Thinking in Graphs

A Tech Talk, 3/3/23

What We'll Cover

What data graphs are / can be
How to think about your data in a graph
How data graphs differ from other data modeling approaches
The value of data graphs
When to use a data graph

What is a data graph?

A data graph is an approach to structuring data that is focused on relationships between entities.
It's non-linear architecture allows for more flexibility and enables you to create connections between disparate pieces of data to provide new and powerful insights.
When we talk about data within a data graph we generally refer to entities as "nodes", and connections between entities as "edges".

How do they work?

Based on graph theory, a mathematical approach to model pairwise relations between objects.
It connects vertices in a network through edges.

How we got here

Driven by a more recent change in how we think about data, which impacted how we structure it.
Thinking shifted from how to most efficiently store data to how to get the most value out of it.
Realization that data is inherently more valuable when it is connected.

The gold rush

GraphQL

Stardog

DGraph

Notable implementations

A shift in mindset needs to occur in order to see the value [in data graphs]. This mindset is a shift from thinking about your data in a table to prioritizing the relationships across it.

- Matthias Broecheler

Shifting our mindset

Trees vs. Graphs

All trees are Graphs, but not all graphs are trees...
Trees are linear, they have a single root node and flow in a single direction.
Can connect to other leaf nodes or more subtrees, but connections happen in a specific order.

Trees vs. Graphs

Graphs are omnidirectional
They have no defined end or beginning, they have multiple "entrypoints"
Connections to nodes within a graph can be

Multiple Entrypoints

Relational vs. Graph

Relational model uses foreign keys to relate data together.
Relational focuses on storing tabular data for people, places, and things.
Graphs make it easier to create richer data connections with edges.

What does a healthy data graph look like?

Many connections on each node, and many paths to each node through various entrypoints.
Common entities referenced throughout the graph, enriched with relationship-specific data via edges.
Few if-any "dead ends" within the graph where nodes are only accessible via a single entrypoint.

Edges

Edges are the connections between nodes within a graph
They often contain data that can be attributed to the relationship of the nodes
You can have directed or bidirectional edges within a graph

Bidirectional Edges

Bidirectional edges represent connections to nodes within a graph regardless of direction.
In this example, the relationship between me and my dad can always be considered "family".

Directional Edges

Directional edges represent connections to nodes where directionality does matter.
In this example we are looking at a graph of twitter followers. Me and Elon Musk are followed by Russian bots, and I follow Elon musk, but Elon doesn't follow me and neither of us follow the Russian bots.

The value of a data graph

Allows us to enrich connections from disparate data sources with highly valuable information.
We can use edge data to personalize experiences for the user.
Personalized experiences can lead to better customer engagement & ultimately revenue.

How we leverage data graphs at NerdWallet

{
  review {
    title
    authors {
      name
    }
    categories {
      id
      articles {
        title
        authors {
          name
        }
      }
    }
    marketplaceEntity {
      id
      name
      review {
        id
        title
      }
    }
  }
}

When to leverage a data graph

Relationships within your data help you understand your problem better.
A lot of the data you need is derived from the relationship between entities.
Most of the complexity within your implementation will be around creating relationships between entities.
Your entity models change often, and are highly connected to other models.

Takeaways

Data graphs can be an effective tool for extracting more value from your data.
Data graphs can help create more meaningful connections by enriching them with edge data.
When working with data graphs we need to stop thinking about data linearly, and embrace graph thinking when modeling nodes and edges for a data graph.

Thanks for coming! Questions?