What is Graph Theory?
The Technical Stuff
- A graph is a mathematical construct consisting of two types of elements: vertices (V) and edges (E). In the Connexity graph, consumers and products are vertices, connections and engagement are edges.
- The order of the graph is the number of vertices. The size of the graph is the number of edges. The Connexity graph can handle billions of vertices and trillions of edges.
- Each edge joins two vertices and represents a relationship or connection between them. Two vertices joined by an edge are said to be adjacent, or neighbors.
- Two vertices can have multiple edges connecting them. A multigraph is a graph that has multiple edges connecting the same vertices. The Connexity graph is a combination of three different multi-graphs.
- Edges can have labels with weights. The term network is a synonym for a weighted graph. Connexity graph edges are labeled and weighted.
- The degree of each vertex is its number of neighbors.
- The distance between two vertices is the length of the shortest path, or lowest count of edges, between them.
- Vertices adjacent to v not including v itself, form an open neighborhood. Connexity uses open neighborhoods to expand based on consumers who have converted but exclude the converted consumer.
- Connectivity extends the concept of adjacency and is essentially a form (and measure) of concatenated adjacency. The Connexity graph allows for custom settings of connectivity (distance and degree) for each campaign from zero connectivity (retargets only) to infinite connectivity (run of audience, ROA).
Wildfire is horizontally scalable; shared nothing
Wildfire is scalable to billions of nodes and trillions of edges
Wildfire supports concurrent updates, bulk traversals and prioritized (tactical) traversals
Updates can be bulk (core graph) or streaming (engagement)
The Wildfire bulk traversal algorithm substantially reduces system load
Wildfire manifests as directed or symmetric multigraphs with attributed nodes and edges
We start with a blend of audience data based on behavior, engagement, retargeting, transaction history, content,
context, geography, demographics, technographics, audience day part and channel.
The audience is then expanded from the “seed” or “core” engagement vertices into the hybrid multigraph based on
adjacency using either closed or open neighborhoods, to the degree and distance necessary to achieve the scale
and performance that meet the your needs. A custom audience develops for each campaign, the parameters of which
are editable in real-time to make adjustments on the fly. We believe this is the future of programmatic,
optimized audience buying.
Our optimization algorithms extract relevant explicit and implicit connections to identify friends, likes,
households, coworkers, shared hotspots, cross-channel activity, multi-device use and shared-user devices.
This multi-graph analysis applies a network architecture to the entire available audience and treats every
ad as a product recommendation.