User persona analysis — Conversate smart glasses

30-day window ending 2026-05-23 · 210 sampled users (70 per tier × 3 geographies)

Important caveats — read before interpreting

⚠ Conversate-usage sample, not a glasses-customer sample. Every user in this analysis actively triggered Conversate AI-cue sessions during the 30-day window. They are not a representative sample of all Even Realities glasses customers — they are the subset who turned on and engaged with the Conversate feature. The persona mix below reflects who finds the AI-cue feature useful enough to keep using it, which may deviate significantly from the broader customer distribution (e.g. customers who only use display/translation/HUD features without Conversate are invisible here). Use these personas to inform Conversate marketing and product decisions; do not generalize to overall glasses positioning without separate validation.

⚠ Wearer-unknown attribution. Transcripts contain multiple speakers and the wearer is not identifiable. Themes are inferred from recurring context across many sessions (≥ 3-session evidence bar), not from utterances attributed to a specific person. Phrasing like "sessions recur in X" is intentional — do not read it as "the wearer is X".

⚠ Sample-size limits. 70 users per tier × 3 geographies = 210 users total. Persona counts ≤ 5 are directionally interesting but not statistically reliable. Population extrapolations below are rough sizing estimates (±large), not forecasts.

⚠ How percentages in this report are computed. We deliberately oversample power users in the sample (70 power / 70 regular / 70 casual) so we can characterize the small power tier reliably — even though they are only ~13 % of the eligible population. All headline percentages and ≈-counts shown for the overall population (executive summary and the right-most column of each rollup table) are reweighted to the true tier mix (1,444 power · 5,370 regular · 4,562 casual). The per-tier columns are not reweighted — each column already has its own tier-specific denominator (70), so the percentages there read as "share of sampled users in that tier". Without this reweighting, themes that skew to power users (e.g. industrial / financial / B2B) would look ≈3–4× larger than they actually are at the population level.

Executive summary

Two numbers per row, so the picture is unambiguous:

• Share in sample — how many of the 210 sampled users showed this persona / use case. Useful for understanding the evidence base.
• Estimated share in user base — the same signal, reweighted to the true tier mix in the eligible Conversate population (11,376 users). This is the marketing-quotable number.

The two columns can differ noticeably for personas that skew strongly to one tier (power users are oversampled 5× to give us reliable signal on the small power tier).

Top persona themes

Top use cases

Methodology and sample composition

• Population: 11,376 Conversate-eligible users out of 15,317 total Conversate-active users in the last 30 days (one-shot users excluded for persona signal).
• Sample: 210 users, stratified by tier × geography.
• Tiers (derived from 30-day activity rollup):
  • Power — ≥ 25 sessions and ≥ 7 active days
  • Regular — 5–24 sessions and ≥ 3 active days
  • Casual — 2–4 sessions
• Method: each sampled user's full session history is summarized, then analyzed for recurring themes and use cases that appear across at least 3 sessions; similar labels are consolidated across users into a single canonical set for the rollup tables below.

Eligible population by tier (Conversate-active users meeting the tier cutoff in the 30-day window):

Engagement by tier

Counts are over the 30-day window. Cue totals are computed from up to 50 most-recent sessions per user (the analysis cap), so heavy power users may be undercounted on cue volume but their rates are still accurate.

cue_open_rate can exceed 100 % when users repeatedly re-open the same cue — observed on a handful of power users. Median is the more robust signal.

Users with no persona signal (no theme met the 3-session evidence bar — usually because most of their sessions had no transcribable content, e.g. silence or background audio):

Casual-tier signal collapses sharply: with only 2–4 sessions per user, there is often not enough recurring content to confidently identify a persona. Persona claims about the casual tier should carry the heaviest scepticism.

Persona themes by tier

Bars show the share of sampled users in that tier exhibiting the theme (denominators: Power=70, Regular=70, Casual=70). Rows are ordered by the population-weighted share (right-most columns) so the most population-significant themes appear first — not the themes that happen to over-index on power users.

Persona themes by geography

Bars show the share of sampled users in each geography exhibiting the theme (sample sizes: US=69, JP=69, other=72; users are evenly split across the three activity tiers within each geography).

Use cases by tier

Bars show share of sampled users in each tier exhibiting the use case. Rows are ordered by population-weighted share (right-most columns).

Use cases by geography

Sampled persona profiles (sensitive details removed)

Per-user roll-up of themes and use cases. Free-text descriptions, specific evidence (city names, brand names, individual session references), and per-user engagement rates are intentionally removed to prevent identification of any single user. The 8-character codes are anonymized user identifiers — they cannot be linked back to a real account.

Tier: power (70 users sampled)

Tier: regular (70 users sampled)

Tier: casual (70 users sampled)

Appendix: full theme and use-case glossary

Themes

• Smart Glasses Early Adopter / Evaluator — Users who actively test, demonstrate, evaluate, or advocate for smart glasses and wearable AR technology.
• Tech & AI Enthusiast / Professional — Users with strong interest or professional involvement in software, AI, emerging technology, and digital tools.
• Corporate & Business Professional — Users working in corporate, enterprise, or organizational roles including management, operations, finance, and business development.
• Industrial & Trades Worker — Users involved in manufacturing, construction, trades, industrial operations, or hands-on technical work environments.
• Healthcare & Life Sciences Professional — Users working in or adjacent to healthcare, clinical, biomedical, pharmaceutical, or medical education settings.
• Faith & Religious Community Member — Users who are active participants in religious communities, faith-based organizations, or spiritual leadership roles.
• Family-Oriented Domestic Lifestyle — Users whose primary context is family life, parenting, household management, and domestic social routines.
• Casual Social & Leisure Lifestyle — Users whose primary context is informal socializing, dining, entertainment, and everyday leisure activities.
• Education & Academic Environment — Users involved in educational institutions, academic research, teaching, training, or student life.
• Multilingual & Cross-Cultural Professional — Users operating in multilingual, multicultural, or internationally diverse personal or professional environments.
• Government, Public Sector & Regulated Industry — Users working in government agencies, public policy, regulatory compliance, aviation, or other regulated sectors.
• Entrepreneurial & Startup Ecosystem — Users engaged in entrepreneurship, startups, venture-building, sales, or business development roles.
• Accessibility & Assistive Technology Advocate — Users focused on assistive technology, accessibility features, or serving communities with disabilities.
• Science & Engineering Researcher — Users involved in scientific research, engineering, materials science, chemistry, or laboratory work.
• Urban Commuter & Local Community Member — Users defined primarily by their geographic locale, urban commuting habits, and local community participation.
• Media, Culture & Entertainment Consumer — Users with strong engagement in pop culture, anime, gaming, performing arts, sports, or entertainment media.
• Miscellaneous / Broad Interest User — Users with general curiosity, broad intellectual interests, or contexts that do not fit neatly into other categories.

Use cases

• Professional meeting capture and assistance — Passive or active capture, transcription, note-taking, and real-time cueing during workplace meetings, strategy sessions, briefings, and corporate discussions.
• Product demonstration and feature testing — Showcasing, evaluating, or testing smart glasses capabilities, features, and hardware to colleagues, clients, or third parties.
• Casual social and everyday ambient capture — Ambient or passive recording during informal social gatherings, family interactions, domestic settings, and everyday outings.
• Conference, seminar, and lecture attendance — Passive listening, transcription, or note support while attending academic, technical, or professional presentations, conferences, and seminars.
• Real-time translation and multilingual support — Live translation, multilingual transcription, and cross-language communication assistance during conversations or meetings.
• Religious and community event participation — Wearing or using glasses during religious services, worship gatherings, Bible study, community ceremonies, and civic forums.
• Retail, transactional, and errand settings — Ambient capture or documentation support during shopping trips, retail transactions, dining, and routine service errands.
• Healthcare and clinical encounter support — Capture, transcription, or real-time reference assistance during medical consultations, clinical visits, telehealth calls, and healthcare meetings.
• Education, tutoring, and learning support — Assistance, transcription, or capture during tutoring sessions, classroom instruction, academic advising, and structured learning contexts.
• Industrial, field, and technical work support — Hands-free capture, troubleshooting assistance, and documentation during construction, manufacturing, engineering, or field operations.
• Transit, travel, and commute capture — Ambient capture or navigation assistance during public transit commutes, in-vehicle travel, and on-the-go urban movement.
• Real-time information lookup and fact-checking — On-demand AI-assisted information retrieval, fact-checking, term definition, and trivia assistance during conversations or activities.
• Device setup, testing, and troubleshooting — Technical evaluation, hardware pairing, UI calibration, audio quality testing, and troubleshooting of the smart glasses device itself.
• Accessibility and hearing assistance — Real-time captioning, subtitle display, and hearing or vision accessibility support for users with sensory needs.
• Media, entertainment, and broadcast monitoring — Passive listening or capture during media consumption, live sports, broadcasts, podcasts, radio, and entertainment events.
• Compliance, legal, and regulatory documentation — Capture and documentation support during compliance briefings, legal proceedings, regulatory hearings, audits, and HR or contract discussions.
• Sales, client, and business development meetings — Capture and assistance during client-facing sales presentations, business development discussions, and enterprise sales sessions.
• Miscellaneous and incidental use — Idle or background wear, personal monologue capture, exploratory use, and other incidental or unclassified contexts.

Limitations and recommended follow-ups

• Conversate-usage selection bias (see caveat above): every finding describes the Conversate-engaging subset, not all glasses customers. To characterize the full customer base, separately sample users who own glasses but don't engage with Conversate.
• Wearer attribution: every theme is contextual inference. Confounders: shared-device use, sessions where the wearer is silent, ambient capture of unrelated speakers.
• Sample size: 70 users per tier (210 total); counts ≤ 5 are directional only.
• 30-day window: lapsed cohorts and lifetime-tier signal are not captured. Re-run with a wider window for retention or churn questions.
• JP / non-English nuance: JP and 'Other' geographies may deserve a separate analysis run with native-language summarization for better signal extraction.
• Population estimates are rough: extrapolations from 30/tier sample have wide confidence bands; treat them as order-of-magnitude.