How a DOOH Ad Platform Replaced Unsellable Inventory
An ad platform team layered a custom image-expansion model and a 96%-accurate moderation algorithm into uploads in two weeks — turning every image into a billboard-ready, revenue-generating ad.
Significant Revenue Lift
From unsellable inventory
Arman Hezarkhani
Co-founder & Managing Partner
Tenex
Connect ↓
The Challenge
A digital out-of-home advertising platform was losing revenue every time an advertiser uploaded an image. If dimensions didn't match available billboard formats, the ad couldn't run — and inventory went unsold. Making it worse, human moderators reviewed every creative manually before it could go live, creating a slow, expensive bottleneck between upload and launch. Every hour of delay was another hour of unmonetized inventory sitting idle.
What They Built
Tenex built a custom AI model that automatically expands any uploaded image to fit any available billboard format — turning every upload into a viable ad unit. Alongside it, they developed a proprietary content moderation algorithm that handles the first review layer with 96% accuracy, dramatically reducing reliance on manual review. The system deployed in just two weeks. What started as a revenue-recovery tool became a platform differentiator: every image that walks in the door now walks out as revenue-generating inventory, with compliance handled before a human ever needs to look.
Tenex identified two compounding revenue problems: dimension mismatch leaving billboard inventory unsold, and slow manual moderation delaying ad launches. Rather than patching the existing process, they built two parallel AI systems. The first was a custom AI image expansion model that takes any uploaded creative and automatically generates versions sized for every available billboard format — turning previously unusable uploads into viable ad inventory. The second was a proprietary content moderation algorithm built to handle the first review layer with 96% accuracy, dramatically reducing the volume reaching human reviewers and accelerating time-to-launch. The two systems were designed to work together: an image uploads, expands into all formats, passes first-pass moderation, and surfaces for human review only when flagged. Both systems were built and deployed in two weeks. The immediate impact was the monetization of previously unsellable inventory, generating a significant revenue lift across the platform. What began as a revenue-recovery project became a differentiated platform capability.
Infrastructure
Custom AI image expansion model (proprietary, multi-format output) • Proprietary content moderation algorithm (96% first-pass accuracy) • Digital out-of-home ad platform (client-side infrastructure) • Human moderation review interface (for flagged content)
Integration Points
Advertiser image upload → AI expansion model → format-matched ad units for all billboard dimensions • Expanded ad units → content moderation algorithm → pass/flag determination • Flagged content → human review queue → approval or rejection • Approved units → platform inventory → live ad serving
Impact
Previously unsellable inventory — ads that couldn't run due to dimension mismatches — became monetizable, increasing revenue per upload across the platform.
A custom-built AI moderation algorithm achieved 96% accuracy on first-pass content review, replacing a slow manual process without sacrificing compliance standards.
The full solution — image expansion model plus moderation algorithm — went from concept to live production in under two weeks.
Implementation Complexity
Best Fit For
Digital advertising platforms, DOOH networks, and ad-tech companies struggling with creative compliance bottlenecks or inventory utilization gaps.
