The methodology that makes the platform work.

Behavioral measurement captures actions, choices, and reactions that don't pass through stated-preference filtering. Drag-and-drop exercises, choice mechanics (MaxDiff, conjoint, choice-based scaling), shelf-simulation studies, and reaction-time tests all fall here.

The methodological case for behavioral measurement is the say-do gap: across decades of research, stated preference and observed choice diverge by 40-50% on average. The Tropicana redesign, the Bertolli case, New Coke — all stated-preference success stories that failed in observed behavior.

RIWI's Research product runs behavioral methods in the same study session as surveys and implicit tests. They're not separate studies; they're parallel signals that triangulate.

• Choice mechanics: MaxDiff, conjoint, choice-based scaling
• Shelf simulation: virtual shelf, FMOT studies
• Reaction-time methods: implicit association, priming
• Drag-and-drop and card-sort exercises

Physiological measurement at scale means webcam-based methods — eye-tracking, facial coding for emotion, attention measurement — that used to require lab equipment but now run in respondents' own browsers.

Webcam eye-tracking is calibrated, validated, and widely accepted as a reliable signal for attention and saccade tracking on standard tasks. Facial coding identifies emotional expressions (valence and arousal) frame by frame across video stimuli. Both run in-browser, on the respondent's device, without specialized hardware.

This is what makes physiological measurement available outside the lab — and what makes it usable at the sample sizes consumer research demands.

• Webcam eye-tracking — heatmaps, fixations, saccades
• Facial coding — emotional response, frame by frame
• Attention diagnostics — dwell, return, escape
• All in-browser, no special hardware

RDIT is the methodology that powers RIWI's Audience product. Instead of recruiting respondents into a panel and surveying them repeatedly (the panel methodology), RDIT samples respondents randomly across the open internet — across millions of partner websites and apps — without prior recruitment, without incentivized participation, without survey fatigue.

Methodologically, RDIT is a closer approximation of random sampling than panel methods. Panels have selection bias (the people who join panels are different from the people who don't), incentive bias (responses shaped by reward structure), and fatigue bias (respondents in panels too long give different answers). RDIT minimizes all three.

RDIT enables work no panel can: hard-to-reach populations, fragile states, low-incidence professionals, country coverage panels can't supply.

• Random sampling across millions of partner sites
• No panel registration, no incentive bias, no fatigue
• Reach: 200+ countries, including hard-to-reach
• The methodology behind RIWI's Audience product

RIWI's research team is exploring synthetic audience methodology — the use of AI-generated respondents alongside (not instead of) verified humans, for hypothesis generation and study scoping.

This is a research direction, not a current capability. The methodological questions are real: where synthetic respondents reliably approximate human response, where they don't, and where the boundary lies. RIWI's view is that synthetic methods will be useful in narrow contexts where they earn their keep — hypothesis testing, methodology pre-validation, scoping work — alongside, never replacing, the verified human respondents that grounded research depends on.

We'll publish methodology and validation work as it develops.