// HUNTINGTON ANALYTICS

The independent read on whether your creative will work.

We don’t sell media, and we don’t grade our own homework. Huntington Analytics is building the neutral, cross-platform benchmark for what actually moves an audience — so the verdict on your creative is yours, not a platform’s.

Apply to the pilot
// WHY INDEPENDENT

Every platform grades its own homework. We don’t.

Meta, Google, and TikTok each score the creative you run on them — and each earns more when you spend more. None can be the impartial, cross-platform read. That’s the gap we fill: one neutral standard for whether creative works, independent of where it runs.

Scores its own creative · paid when you spend

MetaGoogleTikTok

Neutral · cross-platform · paid only by you

Huntington Analytics
// THE BENCHMARK

Building the outcome benchmark for creative.

A credit bureau is valuable because it links a record to what actually happened. We’re building the equivalent for creative — an accumulating, cross-brand library of which creative moves drove real lift, in which category. We start where outcomes are clean, become the benchmark there, then widen.

01

Start clean

Begin where outcomes are unambiguous.

02

Become the benchmark

Set the neutral standard in that category.

03

Then widen

Extend the library category by category.

Early-stage: we lead with the neutral read today and validate against outcomes as the benchmark grows.

// THE INSTRUMENT

See your video’s emotional journey, second by second.

How we read the creative: NeuroLens reads a finished video the way an audience does — and maps the whole emotional arc automatically.

Characters

What the characters feel

Read from their voice and their face. When the two don't match, that gap is the performance.

Audience

What the audience feels

Separately, from the music, camera, and editing pace. The director's craft.

Subtext

What's said vs how it's said

The gap between the words and the delivery, where irony and subtext live.

Arc

The shape of the story

The emotional arc and its exact turning points, found automatically.

Point it at any film or ad and it finds the emotional turning points on its own — no script, no setup.

// Feedback Loop

How the measurement signal conditions the next generation.

When NeuroLens partners with a generative-media model, scoring stops being a post-hoc judgment and becomes part of the production loop. Variants drafted, variants scored, convergent winners returned to the generator as conditioning — every cycle sharpens the next.

01

Generate

Variants drafted in parallel — copy, voice, frame, composition. Hundreds at a time, not a handful.

02

Score

Each variant scored across every emotional beat — face, eyes, voice, language, social — before any human looks at it.

03

Converge

A variant advances only when at least two of nine signals agree on the audience response. Single-signal lift gets flagged, not promoted.

04

Condition

Convergent winners — and the reasons they won — are returned to the generative model as next-pass conditioning. The loop closes.

Currently piloting with generative-media partners · seeking design partners to test

// The Stack · Eight More Signals

The pre-flight read and the A/B readout are two surfaces over a nine-signal stack.

Each signal below feeds the full methodology. A finding is reported only when at least two of the nine agree. Disagreement gets flagged, not hidden.

Eye Tracking

Gaze · dwell · attention curve

Per-frame gaze coordinates and dwell times. Surfaces where attention actually lands versus where the creative is asking it to.

60 Hz gaze · ~28 ms median fixation latency

Facial Expression Recognition

Micro-expression coding

Frame-rate facial-action-unit detection mapped to seven canonical emotions plus contempt. Catches the half-second tells that survey data misses.

8 FAU classes · per-frame confidence vectors

NLP Sentiment

Linguistic feeling, not keyword counting

Transformer-based sentiment scoring on copy, scripts, captions, and customer voice — reads positive vs negative, intensity, and context, including irony detection.

Continuous −1.0 → +1.0 positive/negative + intensity

Social Monitoring

Brand-signal stream

Continuous monitoring of brand-relevant social signal — volume, sentiment, semantic drift, and competitor share — into a single time-series view.

Hourly NPS proxy · 7-day rolling baseline

Sentiment Databases + Cultural Tropes

Movies · Music · Cultural references

A semantic library of how cultural references actually land — drawing on movie scenes, music eras, regional idiom, and in-group signal — used to flag tone-brand mismatches before they ship.

200K+ tagged references across film, music, and regional language

Semantic-to-Behavior Prediction

Language → action models

Predicts purchase intent, share probability, and call-to-action follow-through from language features alone. The closer the language sits to the buyer's, the higher the score.

Predicted intent vs measured intent · r ≈ 0.71

Attention Mechanisms

Cutting-edge attention modeling

Cutting-edge attention mechanisms — borrowed from the same research powering modern foundation models — surface where audience focus actually concentrates in a creative, versus where it just drifts past.

Per-spot attention concentration · attention-shift timeline

Generative-Media Feedback

Pre-screening generated variants

Scores AI-generated and AI-augmented creative variants against the same nine signals — so the production model can condition on what scored, not on what was guessed.

50 variants screened → 5 defensible in a morning

Sample metrics · illustrative · methodology calibrated against the 60-clip reference corpus

// The science under the hood · Pre-flight scoring

Read a spot before you spend.

The scoring engine behind NeuroLens. It reads the attention curve, every emotional beat, and predicted recall, then returns a clear signal: APPROVED, CONDITIONAL, or REJECTED — along with the time-stamped seconds your team can recut. Click any row to drill into the per-class scorecard.

Reference creativeSpotDune: Part Two · Warner Bros.
APPROVED
Cleared for media buy
All twelve classes inside target band. Ship as cut.
Premium sci-fi positioning · 60-second global launch spot
Peak 0:23–0:34Awe + Curiosity peak — desert reveal lands
Dip 0:51–0:55Calm dip during transition, within tolerance
ConvictionHIGH
Per-class scorecard · 12 emotional responses · 0–100on-spec 11 · under 0 · over 1
JoyON-SPEC · 72
TrustON-SPEC · 68
AnticipationON-SPEC · 64
PrideON-SPEC · 51
CuriosityON-SPEC · 78
AweOVER · 81
SurpriseON-SPEC · 55
CalmON-SPEC · 42
FearON-SPEC · 22
SadnessON-SPEC · 14
AngerON-SPEC · 8
DisgustON-SPEC · 5
Target band
Score in-spec
Outside band
Reference creativeSpotBarbie · Warner Bros.
CONDITIONAL
Recut 0:34–0:52, then re-test
Spot is mostly on-spec. Targeted recut at flagged seconds, retest, ship.
Youth-segment campaign · brand refresh seeking edge
Peak 0:08–0:14Joy + Surprise peak — color reveal works
Dip 0:34–0:52Calm and Awe both dip below target band
ConvictionMEDIUM
Per-class scorecard · 12 emotional responses · 0–100on-spec 10 · under 2 · over 0
JoyON-SPEC · 78
TrustON-SPEC · 48
AnticipationON-SPEC · 55
PrideON-SPEC · 32
CuriosityON-SPEC · 71
AweUNDER · 38
SurpriseON-SPEC · 64
CalmUNDER · 18
FearON-SPEC · 14
SadnessON-SPEC · 22
AngerON-SPEC · 12
DisgustON-SPEC · 8
Target band
Score in-spec
Outside band
Reference creativeSpotOppenheimer · Universal Pictures
REJECTED
Don't run — fundamental tone-brand mismatch
Multiple classes outside brand-safe band. Don't run; rebuild from concept.
Family-brand expansion · holiday spot (tone mismatch)
Peak 1:14–1:22Awe + Anticipation peak (wrong genre signal)
Dip ThroughoutFear + Sadness + Anger all above brand-safe band
ConvictionHIGH
Per-class scorecard · 12 emotional responses · 0–100on-spec 5 · under 4 · over 3
JoyUNDER · 28
TrustUNDER · 32
AnticipationON-SPEC · 51
PrideUNDER · 24
CuriosityON-SPEC · 62
AweON-SPEC · 71
SurpriseON-SPEC · 48
CalmUNDER · 22
FearOVER · 48
SadnessOVER · 38
AngerOVER · 24
DisgustON-SPEC · 14
Target band
Score in-spec
Outside band

Illustrative · sample reads · not client data · target bands derive from the 60-clip reference corpus

// The science under the hood · A/B scoring

Two cuts. One audience.

The same engine scores A/B variants head-to-head. Two cuts of the same 60-second spot, scored across ten emotional dimensions and five audience-response metrics, then collapsed to a verdict. The example below tests an environmental opening cut against a stakes-forward opening cut from the same trailer.

Cut A — env. open reference contentCut A — env. openAvatar: The Way of Water · 20th Century Studios · 0:22
vs
Cut B — stakes open reference contentCut B — stakes openAvatar: The Way of Water · 20th Century Studios · 1:15
Statistical Summary60-clip corpus · head-to-head
2
A Wins
34.5%
Audience Response Difference
3
B Wins
EngagementB Win
55.2% vs 54.8%
Shared ResponseA Win
0.222 vs 0.180
ConnectivityB Win
0.422 vs 0.425
Transition RateB Win
0.057/ms vs 0.072/ms
Peak NetworkA Win
78.1% vs 65.5%

The two cuts produce meaningfully different audience responses (diff score 34.5%). The stakes-forward cut drives deeper engagement and faster emotional state transitions; the environmental cut holds attention more consistently and audiences respond more strongly together. Both pass; the choice is which arc your media plan is buying for.

Emotion Dimension Differences
A stronger
B stronger
Awe / Wonder
A +16
Happiness
A +6
Place Recognition
A +22
Engagement
B +20
Surprise
B +24
Memory
B +12
Fear / Tension
B +42
Anger
B +22
Sadness
B +20
Mind Wander
A +30

Illustrative · sample comparison · derived from NeuroLens ab-testing.html dashboard

// The science under the hood · Why it works

Nine signals reading the same moment. A finding only when they agree.

Face, eyes, voice, language, and social signal — read in parallel. A result is reported only when multiple signals converge. That’s what makes NeuroLens’s reads defensible, not just fast.

// The science under the hood · How we know it works

How we know the reads are defensible.

NeuroLens’s verdicts are backed by rigorous methodology and continuous validation — not intuition.

01

Peer-reviewed lineage.

Built on decades of published work in measurement science.

02

Rigorous calibration.

Every read benchmarked against an internal reference corpus.

Continuously refined

03

Ongoing validation.

Predictions continuously checked against measured response.

Multi-modal · continuous

04

Convergence over confidence.

Findings reported only when multiple signals agree. Disagreement is named, not hidden.

Convergence required

Methodology brief available on request before engagement

// Where We Are

Six months ago this was a methodology. Today it’s a platform looking for design partners.

2022

Custom media taxonomies

Custom media taxonomies built and compared against psychological models in embedded space.

Step 1

2024

Machine learning research

Research into multi-modal emotional inference and cross-linguistic colexification mapping — the semantic groundwork the platform grew from.

Step 2

Oct 2025

NeuroLens testing begins

First audience-response predictions run against the 60-clip corpus. Convergence rule shipped.

Step 3

Jan 2026

Huntington Analytics, LLC

West Virginia entity formed. The firm launches as the commercial face of the platform.

Step 4

Apr 2026

Methodology validated

Internal validation runs confirm predicted-vs-measured emotional response across a wide variety of content.

Step 5

May 2026 · now

gBETA Marshall cohort

Selected into the Summer 2026 gBETA Marshall accelerator.

Step 6

Q3 2026

First generative-media partner pilot

First generative-media partner runs the feedback loop in production — variants scored, convergent winners returned as conditioning.

Step 7

2027

Generative-media partnerships scale

The feedback loop closes at production scale: scoring → conditioning → next-generation.

Step 8

Dates after May 2026 are scheduled, not promised · selection is rolling

// Pilot Cohort · What You Take Home

What a design partner actually takes home.

Specifics, not pitch. A six-month, no-fee engagement, focused scope, four deliverables we’re putting our name on.

Readout

Defensible written readout

A written report your team can put in front of a CMO without translation. Verdict, reasons, recut recommendations, methodology appendix. Not a dashboard screenshot.

Scorecard

Per-spot scorecard

Every emotional beat scored against the reference band, with time-stamped peaks and dips. The kind of file an analyst can drill into and a director can skim.

Founder time

4 hours founder time / month

Strategy + roadmap, not just diagnostic. The person who built the stack reviews the read with your team and helps map what to test next.

Reference network

Reference network + case study

Co-authored case study (review and approve before publication), logo rights, and two investor reference calls during the engagement window. We borrow your credibility; you borrow ours.

Six-month engagement · no fee · selecting for fit, not budget

// NEWSLETTER

Stay Ahead with Data Insights

Get monthly tips, case studies, and analytics strategies delivered to your inbox. Learn how to turn your data into competitive advantage.

No spam, ever. Unsubscribe anytime.

// Pilot Program

We’re selecting design partners to pilot the platform.

The methodology is built. The validation work is in front of us. We’re looking for a small cohort of brand and creative teams to put the platform on a real decision and tell us what holds up.

6 mo

No-fee design-partner engagement. Focused scope, defensible readout your team can put in front of a CMO.

6 seats

We’re selecting a small cohort of brand and creative teams to pilot the platform. Selection is rolling.

20 min

The entire ask up front. One conversation about how creative gets approved inside your organization.

// Frequently Asked

The questions we always get.

If you have a sharper one, the “Apply to the Pilot” CTA gets you a 20-minute conversation in front of the person who built the platform.

What do you need from us to run a single-spot read?

One creative file (final-cut or near-final, mp4/mov), plus a one-paragraph context — brand, audience, what you’re testing for. We handle calibration against the internal reference corpus. Mutual NDA standard before any file moves.

How long does a pilot take?

A single-spot pre-flight read returns a written verdict in 2 hours from file receipt. An A/B comparison runs in the same window. The full six-month design-partner engagement scales to roughly one read per month plus founder-time review.

What is the methodology validation behind the platform?

Built on decades of peer-reviewed work in measurement science. Calibration baseline is an internal reference corpus, continuously refined. Predictions checked against measured response through multi-modal validation runs, ongoing.

How do you handle pre-release creative under NDA?

Mutual NDA before any data is shared, in either direction. The creative stays on a single-tenant ingest pipeline; no model is trained on your content. Security posture documented in SECURITY.md, available on request before engagement. Subprocessor list disclosed.

What if the verdict says don’t ship?

That’s the work. The readout names the dimensions that are off, the seconds that drive them, and recut recommendations. We’ll re-run a second pre-flight at no additional fee if the recut comes back within the engagement window. A NO-GO verdict is a 1-week head start on a salvage plan, not a dead end.

What does a design-partner engagement cost?

No fee. Six-month engagement, focused scope, defensible deliverable. We’re selecting partners for fit, not budget. We get co-authored case study + logo rights + two investor reference calls; you get the readouts, four hours of founder time per month, and a methodology your team can defend in a CMO review.

// DESIGN PARTNERS

Looking for short conversations with creative leaders, brand marketing decision-makers, and agency principals.

Especially in healthcare, public-health, behavior-change creative, or any campaign where the brief is to actually change behavior, not just to be remembered.

For organizations that are a strong fit, we offer a six-month no-fee design-partner engagement, which includes logo rights, a co-authored case study, two investor reference calls, and four hours of founder time per month. Twenty minutes on the phone is the entire ask up front.

Apply to the pilot