Neo4j CEO talks growing enterprise graph adoption and why partnering with Google makes sense

I last met Emil Eifrem, the highly energetic and persuasive founder and leader of graph database and software tools firm Neo4j for diginomica back in May 2017. As he was quick to point out in our latest meeting, a lot’s changed for his company and the market he’s tried to so hard to dominate.

But then, as he also said, a lot’s changed for us all in business in the past few months anyway, due to various external business and epidemiological factors:

If you just look at the news the macro uncertainties are definitely a lot higher now than I thought even at Christmas time. Coronavirus is the definition of the Black Swan event, coming in out of nowhere. So as you can't control stuff like that, I tend to focus on what we can control, which is how we run our company; we've raised an additional $80 million worth of capital since we last spoke, the vast majority of which is still in the bank. We are not profitable today, we could be if we want to, but now is the time to invest.

Invest in what? Back in 2017, Neo4j and graph were still being mostly ignored by the enterprise-level CIO, as well as analysts, with most of the interest in NonSQL databases going to the likes of Cloudera (hey, remember ‘Big Data’?). Things may have changed here, as we’re about to see. But two big promises were made by Eifrem from the stage of ‘GraphTour’ in London - that his company would be profitable in 2018 and that it would definitely IPO by 2020, too.

By his own admission, those are things we won’t see for a while, so what is he investing in? Let’s unpack his reasons for believing that it’s still very much worthwhile leading a native graph company in a world still dominated by traditional SQL relational database management systems, and that there are good times ahead, Black Swan events or not.

Enterprise interest in graphs at last...

First off, between 2017 and the end of 2019, graph finally started to find its feet in the Fortune 500. Neo4j won new customers in financial services (JPMorgan Chase, Citi) insurance (die Bayerische, Optum Healthcare and Allianz), retail (eBay, Walmart), even aircraft manufacturing (Airbus and Boeing). The company now claims that it’s now racked up more than 400 customers, including Adobe, Volvo, and Orange, Eifrem told us on our call.

That makes it, it claims, the world’s most widely deployed graph database, enabling connected data applications for more than 75% of the Fortune 100, and that it also gets 80% of its revenues now from the Global 2000:

We probably talked a lot about enterprise adoption last time we spoke; that was very top of mind, because we were right at that inflection point where we'd grown up from a startup with a fairly even distribution in our customer base between mid-market, mid-size companies and big enterprises. 

What is very clear that the enterprise segment was really taking off for us, which we define as a billion dollars and above in revenue. In 2019, we added over a 100 new customers and those are customers like Caterpillar, Comcast, and UBS: these are the big global brands, from financial services, or telecom or healthcare, life science, big pharma, that kind of stuff. And that train has just accelerated since then.

Next up: analysts have finally started to give him air-time. The company was featured in Gartner’s Magic Quadrant for Operational Database Management Systems, got a namecheck in the same company’s Gartner’s Critical Capabilities for Operational Database Management Systems report and was included in rival analyst giant Forrester’s  The Forrester Wave™Big Data NoSQL, Q1 2019.

It's taken them a while but they say things like, ‘Graph processing, graph databases will grow at a 100% annually through 2022.’ On the record, right? Which is pretty outstanding.

In parallel with growing acceptance at the heavy end of the IT market, Neo4j also points to what it sees as milestone accomplishments in terms of making it easier for developers of all sorts to work with graph. For Eifrem, the central pillar of this is the establishment of a solid standard to achieve interoperability between graph products in the shape of GQL, Graph Query Language. In the Neo4j world, this was an effort it championed that saw the same international committee that developed the SQL standard voting to baptise the first ISO/IEC international standard database languages project since SQL itself.

As ratified last September, this does really seem to be a move toward a true international standard for graph databases and could, arguably, spark as rapid a rise in acceptance of graph-based ways of working as SQL did back in the 1980s.

There's been emerging new types of databases for quite some time; in the mid-90s there were the object databases and the early 2000s, there were XML databases, in the early 2010s document databases. All of them have gone to the SQL committee and said, ‘Hey, let's standardise a query language for X,’ whatever the category may be. 

Every single time, what the SQL committee have done is that they've said, ‘Well, I know that you guys look at this as a distinct category but actually we see that we can just integrate it into SQL.’ For the first time in the history of databases, the SQL committee looked at GQL and said, ‘This is its own distinct category of data organising.’

… but a lot more CIO choice too

But if you ask why this is significant, we also get an insight into the reality of the world Neo4j operates in; everyone else has got interested in graph, too, and he’s no longer the only source of this distinctive way of working with data. As it stands, we have smaller companies like Neo4j in there, but also players like SAP, IBM has an offering here now, Amazon has its Neptune, a list that’s only likely to grow.

To his credit, Eifrem knows this:

GQL means that if you are a CIO then you can write your applications using GQL and while it’s never 100% interoperable between vendor products you’ll probably get to 95-plus. That's a huge deal if you're a CIO in terms of skills they can access and investment in training your developers on GQL, knowing fully well that it can work on Neo4J and on other graph interfaces too from Oracle, from Microsoft, from Amazon over time.

Also to his credit, though it may be dismissed as false modesty, is that he’s also sanguine enough to realise that Neo4j will likely remain one of many players in a market, and that the time for bold dreams of becoming ‘the Oracle’ of a whole new database world may have passed.

But that may not be fatal to his long-term chances, he argues, as instead Neo4j will win hearts and minds as being a high-profile graph player with its own distinct identity:

In November, we launched the fast version of our product called Neo4J Aura and announced a major partnership with Google. Whereas other cloud platforms chose to build their own or acquire a smaller graph database startup Google said, ‘Hey, we want to partner with the best of breed’; they want to be the open source partner-friendly cloud, so they identified six leading top tier open source vendors and we were one of them.

They offered us this completely unique, never been done before in the cloud world partnership opportunity, where we are deeply integrated into their product. You have what they call a console which is where you add services; you log on to your account and you can spin up the feeds, you can shutdown machines and you can also add services.

Side-by-side with the Google services is Neo4J, so it looks and feels like a Google service. It is like, I don't know, if you were on the first screen of the iPhone.

My Take

No-one now pretends that rdbms will wither away and that NoSQL will become the dominant paradigm. If that logic were to prevail, Neo4j would be in pole position to become a genuine global enterprise database, as it would be the only game in town. But what’s really smart about what Eifrem and his team are doing is to accept that but to position their software suite as the ideal alternative to the other work the CIO wants doing— all that complex data stuff that it seems relational hasn’t quite been able to extend itself to be able to handle.

This is a great place to be in, potentially, as Neo4j, which let’s remember is only ten years old (1.0 was released in March 2010, and written basically by Eifrem and a couple of his buddies) can argue it’s got the independence and focus that a large database company saying it speaks ‘graph’ just can’t… and if GQL really does work, that’s a huge safety factor for CIOs opting for the Neo4j approach as opposed to bigger rivals.

I think Neo4j will continue to win interesting business and develop in new ways. Though its current claim that it’s the ideal companion to AI projects as it can best explicate why decisions were reached does need some proof points, it also seems to have found a great niche as the tool you choose when you have a big complex data issue that would profit from a better understanding of the inter-relationships in the information.

And, to be honest, until relational can do that, too, we’ll always need something like Neo4j.