We were not really planning on a recap of the graph market. But sometimes news has a way of getting in the way of plans, and the news of TigerGraph announcing it has raised $105 million in Series C funding changed our plans.
TigerGraph is a graph database vendor we have covered since it exited stealth in 2017, and we see the progress it has made in a little over 3 years as a tell-tale for graph at large. TigerGraph’s Series C was led by Tiger Global and brings TigerGraph’s total funding raised to over $170 million.
This was the backdrop of our conversation with TigerGraph CEO Yu Xu and COO Todd Blaschka. We discussed TigerGraph’s evolution and the evolution of the graph landscape at large.
Graph is growing, and moving up the stack
The last time we connected with TigerGraph was almost a year ago, at the beginning of the COVID-19 crisis. In what feels like much longer than a year, TigerGraph has gone through the adjustment period many businesses have. Among them, data and analytics vendors may even be on the winning side results-wise, owing to the accelerated pace of digital transformation.
For TigerGraph, things worked that way, according to Xu. The company closed 2020 with the best quarters in its history. Xu and Blaschka went over different success stories, with customers ranging from Intuit and Jaguar Land Rover to the Australian Tax Office.
They also mentioned a number of use cases, ranging from the typical in graph such as customer 360 and supply chain analytics, to more unusual ones like Blockchain analytics and tax anti-fraud. That’s all good, but it almost begs the question: Why a funding round then?
There are a few things to take into account to be able to sketch an answer to that. The picture that emerges from TigerGraph’s experiences reaffirms what we have also seen with other vendors in that space: They are moving from databases to platforms, closer to solving customer problems and creating value.
Xu and Blaschka described how they saw getting a fast and scalable distributed graph database just as a starting point. That enabled them to get their foot in the door in many organizations, despite initially not having much of a name for themselves or success stories to show for. As Xu put it, organizations “had no other choice” but to adopt TigerGraph for a certain type of use cases.
Those use cases can be described as real-time graph analytics: getting answers that demand joining and going over many datasets, often massive ones, in near-real-time. In many cases, Xu said, TigerGraph was the only option for such use cases. Once it was adopted, customers also started using it for other use cases, and today TigerGraph is often adopted as an offline-analytics first solution too, Xu went on to add.
Moving up the stack for TigerGraph translates to things such as having added a visual IDE and querying capabilities, which the company aims to develop further, and expanding to areas such as what Xu referred to as “Graph Business Intelligence.” Xu detailed TigerGraph’s ambition to build a “Tableau for Graph.” Admittedly, this type of ambition probably requires funding to fuel.
But that is not to say that TigerGraph does not have more down to earth, operational aspects in its roadmap. TigerGraph has been running its database-as-a-service offering for a while, with support for AWS and Microsoft Azure. Adding Google Cloud support, as well as expanding its team to meet increased product demand is in the company’s plans, but there’s more.
The years of the graph are just beginning
When discussing its cloud offering, TigerGraph executives mentioned that they don’t just want to add Google Cloud support, but also to add more features and better integration to their existing AWS and Microsoft Azure layers. Discussing what that may include, Xu emphasized integration with cloud vendor-supported machine learning libraries as a prime example.
Xu noted that the integration of machine learning capabilities is happening across a wide range of data management platforms, citing the example of Google’s BigQuery. The idea is simple – to be able to shorten the data pipelines needed to process data for machine learning. The goal is to make the job of data engineers and data scientists a bit easier.
The way to do this is by integrating machine learning-oriented extensions in SQL, for example, said Xu. TigerGraph has its own query language, called GSQL, but this idea is something they have been working on for a while. In fact, there is an additional reason for graph vendors to do that.
As we have noted, and Xu confirmed, graph-based machine learning is an area seeing huge traction. Put simply, graph-based machine learning is about working with multi-dimensional data and leveraging connections, rather than reducing everything to 2 dimensions. So it makes sense to use a graph platform for this.
Speaking of graph query languages, however, Xu also referred to GQL. GQL is the standardization effort for graph query languages currently underway under the auspices of ISO, with support from many vendors. Since we have not had much news from that front for a while, we wondered what the status is.
Xu was reassuring, mentioning that GQL is making solid progress, and we may see results even within 2021. Like all standardization efforts, things tend to move a bit slow. Considering how many people and vendors are involved, that is to be expected. Xu went on to add this is only the second query language to be standardized by the ISO in 40 years, after SQL.
Another point Xu made on GQL is that graph is not like key-value or document databases, which don’t have a standard query language and possibly don’t need one. Graph is a richer, more complex data model, richer than relational as well, and accessing it programmatically makes less sense.
Does that mean that organizations are ripping and replacing their legacy relational databases in favor of graph? Not exactly, at least not yet, but that’s fine. Xu referred to examples in which TigerGraph is operating as a system of record, mentioning however that the focus remains on analytics. That said, however, more and more applications going forward will be graph-first.
The years of the graph are just beginning.