Our Data Integrity Principles
Brands buy attention on LaunchPad.AI because the humans on the other side are real, qualified, and consenting. We enforce four non-negotiable principles across the entire platform:
Verified before rewarded
A participant cannot earn a single dollar until identity, device, and household have been verified through independent signals.
Continuously re-scored
Trust is not a one-time gate. Every session, survey, video, and payout re-scores the participant against 130+ behavioral and technical signals.
Consent-first data
Every consumer insight is captured under explicit, versioned consent. Nothing scraped, nothing inferred without disclosure.
Auditable end-to-end
Every verification, decision, and payout is logged immutably. Brands and regulators can trace any row back to its evidence.
Identity & KYC Verification
Every participant clears a multi-layer identity gate before they can join a campaign. This is the same KYC framework used by regulated fintech partners.
- 1
Email + phone double opt-in
Confirmed via time-limited codes. Disposable domains and VoIP numbers are blocked.
- 2
Government-issued photo ID
Driver's license, state ID, or passport captured through a liveness-checked flow. Documents are OCR'd, MRZ-verified, and cross-checked against issuer databases.
- 3
Selfie liveness match
Active liveness (blink, turn, distance) matched against the ID photo. Deepfake and mask attacks are rejected.
- 4
Address of residence
Confirmed against utility data, credit-header data, or a household member on file.
- 5
Device & network fingerprint
Hardware fingerprint, IP reputation, ASN, geo, VPN/proxy/Tor status, and emulator detection.
- 6
Sanctions & PEP screening
OFAC, UN, EU consolidated lists, and adverse-media screening for every US-adult participant.
- 7
Household de-duplication
Cross-account graph matching prevents the same household from farming multiple accounts against a single campaign.
Institutional Verification
Some LaunchPad.AI audiences require an additional layer of institutional proof. We verify these attributes through primary sources — not self-declaration.
Renters (RentGain.AI)
Verified against lease-data partners, property-manager rosters, and rent-payment history. A renter is only a renter if a landlord agrees.
Students (CollegeGain.AI)
SheerID and .edu domain verification, plus enrollment-file matching for participating institutions.
Homeowners
County assessor and title-data matching against the verified address of residence.
Parents & households
Household composition confirmed against credit-header household graphs and school-district enrollment data where consented.
Professionals
NPI, bar, and professional-license registry checks for gated verticals (healthcare, legal, real-estate).
Income & employment
Optional payroll and bank-linked verification via Plaid for income-gated studies.
AI Assessment & Scoring
Every account carries a live Trust Score maintained by our scoring engine. The score blends verification evidence with behavioral history and is recomputed on every session.
Fully verified, long tenure, zero flags. Eligible for premium studies and highest-tier payouts.
KYC + institutional verification complete. Eligible for standard campaigns.
Partial verification or recent flags. Limited campaign access; capped rewards.
Multiple behavioral flags or failed verifications. Read-only; cannot participate.
Confirmed fraud, sanctions match, or terms violation. Removed from all campaigns and reported to the fraud network.
What feeds the score
- • Identity & document verification
- • Device / IP / geo consistency
- • Response latency & attention patterns
- • Open-ended text quality (LLM-graded)
- • Video engagement telemetry
- • Cross-campaign consistency
- • Brand feedback + audit outcomes
In-Field Monitoring
Once a campaign is live, monitoring never stops. Every response is streamed through a real-time integrity pipeline before it's counted toward a brand's results.
Attention telemetry
Focus events, tab switches, dwell time per question, and scroll velocity flag disengaged sessions.
Speeding & straight-lining
Sessions completing below the 25th-percentile speed threshold, or flat-lining across grids, are auto-quarantined.
Open-ended LLM grading
Every free-text answer is scored for relevance, coherence, and originality. Boilerplate, gibberish, and LLM-generated responses are flagged.
Video review authenticity
Face presence, lip-sync consistency, and audio-liveness confirm the reviewer is the verified participant.
Duplicate detection
Semantic hashing across responses catches copy-paste rings even when text is lightly rewritten.
Live source optimization
Traffic sources delivering low-quality respondents are throttled or paused mid-campaign, protecting the sample.
Fraud & Bad-Actor Defense
We combine our own signals with the industry's leading fraud-defense networks to keep bad actors out — and, when they get creative, to catch them fast.
130+ Fingerprint Signals
Every session identified through hardware, browser, and behavioral fingerprints.
No-Fly List
Known bad actors blocked across our platform and shared fraud networks.
24-Hour Activity Watch
Unusual velocity, farming rings, and coordinated attacks detected within hours, not days.
Third-Party Fraud Networks
Live integrations with Research Defender, IPQS, and adjacent-industry fraud lists.
Emerging Threat Detection
AI models flag novel evasion patterns — new VPN ranges, headless browsers, and automation frameworks.
Bounty & Manual Review
Brands can flag any suspect row; our trust team investigates within 48 hours and refunds if confirmed.
Post-Field Auditing
Before results are delivered to a brand, every dataset passes through a three-tier audit. Rows that don't clear the audit are removed and refunded automatically.
Automated integrity sweep
Speeding, straight-lining, duplicate, and fingerprint checks are re-run against the full dataset. Confidence bands attached to every row.
AI qualitative review
Open-ended answers, video reviews, and photo uploads are re-scored by our LLM/vision models for coherence, originality, and prompt-adherence.
Human spot-check
Our data-integrity analysts sample 5–15% of every campaign (100% on Advance-tier studies) and sign off before delivery.
Every delivered row includes a provenance record: participant Trust Score at completion, verification evidence, device fingerprint hash, and audit outcomes. Brands can inspect it anytime from the campaign export.
Governance & Compliance
Consent & privacy
All data collection is governed by our Privacy Policy, Terms of Service, and Data Processing Addendum. Consent is versioned and revocable.
Regulatory framework
Practices aligned with GDPR, CCPA/CPRA, COPPA (18+ only for participation), and the Insights Association Code of Standards.
Access controls
Row-level security on every table. Least-privilege access for internal staff. Service-role keys are rotated and never exposed to clients.
Auditable log
Every verification decision, score change, and payout is immutably logged for a minimum of 24 months.
Vendor due diligence
Every KYC, ID, and fraud partner is security-reviewed at onboarding and re-audited annually.
Incident response
Documented process for detection, containment, notification, and remediation. Response SLAs published to enterprise customers.
For legal detail see our Privacy Policy, Terms of Service, and Data Processing Addendum.
Quality SLAs
| Metric | Standard | Advance |
|---|---|---|
| Verified-identity rate | 100% | 100% |
| Post-field bad-actor removal | ≥ 95% | ≥ 99% |
| Human spot-check coverage | 5–10% | 100% |
| Bad-row refund window | 14 days | 30 days |
| Fraud investigation SLA | 72 hours | 48 hours |
| Provenance record per row | Included | Included + raw evidence |
Have a question our team should answer?
Enterprise buyers can request our full data-integrity whitepaper, vendor questionnaire responses, and a live walkthrough with our trust team.