Pixel-Level Viewability Methodology | 4D Sight Docs

Why Pixel-Level Measurement

Traditional sports sponsorship measurement was built around media-equivalent value: count the seconds a logo is on screen, multiply by the cost of an equivalent TV spot, ship the report. The number was always flattering and almost never auditable. Virtual advertising changes the unit of analysis. Because every sponsor frame is rendered into the broadcast by software, every frame is also instrumented by software — and the measurement product can finally be as precise as the inventory it measures.

4D Sight measures every placement at the pixel level, every frame, in every encoded output. There is no panel, no statistical inference, no model that approximates exposure. The signal is the ground truth of the broadcast itself.

The Pipeline

For each rendered frame, the measurement pipeline runs four passes:

Surface detection — computer vision identifies the exact pixel polygon occupied by the virtual surface (perimeter board, pitch carpet, octagon canvas, courtside apron) after camera projection and lens distortion.
Sponsor segmentation — the rendered creative is matched to the surface polygon, producing a per-sponsor bounding mask in screen space.
Occlusion analysis — players, referees, on-screen graphics, and overlays are subtracted from the mask. What remains is the pixels actually visible to the viewer.
Frame logging — the visible pixel count, surface area percentage, occlusion percentage, and on-screen duration are written to the impression ledger with a frame-accurate timestamp and the feed identifier.

The Viewability Formula

The viewability percentage 4D Sight reports per impression is not a flag, it is a continuous value:

viewability_pct = (visible_pixels ÷ rendered_pixels) × on_screen_ratio × 100

Where visible_pixels is the post-occlusion pixel count, rendered_pixels is the surface mask before occlusion, and on_screen_ratio is the fraction of the impression window the surface remained in frame. A logo briefly clipped by the broadcast graphic, partially blocked by a player, and visible for 70% of its window will report a specific viewability number — not "yes" or "no."

Alignment with IAB and MRC

The methodology aligns with the IAB Display Viewability Guidelines and the MRC Cross-Media Audience Measurement Standards by adopting the same continuous-pixel basis those standards use for digital display. Where the sports broadcast context required extension — primarily around occlusion handling and per-feed regional rendering — the deviations are documented in the technical addendum, not buried in the report. Auditability is the point of the spec; vendor-specific math defeats it.

What Gets Reported

For every impression, the platform exposes: event_id, broadcast_id, feed_id, placement, sponsor, region, viewability_pct, on_screen_ms, surface_px, occlusion_pct, brand_safety_state, and frame_timestamp. Every field is queryable, every field is exportable, and every field is reconciled nightly against the encoder log.

What This Replaces

This methodology retires three categories of legacy report: "estimated visibility" (which was always a function of duration alone), "media equivalent value" (which was always a function of estimated visibility), and "share of voice" (which was always a function of media equivalent value). The new chain runs from rendered pixel to dollar — and every step is reviewable.