What If Media Group Cuts Ad Costs by 33% with Machine Learning on Neo4j

In a world where the average person sees thousands of ads daily, most marketing messages become background noise. What If Media Group turns this digital chaos into opportunity by doing something radical: listening to what consumers actually want. The company analyzes billions of monthly advertising interactions in its Neo4j graph database to match relevant offers with audiences.

Every online interaction reveals something about what people want and need. A parent researching school districts signals an upcoming move. A homeowner browsing garden supplies in February shows early planning for spring. What If Media Group transforms these digital clues into connections, helping brands provide helpful information when consumers need it most. The firm’s ARIA platform doesn’t just track what consumers do; it understands the web of connections between people, their interests, and their behaviors.

When Uber needed new drivers, What If Media Group didn’t simply blast ads about driving opportunities. They analyzed patterns of behavior that indicated someone might be interested in flexible work opportunities, then provided targeted information about earnings potential and schedule flexibility. The result: tens of thousands of new drivers at 33% below Uber’s target acquisition cost.

Influencing That Next Customer Action with ARIA and Neo4j

“In the earliest stages of our business, we relied on a more human-driven process,” says Pershing.

What If Media Group initially outperformed competitors by manually segmenting and customizing email campaigns for different audiences. But as their data volume grew into the billions, this approach became a bottleneck.

Pershing and the team at What If Media Group saw an opportunity to combine information from proprietary and third-party sources, including consumer surveys and more than one billion monthly ad interactions from their network. Traditional databases struggled to handle these complex data relationships.

“We had to think outside the box, beyond table lookups, rows, columns, and spreadsheets. If you use a relational database, data warehouse, or even a NoSQL database, you have to extract all of that data and crunch it together,” explains Nick Mossie, Chief Architect of Messaging at What If Media Group. “We chose Neo4j because we can have all these joins from the start, sitting in a model in a data graph.”

This realization led to an expansion of ARIA, their AI-powered decision engine, to integrate a Neo4j graph database. ARIA can instantly analyze the intricate web of connections between individual consumers, their behaviors, and their preferences. When ARIA detects a signal of interest in a particular topic or product, it can deliver relevant marketing content within seconds—a critical advantage in today’s fast-paced digital environment.

Graph databases revolutionized how What If Media Group understood consumer similarities. “Neo4j helps us analyze how similar people interact with offers in our ad network. This helps us ultimately find the best offer for a consumer,” says Pershing. Unlike their previous systems that could only analyze data at a broad cohort level, Neo4j enables real-time offer recommendations by comparing a prospect’s interactions with extremely recent, similar interactions at the individual level.

What If Media Group chose Neo4j after evaluating several graph databases. “The Cypher query language is a standout differentiator that makes Neo4j dramatically easier for programming and running queries,” Mossie notes. Cypher is Neo4j’s declarative, schema-flexible graph query language, allowing users to unlock the full potential of property graph databases.

What If Media Group procured Neo4j Enterprise Edition through AWS Marketplace, making deployment and billing easy. “From our CFO’s perspective, the costs are all in one easily visible bucket tying the compute and the licensing together,” Pershing adds. Today, ARIA processes incoming activation signals using AWS Lambda, which connects to the Neo4j data graph to power its machine learning pipeline for offer recommendations.

The platform continues to evolve, with the data team recently enhancing their matching algorithms by incorporating Cosine similarity functions from the Neo4j Graph Data Science library, alongside their existing Jaccard similarity formulas. ARIA can identify subtle behavioral clusters that traditional analytics would miss — like someone researching insurance shortly after browsing moving companies — giving What If Media Group a competitive edge.

Matching Genuine Interest With Personalized Offers at Scale

These fine-tuned techniques for optimizing offers set What If Media Group’s ARIA platform apart.

“When consumers interact with our platform, we quickly discover their genuine interests. And if the Neo4j recommendation engine identifies a winning offer to present to a consumer, it’s usually right. That personalized offer has the highest probability of resulting in a conversion compared to other potential offers,” says Pershing.

Brand advertisers care most about driving results. The same goes for What If Media Group. When Uber needed to recruit qualified new drivers, What If Media Group earned a fee only after a driver took a passenger on a first ride. The company used ARIA to target likely drivers with ads tailored to their interests. Tens of thousands of new drivers signed on and Uber achieved a cost per acquisition (CPA) 33% below its target.

In another success, What If Media Group helped insurance firm Nationwide boost its customer acquisition volume by 30x while cutting its CPA rate in half. The mobile-first campaign generated millions of dollars in sales at a margin of over 100% for local insurance agents. Nationwide’s return on ad spend (ROAS) turned positive in less than one month.

“The more sophisticated we make ARIA, the more our brand advertising clients benefit. We can achieve better ROI for our clients by sending consumers fewer messages to achieve the engagement and conversion rates brands need,” Pershing says.

What If Media Group operates in a competitive, dynamic ecosystem of ad agencies and performance marketing firms. “If we’re not evolving our capabilities, our clients will move over to our competitors,” says Pershing. “With Neo4j, we can iterate quickly on the R&D side to do an even better job of offering consumers what they want. Our strong partnership with Neo4j frees us to focus on analytic innovation crucial to our success as we scale up performance marketing for brands.”

Get in Touch

Curious about what insights you could unlock for your business with graph-powered solutions? Let’s talk — reach out, and we’ll get in touch.