Fracture PointDetecting…
Detecting story elements0 coded
Core audience
A18–49 sci-fi thriller
0
Top driver
City Catastrophe
0
Top storyline
One man moves while the world freezes
0
Top comp
Grounded sci-fi mystery
0
Freshness
Time-fracture visual hook
0
Hybrid: Panels + AI

The best of traditional research and predictive AI. Finally.

Vault fields traditional panel research through the vendors you trust, runs real predictive AI on the same title, and aggregates both into one report. Built for the calls creatives actually have to make.

Book a Demo → See how it works
How it runs

Traditional research and predictive AI. One read.

Vault fields the traditional panel through established research vendors and runs the predictive AI study in parallel, same title, same week. We aggregate both into a single, calibrated read. Your team gets one answer, instead of two reports to reconcile.

TRADITIONAL PANEL EXECUTION · NUANCE · LIVE VOICE PREDICTIVE AI UNBIASED · CALIBRATED · AT SCALE THE CUT Rooster S1 PILOT JM AR TK SP STAGE 01 Recruit + field 1,840 RESPONDENTS STAGE 02 Live reactions DIAL · OPEN-END STAGE 03 Voice captured WHAT THE AUDIENCE SAID 1,840 LIVE VOICES STAGE 01 Read 60K dataset COMP SET · BENCHMARKS STAGE 02 Score the drivers STORY DNA · PULL STAGE 03 Forecast every audience EVERY DEMO, EVERY MARKET POPULATION SCALE VAULT Aggregates ONE Report PRE-SYNTHESIZED THE CUT Rooster S1 TRADITIONAL PANEL EXECUTION · NUANCE · LIVE VOICE STAGE 01 · RECRUIT + FIELD 1,840 respondents in field STAGE 02 · LIVE REACTIONS Dial trace and open-end STAGE 03 · VOICE CAPTURED What the audience said PREDICTIVE AI UNBIASED · CALIBRATED · AT SCALE STAGE 01 · 60K DATASET Comp set and benchmarks STAGE 02 · SCORE DRIVERS Story DNA and audience demand STAGE 03 · FORECAST Every demo, every market VAULT Aggregates ONE Report PRE-SYNTHESIZED FOR THE ROOM
Distinct methodologies
Run in parallel
Aggregated by Vault
The traditional side

Traditional panel research, fielded by the vendors you trust.

Vault commissions panel work through the established research vendors the industry already uses. Real humans, recruited through established vendors. Your creative stays inside the panel partners and security protocols your research team has already cleared.

+Fielded by established research vendors. Same screeners, same field rigor.
+Live reactions captured during the watch.
+Your creative ships through panel vendors your team has already cleared. No new vendor to vet, no new security review.
Verified respondents · established vendors
In field
n = 1,840 respondents recruited per study. All human, all verifiable.
The AI side

Not a chat wrapper.

Plenty of platforms now claim to have AI. Most are pointing a general-purpose chat model at last year’s headlines and asking it to guess. Vault is a domain-specific predictive system trained continuously on every real audience outcome since 2015. Calibrated on the metrics buyers and bosses actually use to defend a call.

+300+ specialised models, not one general-purpose chat model.
+Calibrated on real audience outcomes.
+Retrains continuously as new titles close the loop.
Why hybrid

For creatives to move the needle, there is no other way.

Pure panels miss the scale. Pure AI misses the room. Synthetic panels miss the humanity. Vault Hybrid is the only approach that delivers all three, with the bias of each method checked by the other.

Limits the bias.

The panel keeps the AI honest. The AI keeps the panel scaled. Two methodologies, each correcting the blind spots of the other, vocal-minority bias, sample bias, demographic skew, all checked.

The best of both worlds.

The voice and the forecast. The texture and the scale. The lived reaction and the population read. No trade-off between rigor and reach. Both, on one report.

Built for the creative.

Iterate the cut. Test the scene. Score the trailer. For the first time, the creative gets a tool that helps them move the needle before launch.

There is no other way.

Synthetic panels miss humanity. Pure panels miss the slate. AI-only platforms miss the room. Only Vault Hybrid delivers all three at once.

What everyone else is missing

Only Vault sits in the middle.

A real panel tells you what 1,000 viewers felt. Predictive AI tells you what the next 10 million will look like. The decision-maker brings the craft and pattern memory the model cannot replace. Vault is the only place all three meet on the same dashboard.

PANEL · SURVEY 100s to 1,000s OF VIEWER VOICES PREDICTIVE AI CALIBRATED ON INDUSTRY CURRENCY THE DECISION-MAKER RESEARCH · MARKETING · DEVELOPMENT DATA + ALGORITHM TRADITIONAL RESEARCH NO PREDICTIVE LAYER STRATEGY + MODEL VAULT HYBRID TESTING ALL THREE
Panel · Survey
100s to 1,000s
Vault runs these in the field, same recruit, same screener, same field rigor as the leading panel providers. Real viewers, real answers, captured live. Not simulated, not modelled off someone else’s panel.
Predictive AI
Millions
Calibrated on the industry’s real currency: viewership, demo lift, audience demand. 60K+ comp set, population-scale forecasts, on the metrics buyers and bosses actually use to defend a call.
The decision-maker
The room
The research, marketing, or development exec running the work. Brings craft, instinct, and pattern memory the model cannot replace, and uses Vault to navigate the next move.
“
Mixing traditional with Vault’s predictive AI is a game changer. It allows us to get the voice and the accuracy of viewership that no other market research firms can enable. Meanwhile, we still get the insight that AI cannot match. It’s the only product we use.
Head of Studio
The whole picture

Traditional gets you some. Vault gets you the rest. Hybrid gets you all of it.

Twelve things a panel or a model alone can deliver. Only Hybrid hands you every one of them on the same page.

What you get
Traditional
Panel only
Vault
Predictive AI only
Only way to get all of it
Vault Hybrid
Panel + Predictive AI
Real people in the field
Live, in-the-moment reactions
Open-end quotes and detail
Sentiment and emotional read
Predictive performance scoring
Driver and theme extraction
Comp set benchmarking · 60K titles
What-if simulations
Reach beyond home market
Bias detection and correction
Cost-efficient at scale
Speed to insight
Only a panel gives you this
Traditional
Vault
Vault Hybrid
  • Real people in the field
  • Live, in-the-moment reactions
  • Open-end quotes and detail
  • Sentiment and emotional read
Only predictive AI gives you this
Traditional
Vault
Vault Hybrid
  • Predictive performance scoring
  • Driver and theme extraction
  • Comp set benchmarking · 60K titles
  • What-if simulations
  • Reach beyond home market
Where traditional falls short
Traditional
Vault
Vault Hybrid
  • Bias detection and correction
  • Cost-efficient at scale
  • Speed to insight

Frequently asked questions

What is Vault, in one paragraph?

Vault is an AI audience prediction platform built on a world model trained on 60,000+ titles across film, TV, and games. We turn any input, whether a logline, script, trailer, slate calendar, or a question in plain English, into an audience prediction calibrated against every release that came before. Studios, networks, and streamers use Vault to pressure-test before they greenlight, position before they launch, and simulate before they release.

How accurate are the predictions?

Vault AI maintains over 85% average accuracy across core predictions. Accuracy is measured by comparing every prediction against actual in-market performance once a title launches.

How is Vault different from a general LLM?

General LLMs are trained on the open web. They generate text that sounds plausible. Vault is a domain-specific world model trained on a structured dataset of every released title and every signal we could index against it. The model returns predictions calibrated against ground-truth performance data, with the comp evidence one click underneath every score. We don’t hallucinate audience numbers.

How long does setup take?

Most teams are running their first reads in week one. Full integration with internal slate data, brand-specific comps, and custom audience targets is in place in four to eight weeks depending on data scope and security review.

Where does my data sit?

Customer data lives in a dedicated single-tenant environment. Your scripts, slate, and proprietary research never train the public model. SOC 2 Type II, GDPR, and CCPA compliant. Region-specific hosting available on request.

What inputs does Vault accept?

Anything that carries story signal. A two-line logline, a full script, a treatment, a trailer cut, a poster, a slate calendar, or a question in plain English. Vault extracts the DNA from any input and pattern-matches it against the content universe.

How do you measure model lift for our team?

Most enterprise customers run a calibration study in the first 90 days. Your team’s last 12 months of decisions, scored retrospectively against the model’s predictions, plus a forward holdout of 5 to 10 active titles. We publish the lift study to your team and use it to tune the model to your slate.

VaultAI
Predict

See how Vault sharpens your decisions

Book a demo →