Every serious finding should come with receipts.
A number by itself asks to be believed. A number with a record asks to be checked. This page shows exactly what a PolitiMind result can prove about how it was produced, what it cannot prove, and which parts of that proof are available today versus still being built.
The distinction that matters
Integrity and accuracy are different standards, and only one of them can be proven
Accuracy asks whether a finding matches reality. That question is answered by sampling, weighting, and inference, and it always carries uncertainty — no survey method removes it. Integrity asks a narrower and far more answerable question: was this number produced the way its publisher says it was, and has the record changed since?
Legacy research collapses the two. When a result is challenged, the only available defence is the vendor's reputation, because there is nothing an outsider can reconstruct. PolitiMind separates them: uncertainty stays stated and honest, while the production chain becomes inspectable.
Six modules
Integrity is assembled in layers, and each layer states its own limit
Proof is not a single seal at the end of the process. It is six separate assurances, each covering a different failure mode from who was allowed to answer through to whether the published record still matches what was committed. Where a layer is not fully live, it says so.
Published result
- L06
Participant integrity
ProvesThat the people behind a sample were verified as real, unique, and eligible for that study before they could answer.
Does not proveThat the verified sample is demographically representative of any population. That remains a sampling and weighting question.
- L05
Research integrity
ProvesThat the questionnaire was versioned and frozen before fielding, so the published wording is the fielded wording.
Does not proveThat the wording was the best possible wording. Freezing removes silent edits, not design judgement.
- L04
Sample integrity
ProvesThat integrity controls ran against the response set — duplicate, coordinated, and low-quality participation is flagged before a result publishes.
Does not proveThat every manipulation attempt was caught. Controls reduce named failure modes; they do not certify a clean file.
- L03
Processing integrity
ProvesThat the number was produced by a recorded computation run against a registered, versioned method rather than an untracked spreadsheet.
Does not proveThat the method is the only defensible method. It proves which method was used and that it was not changed afterwards.
- L02
Publication integrity
ProvesThat the published result travels with its specification, sample context, methodology, and stated limitations attached.
Does not proveThat the result is free of sampling or measurement error. Disclosure makes uncertainty visible, not absent.
- L01
Audit integrity
BetaProvesThat a commitment to the published record is recorded publicly, so later alteration of that record is detectable by an outsider.
Does not proveThat the underlying opinion measurement is correct, and it is not yet mainnet or externally reviewed evidence.
Verified participation
Anatomy of a record
What actually travels with a PolitiMind result
This is the shape of the record behind a published figure, part by part, with what an outside reader can check and what that part does not establish. The readiness label on each part is the honest line between proof that exists now and transparency still in progress.
Parts of a published research record
- 01
Study specification
Live· available nowA reader can checkWhich questions were fielded, in which order, under which version — as frozen before the first response arrived.
Does not proveThat the specification was well designed for the research question.
- 02
Audience and eligibility basis
Live· available nowA reader can checkThat participation was gated on verification, and the eligibility rules the study applied.
Does not proveWho any individual participant is. Identity never travels with the record.
- 03
Response set summary
Live· available nowA reader can checkSample size, fielding window, and the integrity controls applied to the response set.
Does not proveThat the sample mirrors the target population without weighting.
- 04
Method version and run
Live· available nowA reader can checkThe registered method version used and the recorded computation run that produced the published figures.
Does not proveThat a different defensible method would produce the same figures.
- 05
Uncertainty and limitations
Live· available nowA reader can checkMargin of error discipline and the limitations the study itself declares.
Does not proveThat the stated uncertainty covers every source of error.
- 06
Record commitment
Beta· Base SepoliaA reader can checkA published commitment derived from the record under a documented encoding rule, recomputable by an outside verifier.
Does not proveAnything about representativeness. It establishes only that the record has not changed since publication.
- 07
Independent third-party review
PlannedA reader can checkNothing yet — external cryptographic and methodological review is a stated commitment, not a shipped surface.
Does not proveDo not read anything on this page as third-party certified or independently audited.
Structure is illustrative of what the record contains, not a schema. Beta parts are testnet-stage; planned parts are commitments, not capabilities.
Audit timeline
When each piece of proof comes into existence
Proof is not gathered at the end. Each artefact is created at a specific moment in the study lifecycle, and what an outsider can check depends on which moment they are asking about. Step through the phases below to see what exists at each point, who can verify it, and where transparency is still planned rather than shipped.
Evidence lifecycle, from study design to independent review
Before fielding
At study design, before a single response exists.
- Evidence created
- The research question, the declared sampling frame, and the questionnaire are written, reviewed, versioned, and frozen. The eligibility rules for the study are fixed.
- Who can check it
- Anyone reading the published result: the questionnaire and its version are published alongside the figures, so the fielded wording is checkable against the wording you are shown.
- What it does not settle
- That the design was the right design. Freezing removes silent mid-field edits; it does not adjudicate whether the question was well posed.
Phase 1 of 7: Before fielding
Live evidence versus planned transparencyPhases one through five are in place today. The record commitment is beta and currently anchored on Base Sepolia under Encoding Rule v1. Independent review and mainnet anchoring are planned commitments with no shipped surface — do not treat them as existing assurance.
For the skeptic
How you would check a number yourself
Work through the steps in order. Each one is something a journalist, researcher, or opposing analyst can do from outside PolitiMind, and each one names what it cannot settle so no step is mistaken for a verdict on the finding.
Open the result
- What you do
- Open a published result and read the figures next to the methodology, sample context, and stated limitations that travel with them.
- What it establishes
- That the claim, the method description, and the uncertainty were published together rather than assembled after a challenge.
- What it cannot tell you
- Whether the specification was frozen, or whether the record has changed since. Later steps cover those.
Steps four and five are beta: the encoding rule and verifier are published, and commitments are currently anchored on Base Sepolia pending external cryptographic review before mainnet.
What proof must never expose
Public verifiability without exposing participants or customer strategy
Auditability is often traded against privacy. It does not have to be. The record that outsiders can check describes the study and its computation; it never carries the identity of the people who answered, and it never carries a customer's private research strategy.
Never in the public record
- Participant identity, documents, or contact details
- Any individual's answers attributable to a person
- A customer's unpublished study strategy or internal findings
- Verification internals that would tell a fraudster what to imitate
In the public record
- The frozen study specification and fielding window
- Aggregate response counts and sample context
- The method version and the recorded computation run
- Stated uncertainty, limitations, and the record commitment
What we claim here, and what we do not
- Published results carry provenance: the specification, method version, and computation run behind a number are recorded and inspectable.
- Methodology is stated in plain language — sampling frame, weighting approach, margin of error, and known limitations travel with every result.
- Every study runs behind integrity controls that detect duplicate, coordinated, and low-quality participation before results are published.
- Participant identity and response data are held separately, so buyers receive population-level intelligence and never participant identities.
- BetaResult commitments are anchored to a public ledger so the record is tamper-evident after publication.
- BetaAn independent verifier can recompute a published commitment and check it against the anchored record.
Everything on this page concerns the integrity of the production chain: verified participation, a frozen specification, recorded computation, disclosed methodology, and a tamper-evident record. None of it establishes that a finding is accurate, that a sample is representative, or that sampling, nonresponse, and measurement error have been removed — they have not, by any method. Anchoring and independent recomputation are beta and currently testnet-stage; PolitiMind is not independently audited or third-party certified, and does not claim to be.
Inspect it yourself
Published results carry provenance: the specification, method version, and computation run behind a number are recorded and inspectable.
Open a published resultAn independent verifier can recompute a published commitment and check it against the anchored record.
Read the encoding ruleParticipation requires identity verification, so responses come from confirmed real people rather than anonymous traffic.
How verification worksWhy this is worth your time
Receipts change what happens when someone disputes your number
Research gets attacked when it is inconvenient. Without a record, the dispute becomes a contest of credibility that the challenger can win simply by being loud. With a record, the conversation moves back to the evidence: here is what was asked, here is who was eligible to answer, here is how it was computed, and here is proof the record has not moved.
That is not certainty. It is defensibility — and it is the part of this problem that is actually solvable.
Continue the argument
Why Polling Is Broken
Name the structural failure they have felt but not articulated
Read itIntegrity Infrastructure
Confirm results are tamper-evident and provenance is real
Read itParticipant Network
Understand who is in the network, how they join, coverage, and matching
Read itPublic Results
Inspect a real published result and its methodology
Read it