OmniTrend provides real-time transparency into political advertising by tracking spend, audience targeting, and thematic messaging across fragmented digital ad networks. Elementica led the product from initial concept to production — architecting a machine-learning web scraping pipeline, a time-series ad performance database, and a multi-platform ad creation interface that unifies campaign management across Facebook, Taboola, Outbrain, and Gemini.
Political Intelligence/AI/Analytics/SaaS
Real-time performance tracking via continuous automated scraping and time-series ad spend ingestion
Demographic rule targeting — granular audience slicing mapped directly to strategic campaign briefs
Concept-to-production build with complete technical ownership of architecture, design, and CI/CD deployment pipelines
Client context
Digital political advertising in Australia is spread across competing ad networks, making it difficult for strategists to monitor competitor messaging, analyze public spend patterns, or track demographic targeting strategies in real time.
OmniTrend was conceived to bring structure to this chaotic landscape. The goal was twofold: construct a high-throughput monitoring platform that ingests and categorizes active political ads across the web, and build an integrated campaign management engine allowing political teams to launch targeted multi-network campaigns from a single console.
The challenge
Building an intelligence platform over third-party advertising networks introduced complex data engineering hurdles:
- Unstructured & Volatile Data Streams: Scraping and parsing real-time ad copy, imagery, and target demographic metadata across different networks required processing vast amounts of unstructured media.
- Entity & Theme Classification: Automatically identifying which political figures, parties, or policy themes (e.g., climate, tax reform, healthcare) were featured in an ad required natural language processing and computer vision models.
- Unified API Abstraction over 4 Networks: Combining ad creation workflows for Facebook, Taboola, Outbrain, and Google Gemini into a single unified input schema required mapping four fundamentally different ad network APIs into one interface.
Elementica's role
Elementica acted as the end-to-end technical lead. We established system specifications, designed the database models, engineered the Python background ingestion pipelines, integrated AI classification APIs, and developed the React frontend dashboards.
The solution
Automated Scraping & ML Categorization Pipeline
We engineered an automated ingestion engine powered by Apify scrapers and Python Celery workers. Raw ad creatives are passed through HuggingFace and OpenAI models to classify key political figures, sentiment, and core policy topics. The structured results feed into an administrative workspace equipped with dynamic charts and campaign relationship graphs.
Unified 4-in-1 Ad Creation Console
Rather than forcing campaign staff to log into four separate ad management platforms, we architected a single campaign creation workspace. Campaign managers upload creative assets, write ad copy, and define budgets once; our backend translates the payload and pushes campaigns out to Facebook, Taboola, Outbrain, and Gemini simultaneously via custom API wrappers.
Rule-Based Audience Targeting Engine
We developed a targeting interface that converts strategic briefs into executable audience logic. Campaign managers can define precise demographic slices — such as "Voters aged 40+ interested in renewable energy" — which the system automatically maps to corresponding targeting parameters on each destination ad platform.
Engineering decisions that mattered
TimescaleDB for high-throughput time-series metrics
Ad performance data (impressions, clicks, spend velocity) consists of high-volume, time-stamped events. Standard relational databases slow down under real-time aggregation queries across millions of records. We selected TimescaleDB (a time-series extension for PostgreSQL), enabling sub-second analytical queries over rolling time windows while keeping full SQL support for relational campaign metadata.
Decoupled modular architecture for SaaS spin-off flexibility
We designed the ad-creation engine as an isolated microservice with its own API boundaries. This decoupled structure allowed OmniTrend to evaluate spinning out the campaign creation module into an independent, standalone commercial SaaS tool without needing to decouple it from the main analytics platform later.
Results
OmniTrend operates as a live data platform offering political strategists real-time visibility into digital ad footprints. Its automated pipelines continuously track ad spend and campaign messaging across Australia, backed by automated CI/CD deployment pipelines on AWS.
Product evolution
The unified ad-creation console remains a candidate for expansion into an independent commercial product, while the analytics platform continues to expand its machine-learning models for deeper real-time sentiment analysis.
