of blocking a public claim.
- 01
Define the target population before recruitment or results review.
- 02
Separate residence, citizenship, and diaspora identity instead of treating them as substitutes.
- 03
Monitor predeclared quota cells during fieldwork and retain recruitment-source metadata.
- 04
Use approved population or electorate benchmarks with source, universe, date, and version recorded.
- 05
Publish unweighted sample size, weighted estimates, effective sample size, design effect, trimming, exclusions, and limitations.
- 06
Keep participant identity outside research reporting and suppress result cells that do not meet the predeclared privacy minimum.
- 07
Block official release when coverage, quality, weighting, or independent-review gates are unresolved.
Methodology Passport · Preview 001
Every number carries its own proof.
This is the live disclosure record for the founding Pulse of the Day. Official studies will add questionnaires, translations, weighting diagnostics, design effects, effective sample sizes, revision history, and downloadable documentation.
SponsorDemosight founding product preview
PopulationConsenting adult participants in the founding preview
Sampling frameOpen participation; not a probability sample
Field modeSelf-administered mobile and web questionnaire
WeightingNone for live participation totals
Quality controlsOne pilot entry per participant, duplicate prevention, server-side validation
DisclosureExact wording, field status, sample definition, exclusions, limitations
Key limitationPreview totals cannot be generalized to the U.S. or Philippine population
Response integrity
Quality rules are set before the results are seen.
Identity & duplication
The preview enforces one response per consenting pilot participant. Production studies add declared recruitment-source, identity, and duplicate-risk controls without scoring political views.
Behavioral review
The preview records duration, selection changes, and focus losses. Multi-item studies add predeclared attention, item-response, straight-line, and cross-item coherence rules.
Weight diagnostics
Population claims are blocked until approved benchmarks, weight distributions, trimming, design effects, effective sample sizes, and sensitivity checks exist.
Interpretation guardrails
Participation totals stay labeled as such; statistical significance is not treated as practical importance; associations are never silently rewritten as causes.
We will publish misses, corrections, and revisions—not just flattering results.
Open the research registryMethodological basis