Building intelligent search for out of home advertising

Setting the context

Times OOH manages a large and diverse inventory of advertising sites. Historically, planners relied on structured filters and keyword search to identify locations. While functional, this model struggled when users searched with contextual intent for example, “premium high-traffic corporate zones near tech parks” rather than exact site attributes.

As inventory datasets grew more complex, discovery became increasingly time-consuming. Manual filtering, spreadsheet exports, and internal coordination slowed campaign turnaround cycles and risked underutilization of premium inventory.

The business required a smarter, more intuitive search layer capable of understanding meaning not just matching text.

Why this mattered

In media and advertising, speed and precision directly impact revenue.

Manual search processes consumed valuable planner bandwidth.
High-value inventory could remain undiscovered.
Sales responsiveness was constrained by slow data exploration.
Campaign planning cycles stretched longer than necessary.

Without AI-driven discovery capabilities, scaling inventory would only compound inefficiencies. A modern search infrastructure was critical to unlocking operational leverage and competitive advantage.

The constraint

The challenge was both technical and strategic.

Site data existed in structured and semi-structured formats. Converting multiple attributes location metadata, format types, audience metrics, and descriptive fields into meaningful vector embeddings required careful data modeling.

The POC had to:

• Deliver relevant semantic results without large pre-trained proprietary datasets.
  
• Handle ingestion from CSV-based uploads.
  
• Maintain enterprise-grade data security.
  
• Be deployed rapidly to validate business feasibility.
  

Accuracy, speed, and scalability had to be engineered simultaneously.

Our approach

Saguna Consulting architected and implemented an end-to-end AI-powered search framework designed for rapid validation and long-term scalability.

Solution Architecture Included:

• Data ingestion pipeline for structured site uploads
  
• Preprocessing and normalization workflows
  
• Vector embedding generation using NLP-based models
  
• Similarity search powered by vector indexing
  
• Interactive search UI with dynamic filtering and visualization
  
• Secure cloud deployment environment
  

The platform enabled natural language queries and contextual matching, replacing rigid keyword dependency with semantic similarity scoring.

Tech stack

• Python-based data processing pipelines
  
• NLP embedding models (transformer-based architecture)
  
• Vector database / similarity indexing engine
  
• RESTful API layer for search queries
  
• React-based front-end interface
  
• Secure cloud infrastructure deployment (scalable containerized environment)
  

The POC was designed with production scalability in mind—ensuring that expansion to larger datasets and additional AI capabilities would not require re-architecture.

The impact

The results demonstrated immediate operational and strategic value.

• POC delivered in 7 days
  
• Estimated 30–40% reduction in manual search effort
  
• Instant similarity search across uploaded datasets
  
• Significant qualitative improvement in result relevance

Media planners could now search using intent-based language.
Inventory discovery became faster and more intuitive.
Leadership gained confidence in AI-enabled innovation within core revenue workflows.

The POC successfully validated semantic search as a viable foundation for next-generation inventory management.

Client comment

“Saguna Consulting helped us move beyond traditional search thinking. In a very short time, they demonstrated how AI could meaningfully enhance inventory discovery and planning efficiency. The speed and technical depth of execution gave us confidence to take the next step toward production.”

Looking ahead

The Vector Search POC now serves as a strategic foundation for a full-scale AI-powered search platform.

Planned next phases include:

• Expanded dataset ingestion at enterprise scale
  
• Geo-spatial mapping and advanced filtering
  
• AI-driven site recommendations
  
• CRM and sales planning system integration
  
• Real-time performance analytics integration
  

What began as a Proof of Concept has evolved into a roadmap for intelligent, scalable inventory discovery.

A smarter way to search. A stronger foundation for scale.