How we got to that number.

The product is in preliminary calibration. That means every projection on the calculator and every dollar figure on a Pitch Sheet is anchored to 9 public industry baselines and academic meta-analyses, NOT to a proprietary corpus of measured outcome data. Confidence intervals are intentionally wide (±30% half-width on every dollar projection) and tighten with every post-event reconciliation we ingest.

Two years from now, predictions will be tuned to your specific roster, your venue, and your attendee demographic. The first version is the worst it will ever be at predicting outcomes, and it's already better than the gut calls it's replacing. That's the bar.

Every speaker is scored across four behavioural pillars derived from face, body, and voice signals on uploaded video:

• Poise. Composure under load. Brow stillness, postural stability, vocal cleanness.

• Animation. Visible energy. Expression dynamics, gesture amplitude, vocal range.

• Connection. Audience-direction signals. Gaze, engagement intensity, vocal pacing discipline.

• Conviction. Channel congruence and the absence of leakage cues (suppression, contempt, peak asymmetry).

The four pillars combine into a composite Proof of Impact (POI) score on 0-100, mapped to four bands:

• Foundation0-49
• Capable50-69
• Elite70-84
• Iconic85-100

Confidence is computed per-pillar as the share of input weight that actually had data behind it (face track, body track, prosody). When a pillar drops below confidence floor, it's dropped from the composite and flagged.

Each outcome metric is driven by a specific weighted blend of pillars. Rows sum to 1.00. These are the coefficients in use today:

The mappings are anchored, not arbitrary. Animation and Connection drive immediate-experience outcomes (retention, on-site spend) because they're the pillars that hold a room. Conviction drives durable outcomes (return rate, sponsor renewal, social amplification) because authenticity is what gets remembered.

POI band → outcome multiplier vs. a Capable-median speaker:

The Capable→Elite step (+25%) matches the lower bound of vocal- variation retention research and well-designed upsell uplift. The Elite→Iconic step (+25% on top) matches the source-effects expertise band from the persuasion meta-analysis literature. The Foundation penalty (0.70) corresponds to the audience-drift thresholds observed when weak speakers cause 25-30% session drop-off in EventsAir KPI data.

Where the dollar figures actually come from:

• Median session retention rate: 70%
• Median additional on-site spend per attendee: $26.62
• Median YoY rebooking rate: 45%
• Median sponsor renewal probability: 55%

Caller-supplied event context (audience size, ticket price, your actual on-site spend per head, your sponsor revenue, your prior-year return rate) overrides these baselines wherever supplied. The calculator defaults to a 1,500-attendee corporate conference at $1,200 ticket, $180 on-site, $2M sponsor revenue (the brief's reference event), so the first number you see is anchored to a realistic mid-market scenario.

Every dollar projection on the calculator and every Pitch Sheet carries a symmetric ± confidence interval. Today that's ±30% half-width on every metric. That's wide. It's wide on purpose. Sources for the preliminary calibration are public industry stats, not a calibrated measurement corpus, and pretending otherwise would be dishonest.

The confidence interval on the total dollar projection is computed by RSS (root-sum-of-squares) aggregation of the per-metric half-widths, assuming per-metric errors are roughly independent. That's a generous assumption. Production calibration may correlate them, in which case the total CI will widen.

The path forward: every post-event reconciliation flow upload (actual audience retention from event app, post-event NPS from survey, sponsor renewal commitments, social mention volume from a tagging tool) feeds the calibration loop. The CIs tighten proportional to √N. By 200 measured-outcome sessions, the half-width drops to ±15%. By 1,000, ±10%.

Honest section. We'll keep this updated as the product matures.

• ✕

  Topic fit. The model scores delivery, not content relevance. A 92-POI keynote on the wrong topic will still bomb in front of the wrong audience.

• ✕

  Industry-specific calibration. Today's coefficients are cross-industry averages. Healthcare, finance, and tech audiences respond differently to the same delivery profile. Calibration loop segments this over time.

• ✕

  Live-only adjustment. We score recorded video. A speaker's ability to read a live room and adjust is not yet measured (that requires multi-camera audience capture during the actual event).

• ✕

  Cross-cultural weighting. Pillar attributions are anchored to predominantly Western corporate audiences. Asia-Pacific and Latin American audience-response data isn't in the calibration corpus yet.

Every coefficient on this page traces back to one of these. Click through; the underlying claims are public.

• Antonakis et al. — Charismatic Leadership Meta-Analysisacademic

  76-study, 36,031-individual meta-analysis — charisma is a measurable behavioural construct (vocal delivery, gestures, facial expression, eye contact) and predicts task performance, citizenship, and group performance.

• WiFi Talents — Speaker Industry Statistics 2024industry

  Vocal pitch variation lifts retention scores by up to 25%. Eye contact maintained 90% of the time rates a speaker 20% more persuasive. Storytelling makes information 22× more memorable than facts alone.

• AAE Speakers — 2024 Speaking Industry Benchmark Reportindustry

  67% of event planners say finding the right speaker is the most time-consuming part of their job. 49% of marketers say audience engagement is the single biggest contributor to a successful event.

• Ticket Fairy — Venue KPIs 2026industry

  Industry baseline: organisers generate ≈ $26.62 additional on-site revenue per attendee beyond the base ticket price. Larger events exceed $40 per attendee. On-site spend (F&B, merch, premium upgrades) is 15-30% of total live event revenue.

• PCMA Convene — 55% of Meeting Budgetindustry

  F&B accounts for 29% of exhibition + meeting cost. Combined F&B + AV + speaker/entertainment line items account for 55% of total meeting spend — the speaker is not a rounding error in the P&L.

• Cvent — 390 Event Statistics 2026industry

  72% of CPG event attendees post or share brand content after an event (Spiro). Receiving post-event communications increased purchase intent by 45%. Attendance retention is tracked as a key ROI metric by 59% of marketers (Splash).

• Remo — B2B Event Sponsorship ROI Guideindustry

  B2B event leads typically cost $150-300 each. Average event ROI is 25-34% across the industry, varying with attendance, sponsorship revenue, and brand awareness.

• EventsAir — 10 Event KPIsindustry

  Sponsor booth location: prime locations near registration, coffee stations, and main stages see 3× more traffic than tucked-corner locations. Sessions immediately following high-engagement programming see materially higher booth traffic in the post-session window.

• Source Effects in Communication & Persuasion (J. Acad. Marketing Sci.)academic

  Meta-analysis: speaker expertise has the largest single effect on persuasion outcomes — averaging 16% of explained variance. Trustworthiness and dynamism (perceived energy / animation) follow as the next two largest factors.