X
Business

Neo4j looks beyond the graph database

Neo4j is building a 'graph platform' based on its database. CEO Emil Eifrem explains more.
Written by Colin Barker, Contributor

Video: What's new in the graph database world? Here's a quick recap

Graph database company Neo4j wants to move beyond providing only its graph database, and is working on what it calls a 'graph platform' to help companies make the most of their data.

ZDNet recently caught up Neo4j CEO Emil Eifrem to find out more.

ZDNet: Could you talk me through the new native graph platform?

Emil Eifrem: We have been at this for a while. We coined the term "graph database" when dinosaurs ruled the earth. We defined it. We've given it meaning. We've taught the world about its core benefits, about its top use cases, all those good things.

We are rapidly becoming this kind of default backend for a wide variety of horizontal, and very valuable use cases.

neo4j-emilblue.jpg

Neo4j CEO Emil Eifrem: "Fifty-one percent of enterprises that have graph databases in production... That's pretty crazy -- this is a term that we invented less than 10 years ago."

Image: Neo4j

Now we talk to customers all the time and we saw a pattern emerge where, very frequently, our customers were building out a very similar architecture where our technology was usually deployed in isolation. This came as a great surprise to us.

We live in an ecosystem of Hadoop and Spark, relational databases and data warehouses. People stick other technologies on top of us, for example, visualisation tools and things like that. But I visited three customers one day and two customers today, and four out of those five all had a very similar architecture.

Then I thought, what if we could, out of the box, offer something that is more than a database, which packages up some of these technologies, including data integration and things like visualisation toolkit and so one. And then we could package that up into what we call the Graph Platform which will be a one-stop shop for anything you want to do with connected data.

That's what we are now building. What we are announcing is a vision. This is where we are going. We want to help our customers by giving them a context, a framework of where we are going.

It goes back to what I said 10 years ago: we are building the equivalent of an Oracle database for connected data. What we've built is an awesome database but one that rather than work with tables, works with connected data.

So now we are saying that the heart of that graph platform is the database, but it's now much more. It's a suite of all the technologies that you will need in order to leverage that connected data. We call that the Graph Platform.

You have listed a bunch of things in your new announcement. Are all of these up and running?

Some of those technologies are available right now but this was more about vision. The vision is that anywhere that you have data, you should be able to get it in and out of the graph platform.

What we announced last week was integration with two relational databases, Hadoop and Spark. We also announced the graph analytics suite of algorithms -- we currently have three and ultimately want to get to eight.

Which components are built-in so far?

We have Data Integration and we have another component called Neo4j Extract Transport Load (ETL) and it connects into, for sure, Postgres and I believe MySQL as well.

And so if you have connected data in that database then it will automatically put it into Neo4j. And it has a nice little user interface so that when you point it at a database, it will scroll up and show you a little picture of what the graph database will actually look like.

If you like it, you can click 'import' and if you don't like it, you can just put in some changes and it will automatically make those changes and you will see that immediately. That's one thing that we are releasing.

Another is Neo4j Analytics and that has a range of new features. For example, a thing we call path-finding which will autonomously get you from point A to point B. And then things like clustering which tella you, 'Hey! There's a lot of transactions clustering around this particular node, perhaps that's a fraudulent person'. Things like that.

The third kind of algorithms are things called 'centrality', which is kind of a weird term but is basically a method for identifying the key influencers within a network. For example, let's say that you're a telco and you want to send out some marketing campaigns. You want to know who are the key influencers within your network. That's where centrality comes in.

The third one is a tool that we call Neo4j Desktop and that's really the most tangible part of the platform. You can launch this desktop application and then inside of that you can have all these various graph applications, such as the ones that we ship but you can also build your own.

Can you talk about any specific customers who are using this?

In general terms I can tell you that seven out of ten of the biggest retailers on the planet are now using Neo4j. One we are public on is Walmart.

Another one that I can go public on -- and I think it's a really cool use case -- is eBay. They are one of those customers who have a lot of projects with Neo4j all over the place. Now they have a number of cool things on the go but one that I think is one of the coolest is an application called ShopBot.

Now they are very excited about this -- the CEO talked about it in earnings calls --- and it's what eBay is betting on as the future of commerce. It's called conversational commerce and, basically, it starts off as a Facebook Messenger bot and you connect to it on Facebook and you chat with it just as if you were chatting with a friend.

You ask it, 'Hey, I'm looking for a bag' and then it asks what type of bag and you start chatting from there.

We are a part of that architecture and the way that works is if someone types back and says that they are looking for a bag then usually it is either a woman's handbag or a man's backpack. Now the conversation moves on and all the information is encapsulated in the knowledge graph and that is one of the pillars of machine learning.

I think that that is going to be one of the most popular of our use cases over the next several years.

Machine learning and AI is super-hot and knowledge graphs have emerged as one of the coolest areas. Now Google understands that AI is ripe for machine learning and one of the things that they have talked about is that they use Knowledge Graphs increasingly for AI.

Your business is getting pretty large now, isn't it?

It's always smaller than I want it to be. We are growing pretty fast. This year we have grown from 200 to customers to over 250 and mostly they are global 2,000 companies. I have already mentioned that we have seven of the ten largest retailers and we have over half of the top 25 financial services companies and eight out ten of the top software vendors are using Neo4j today.

In 2014 Forrester said, "We predict that 25 percent of all enterprises will have graph databases in production by 2017", and everyone thought that they were nuts. I thought, Wow, that's really aggressive.

Now, here in 2017, three weeks ago they released their updated report where they surveyed over 2,200 enterprises, and their conclusion is that rather than 25 percent, it's 51 percent of enterprises that have graph databases in production.

Now I thought that that was pretty crazy -- this is a term that we invented less than 10 years ago.

Do you think that among the larger companies, there's a fair amount of experimentation going on?

Yes. The Forrester report looked at how many are using the technology in production and how many were just trying it out. Something like 20 percent are planning to use it in production over the next 12 months. Add that to the 51 percent and you can really see that it is going deep inside the enterprise.

When you look at the graph database versus relational, at one point do you think you are in terms of market acceptance?

I think we are roughly where relational database was in the second half of the '80s. They were invented in the '70s - Oracle was founded in '79 - and then there were two or three competitors and it just took off.

Then, towards the end of the '80s if you took the top 4 relational vendors they were about a billion dollars' worth of revenue. I think we are in the second half of that '80s. Its hitting mainstream, the SQL language was slowly emerging - it hadn't taken off, but you could start to see the potential.

That's where I think we are.

Previous and related coverage

Video: What's new in the graph database world?

Here's a quick recap

Sparkier, faster, more: Graph databases, and Neo4j, are moving on

New players, new features, platforms, and ticking boxes. Let's talk graph with Neo4j.

Infographic: Most companies are collecting data, but aren't using big data solutions [Tech Pro Research]

In a recent Tech Pro Research survey, almost all respondents said their company collected data to glean business insights, but only 39% said their company has invested in big data platforms.

Read more on graph databases

Editorial standards