Neo4j CEO Emil Eifrem
Creating a viable challenger to database giants Oracle and IBM is one of the hardest business problems I have seen. After 18 years, Emil Eifrem looks poised to succeed.
His company, San Mateo, Calif.-based graph database platform supplier, Neo4j, recently raised $325 million at a valuation topping $2 billion to propel it to a possible public listing.
How so? Since making its product available in 2007, Eifrem told me in a June 22 interview that his company has solved two hard problems: “R&D and Go-to-Market.”
If you own shares of Oracle or MongoDB, you ought to keep your eye on what they are doing to keep up with Neo4j.
Neo4j’s Record Funding
On June 17, Neo4j announced a $325 million Series F funding — the largest in private database company history — bringing its valuation to over $2 billion and total funding to over $500 billion. The round was led by Eurazeo with additional investment from GV [Google Ventures].
Neo4j’s fund raising is impressive compared to that of rivals. VentureBeat notes that MongoDB, which went public in October 2017 — and has since soared at a 91% average annual rate to a $22.5 billion market cap — raised $311 million in total. Meanwhile, “Cockroach Labs raised $355 million in funding, including a $160 million round earlier in 2021 at a similar $2 billion valuation,” noted VentureBeat.
A graph database is different from relational databases built from tables — they store words and numbers in rows and columns — that have been widely adopted and used for decades. Data has changed since then and the graph makes it easier to store and retrieve today’s more prevalent unstructured data.
The graph database stores “individual data points (such as key-value pairs or documents) as well the data relationships between them. Those relationships make a graph database easily a million times faster than a relational database” for applications such as fraud detection, according to TechRepublic.
Targeting A Large, Fast-Growing Market
Neo4j is targeting a huge market with gigantic entry barriers. According to TechRepublic, in 2020 companies spent $50 billion on databases — a figure that’s expected to grow at a 15% average annual rate to $100 billion by 2025.
Neo4j’s type of database — graphs — represent a smaller market that’s growing much faster. VentureBeat noted that the graph database market will increase from $822 million at a 24% compound annual rate to $2.4 billion by 2023.
Gartner expects that “enterprise graph processing and graph databases will grow 100% annually through 2022, facilitating decision-making in 30% of organizations by 2023.”
How Neo4j Has Grown
Since Neo4j started selling its product in 2007, it has grown considerably. As Eifrem told me, the company employed 300 in March 2020 and expected to end 2021 with 600. (Sadly, he did not tell me its revenues or growth rate).
VentureBeat notes that Neo4j had received “more than 3 million downloads as of November 2018 and [has] over 300 enterprise subscription users. Current and previous customers [include] Lyft, Walmart, eBay, Adobe, Orange, Monsanto, IBM, Microsoft, Cisco, Medium, Airbnb, NASA, and the U.S. Army.”
Neo4j has solved two very difficult problems to reach its current level of success. As he told me, “Making our product successful meant solving a hard R&D problem and a hard Go-To-Market problem. Databases are at the heart of applications. SQL and Relational Database Management Systems (RDMS) — storing data as rows and columns — have predominated in the 1970s, 80s, 90s, and 00s. When I grew up, Oracle, IBM DB2, Informix and Sybase were 98% the same product.”
Eifrem knew it would be hard to replace RDMS with something better. “There was the opportunity to put a new database in there. But it would face huge scrutiny and deep friction in adoption. Even if you could crack the technical problems, the Go-To-Market would be expensive and very hard to pull off. If successful, it would be massive and would last for decades. The only thing close to solving this challenge is creating a new operating or Enterprise Resource Planning (ERP) system. It’s a huge market. If you can succeed, the prize is fantastic.”
To his credit, Eifrem’s idea for the company was spurred by a problem he faced that then-extant database companies were not solving. Most ventures I studied to research my book, Hungry Start-Up Strategy, were started by people trying to find a solution for a problem that they faced personally. They tried to find a company that was already solving the problem. They started their company because the problem was too important not to solve.
Neo4j followed that pattern. “I came into the idea for a new database because I ran into the problem myself. I was using SQL as a chief technology officer in my early 20s during the first IT bubble. We had the standard stack: Layer 1 was the database, Layer 2 was the business logic, Layer 3 was presentation. For some reason the SQL database was working against us. We had 20 people in the department and 10 of them were fighting the database. In all our previous projects, the database was an accelerant. It solved a lot of hard computer science problems such as storing on a disk, query and backup,” explained Eifrem.
He came up with a critical insight about SQL. “I tried to figure out why the database was working against me. It was that the shape of the data and the building blocks of SQL were in conflict. The table structure was rigid. It made sense when it was first developed — people were putting forms into a computer. But we were experiencing an explosion of data which was messy and inter-connected. I thought ‘What if we had databases that were nodes connected through relationships — a social graph?’”
Benefiting from the ignorance of never having tried it, he embarked with confidence on solving this problem. As he told me, “I was young and stupid enough to think we could build it. We started googling around and couldn’t find anything. We decided, ‘Let’s build it. How hard can it be?’ After 15 years and $100 million we know it is hard. But we concluded that it was worth it because we had a tailwind: the world would become increasingly connected; data would become more connected; and more projects would need it in the future. We were customers. We were sure others would be too.”
Neo4j saw that the most successful users of databases had the same problem. “To get started we needed to build and vet it in the first company. The leap was trying to build it is an independent entity. As we surveyed the landscape of what people were doing with data in 2007, Google, Facebook, Yahoo, eBay, they would use an RDMS. They were forced to build their own [graph database]. We realized that the enterprise would face this problem tomorrow. They were all looking at how web pages are connected. What we built would help Global 2000 companies face their competitors.”
Eifrem looked to develop its product in partnership with early adopters. “We got our first customer by finding early adopters in different industries. They are personalities that want to be on the cutting edge. They see potential and exist in all industries. You look for the level of pain they are solving: Is it a vitamin (e.g., it’s good for you), a drug (like tylenol) or penicillin (without out it, you die). We looked for use cases where it was mission critical for early adopter personalities,” he said.
He settled on working with UBS to develop an easier way to onboard traders. “An early example is entitlements in banking. When you onboard a new trader, which assets do you give them access to. The answer depends on where in the world you are hiring and their previous network. You provision it all out by the Web. In 2010, UBS used Neo4j for the backend of its entitlement systems. We were a million times faster at connecting data than our competitors — the LeBron James [of entitlements processing].”
This industry-leading skill enabled Neo4j to get its foot in the door and improve in other areas where it was relatively weak. “Everything else, we were inferior. It was like the Google Doc which enabled people to share documents much more easily — which saved people time. We had to improve in all the other boring stuff like online backups and compliance. Meanwhile, there was a shift in the platform: there was a move from tapes to disk to the amount of random access memory (RAM) available for storage which changed how you can access data,” he explained.
While Neo4j’s magic power — much faster data connections than rivals — remained the same for different customer applications, the business impact was different but profound. “The business impact for the CFO in entitlements was that we can cut the time to onboard new traders from weeks to seconds. This enables banks to give new hires the royal treatment.”
Another use case was fraud detection. As Eifrem said, “We could see how fraud rings were connected and we enabled Morgan Stanley’s CFO to identify 3% to 4% more fraud — generating a big return on investment. With retailing, our technology enabled recommendation engines to work better — creating a 5% uptick in sales. A retailing CFO with $2 billion worth of e-commerce revenue can do the ROI calculation on a 5% sales increase from our technology.”
Neo4j aims to turn its first customer in an industry into an enthusiastic promoter and thus gain a broader foothold among its peers. “We do marketing within industries. From our first early adopter buying an antibiotic, we add others in the industry. We went from UBS to Goldman Sachs, Royal Bank of Canada, and Royal Bank of Scotland. We also create new categories to which we invite competitors to conferences to help create. We also have practitioner led adoption. We marketed an open source version to developers that is completely free.”
More opportunity is ahead for Neo4j. According to Eifrem, “We currently have a thin slice of the Global 2000 in North America and Europe which represents 80% of Oracle’s database revenue. We are making inroads in APAC and grew there during the pandemic. We have a cloud offering for developers at $65 per month for the individual/SMB market.”
I think Eifrem has a rare combination of technical and business skills which contribute to Neo4j’s success. “It is a blessing and a curse that I am interested in alot of things. Technology is hard. I spent 10,000 hours practicing — taking an athletic training approach coding in the basement instead of cross-country skiing. The business stuff is much easier to learn if you are interested and curious. Managerial skill comes from people skills — it’s just experience,” he told me.
How Incumbents Are Responding
Neo4j is not the only company going after this opportunity. Startups like TigerGraph, MongoDB, Cambridge Semantics, DataStax, and others compete with Neo4j in the graph database market as do incumbents like Microsoft and Oracle. in November 2017, Amazon launched Neptune, a fully managed graph database powered by its Amazon Web Services division, according to TechRepublic.
What’s more, general-purpose databases Postgres and MongoDB are adding graph (and other) functionality. I would not be surprised if some of these rivals can convince their customers that the cost of switching to Neo4j is greater than the benefits.
Will Neo4j Go Public?
I think Neo4j will go public. However, it’s difficult to assess when without knowing its revenues and growth rate.
The company will use the proceeds to improve its product and to strengthen partnerships and expand geographically. Building on a deep relationship with Google Cloud, Eifrem wants to enhance its partnerships with public cloud providers, according to TechCrunch. The company has also opened up an office in Singapore to tap “a major uptick in interest from the APAC region.”
Ultimately, its ability to grow depends on whether it can keep finding new groups of customers and use cases where Neo4j can persuade customers that the benefits of buying its graph database exceed the costs of switching from a rival’s.
