AI-Powered Audience Analytics
Know Who's Watching.
Prove What Works.
ScreenIQ tells you who looked at your screen, for how long, and whether the content landed. Runs on-device. No footage leaves the hardware.
Capabilities
What ScreenIQ Measures
All data is anonymous. No images ever leave the device.
Audience Demographics
Age group, gender. Per viewer, per session. No surveys, no apps, no opt-in required.
Eye Tracking
Did they actually look at the screen? For how long? ScreenIQ measures gaze, not just presence.
Sentiment Analysis
Smiles, surprise, confusion. Picked up in real time and matched to whatever was playing.
Dwell Time
Down-to-the-second measurement of how long each person stays in front of your display.
Engagement Score
A single 0–100 number. Blends attention, dwell, and reaction so you can compare content at a glance.
Footfall Counting
Counts everyone who passes by, even from behind or at an angle. That gives you true OTS numbers.
Staff Filtering
Employees get filtered out automatically. Your reports show customers, not the guy restocking shelves.
A/B Content Testing
Run two creatives, see which one pulls better numbers. Split by age, gender, time of day.
Zone Intelligence
Draw zones on a floorplan. Track traffic, attention, and dwell per zone. Find the dead spots.
Opportunity to See
Were they close enough, facing the right way, long enough to actually see it? That's what OTS answers.
Return Visitors
Session-based frequency tracking. No personal data stored; the device just knows someone came back.
Real-Time Dashboard
Live view, historical charts, CSV exports, heatmaps. One dashboard for your whole network.
AI Pipeline
8 Models. One Pass. Every Frame.
Each camera frame runs through a stack of specialized vision models — all on-device, all in real time. Here's what fires and in what order.
Face Detection
Locates every face in the frame. Works at angles up to 60° off-axis, under mixed lighting, and at distances from 0.5 m to 8 m.
Face Tracking
Assigns a temporary ID to each face so metrics stay tied to the right person across consecutive frames. ID dies when the viewer leaves.
Demographics
Estimates age bracket and gender from facial geometry. No biometric template is stored — the model outputs a category, not an identity.
Gaze Estimation
Calculates where the viewer is looking relative to the screen surface. This is what separates "walked past" from "actually saw it."
Attention Scoring
Combines gaze direction, head pose, and dwell to produce a per-viewer attention score. Updated every 200 ms.
Sentiment Analysis
Reads facial micro-expressions — smile, surprise, confusion, neutral. Mapped to the content playing at that exact second.
Staff Filtering
Identifies recurring faces that appear daily and flags them as staff. They're excluded from audience reports automatically.
Aggregation & Upload
Raw detections get compressed into anonymous statistics — counts, distributions, scores. Only these numbers leave the device.
⚡ Single-pass architecture: all 8 models run in one forward pass per frame. No queuing, no frame drops at 15 fps on standard edge hardware.
Dashboard
One Screen for Your Whole Network
Live numbers, historical trends, demographic splits. Exportable. Filterable by location, time, and audience segment.
Product Tiers
Pick the Depth You Need
Every tier runs the same edge hardware. The difference is which models are active and what data you get back.
Basic
- ✓ Footfall counting
- ✓ Opportunity to See (OTS)
- ✓ Basic demographics
- ✓ Hourly traffic reports
Standard
- ✓ Everything in Basic
- ✓ Gaze tracking & attention score
- ✓ Dwell time measurement
- ✓ Sentiment analysis
- ✓ A/B content testing
- ✓ Staff filtering
Premium
- ✓ Everything in Standard
- ✓ Zone intelligence
- ✓ Return visitor frequency
- ✓ Real-time API & webhooks
- ✓ Custom model training
- ✓ Dedicated support SLA
No hardware purchase required
We supply the edge device as part of the subscription. You provide power and a USB or IP camera — that's it. No capital expense, no procurement cycle.
Billing Models
Pay for What You Measure
Cost per Player
Flat monthly fee per screen. Predictable costs regardless of traffic volume. Best for networks with stable screen counts.
Cost per Mille
Priced per thousand verified impressions. Aligns your analytics cost directly with audience delivery. Ideal for ad-funded screens.
Cost per Viewer
Charged per unique viewer session with attention confirmed. You only pay when someone actually looks. Built for performance-driven campaigns.
Who Uses ScreenIQ
Your Screens Already Have an Audience. Now Prove It.
Retailers, DOOH operators, and brands use ScreenIQ to stop guessing and start measuring what their screens actually deliver.
Retail Media Networks
You have screens in stores. Advertisers want proof that people saw their ad. ScreenIQ gives you CPM-grade metrics: verified impressions, demographics, attention duration. The same data they get from programmatic, except this is physical.
Turn existing screens into sellable mediaDOOH Operators
Mall, transit, QSR, public venue. Doesn't matter. You get standardized impression and attention data across all of them. Price your inventory on who actually saw it, not how many people walked past the building.
Back your rate card with real numbersBrand & Shopper Marketing
You're spending on in-store screens but reporting is a black hole. ScreenIQ shows exactly which creative worked, for which audience, at which location. Run proper A/B tests instead of rotating content and hoping.
Replace gut feel with measured performanceRetail Operations & CX
Zone-level traffic data, peak hours by area, dwell time by display. Use it to rethink screen placement, adjust staffing, or figure out why aisle 7 gets zero attention after 4pm.
Decisions backed by store-level dataConnects with your existing stack
Programmatic Ready
Plug Into the Programmatic Stack
ScreenIQ outputs impression data in formats that programmatic platforms already understand. No middleware translation, no custom adapters.
OpenRTB Compatible
Impression events follow OpenRTB 2.6 structure, ready for SSP ingestion without mapping layers.
VAST / VPAID Hooks
Video ad events (start, quartile, complete) fire in real time, tied to verified viewer attention data.
Audience Segments
Demographics and attention scores exported as audience segment tags. Feed them straight into your DMP or DSP.
Real-Time Bidding Data
Live audience counts and composition available via API — SSPs can factor real-time foot traffic into bid floors.
Metric Definitions
What the Numbers Mean
Every metric has a clear definition and a documented methodology. No black boxes.
OTS (Opportunity to See)
A person was within viewing range, facing the screen, for at least 1 second.
distance ≤ 8m AND head_angle ≤ 60° AND dwell ≥ 1s Verified Impression
A person looked directly at the screen for at least 2 consecutive seconds.
gaze_on_screen = true AND duration ≥ 2s Dwell Time
Total time a viewer spent within the detection zone, regardless of gaze direction.
exit_timestamp − entry_timestamp (seconds) Attention Time
Subset of dwell time where gaze was confirmed on-screen. Always ≤ dwell time.
Σ(gaze_on_screen intervals) within dwell period Engagement Score
Composite 0–100 index combining dwell, attention ratio, and sentiment signal.
0.4 × dwell_norm + 0.3 × attention_ratio + 0.3 × sentiment_pos Audience Composition
Percentage breakdown of viewers by age bracket and gender for a given time window.
count_per_segment ÷ total_viewers × 100 Return Rate
Share of viewer sessions matched to a previously seen face template within 30 days.
matched_sessions ÷ total_sessions × 100 (rolling 30d) Questions You Can Finally Answer
- How many people actually see my screens each day?
- Are they looking, or just walking past?
- Which creative gets the most attention?
- Does the audience change between morning and evening?
- Am I reaching the right demographic?
- What happens when I swap content A for content B?
How Engagement Is Scored
Each viewer gets a 0–100 score, weighted like this:
Privacy by Design
Built for Compliance
No images leave the device
Video gets analyzed on the spot and thrown away. The cloud only ever sees aggregate statistics.
No personal data collected
No names, no photos, no biometric templates. Age goes into a bracket. Identifiers die with the session.
No cross-session tracking
Fresh encryption key every session. There's no way to link yesterday's visitor to today's. That's the point.
No visitor consent required
Purely anonymous statistical processing. Qualifies under GDPR legitimate interest, Article 6(1)(f).
How It Works
Edge AI. Nothing Leaves the Box.
A small device sits behind the screen. It runs all the vision models locally. No video streaming, no cloud dependency, no bandwidth hit.
Works offline
no internet needed for analytics
Plug and play
up and running in minutes
OTA model updates
new capabilities ship remotely
Any scale
one screen or ten thousand
Camera-agnostic
USB, IP, whatever you have
Always on
self-recovers under load
Good to know: We help you spec the hardware. It scales with the project.
Our Standard
We don't sell dashboards.
We ship measurement that works.
If it can't run across a thousand screens, it's a prototype.
If it can't explain why, it's a black box.
If it compromises trust, it doesn't belong in physical space.