Preview access: any email + password demo
Production sign-in: SSO + verified identity. No impersonation — approvals are signature-grade. ← rifftest.com
This action is recorded as a signature — re-enter your password to sign. (Demo password: demo)
Rifftest runs the studies your team doesn't have hours for — designed to your standards, pressure-tested before fielding, synthetically pretested in minutes, and delivered decision-ready. Researchers set the guardrails; the whole organization moves at the speed of the question.
Synthetic personas modeled on this audience — instant, free, watermarked. Verified real respondents open in September; the design carries over unchanged.
Paste your actual messages, TPP attributes, or price points (2–8). The engine scores your words — the per-option power guard applies.
Wording flags are reviewed, never silently rewritten. In production this step routes through your e-signature approval chain — questionnaire → pretest → sample → fielding, each gate signed.
Method: ·
Screener + re-screen consistency probe applied · attention trap + reverse-worded straightliner trap inserted · wording flags reviewed, never silently rewritten · per-option power guard enforced.
Fielding opens this September. Same study, same analysis — verified US HCPs and patients, honoraria passed through at documented vendor cost, quote before anything fields. Join the fielding waitlist →
Fielding runs through verified panels (Prolific, Cint, M3, Rare Patient Voice) with identity checks, attention traps, and consent. Results replace the synthetic read and the validity gate re-scores — you see exactly where synthetic was right and wrong.
The client-approval chain, real-time translations, live quality control, and the focus-room playbook — everything between "approved design" and "clean data."
Each gate unlocks only when the prior one is signed. Approvals bind to a frozen snapshot — changing anything upstream voids every approval downstream automatically. Approved materials are locked artifacts: download or email-share from the gate.
For ex-US fielding, each language is machine-translated, back-translated for semantic checking, scored, and any deviation flagged for the human linguist — then the client approves per language, side by side, in real time. Nothing under threshold is silently accepted.
| Flag | Rule | Action taken |
|---|
Every removal is audited with the rule that fired — replacements auto-recruited so quotas still land. Respondents can pause and resume at any question; state is saved server-side and the resume link re-checks quota + screening before continuing.
| What goes wrong | How the room is built for it |
|---|
Cameras are optional — participants can join video, audio-only, or dial-in; sessions record only with consent on file. Consented recordings stay in your account room for your team to watch anytime; transcripts are auto-generated (speaker-attributed, PII-redacted) and available to view or download right after the session.
| Session | Consent | Recording | Transcript |
|---|
Access is workspace-scoped (viewers can watch; nothing is public). Retention follows your workspace policy and the consent terms — when consent limits retention, the vault enforces expiry automatically. A participant who declines recording can still take part: notes-only mode, clearly flagged in analysis.
Panel data ages: people switch jobs, treatments change, quotas fill. Before any interview begins, 2–3 screener confirmations re-run. A mismatch closes the session gracefully — thanked, partial incentive per policy, never "you failed" — and the quota slot is released for auto-replacement.
Every study's final deliverables, warehoused and alive — exports on your template, an AI presentation your audience can interrogate, the questions people actually asked, and the PV console.
| Study | Status | Exports |
|---|
Word export downloads a real document; PowerPoint renders server-side onto the active template. Every exported figure carries its receipt in the slide notes — numbers stay auditable even inside a deck.
Pick how long your audience has. The presenter ranks findings by decision-relevance and fits the story to the time — 3 minutes gets the verdict and the one thing to do; 60 minutes gets the full walkthrough. Viewers ask questions while it plays; it pauses, answers from the computed results only, and resumes. Watch or listen anytime — unlike a meeting recording, it answers back.
Every question asked in any presentation session is captured with its answer, organized by topic. Frequent questions surface to new viewers up front; unanswerable ones become follow-up tasks for the study owner. This is the institutional memory a meeting recording never becomes.
The manufacturer's own PV form is ingested once. When an interview surfaces a candidate adverse event, the form auto-populates from the transcript and lands here — nothing sends without your one-click review. The 24-hour clock runs from detection, with escalating reminders; the T-minus fail-safe policy (escalate vs. send-flagged) is set with the client's PV team.
Every study includes AI design, a free synthetic pretest, e-signature approvals, live-QC fielding, a receipt-bound report on your template, and an interactive AI presentation. join the launch list for early rates.