[We published this as a subscriber-online note in mid-October.]
Most business applications run on relational databases (RDB) from companies like Oracle. There’s a view that RDB is an “old” technology which is true – it came about in 1970 so it’s not new. However, there have been decades of investment in RDB as a platform which makes it unrivaled when it comes to consistency, integration, performance, transaction processing, and reporting. RDB systems are good business and generate 40% operating margins. Companies like Oracle can afford to build and buy new technologies (like MySQL) and dominate the market.
There are some fundamental problems for which an RDB system is a bad fit. This became very clear as early as 1990 when I was still at IBM working on advanced technologies. At that time we were working an on internal projects like “IDEA – A Semantic Network Database” which was built to address new application types even before the commercialization of the internet began.
The internet changed everything and companies like Google and Amazon took these ideas to the next level. In the mid-2000’s a generation of non-RDB technologies tagged “noSQL” came into being. These systems provided more “natural” representations of information and much greater flexibility for developers.
We already have seen public companies spring up on the Apache side – Cloudera (NASDAQ: CLDR) and Hortonworks (NASDAQ: HDP). Combined these two have a market value of $3B.
The Hadoop product family is about “Big Data” and is used to process massive amounts, typically on “commodity” hardware. Hadoop also handles diverse data including semi-structured (log files, web pages) and unstructured (video, audio), making it easy to “mix and match” like transactions, click streams, social sentiment and geo-location data.
A big problem with Hadoop is that it’s a bear to use. Systems are hard to set up and manage. Queries are complex and often require special optimization to run. Companies wanting to leverage Hadoop found themselves with a massive skills/staffing gap. In the RDB departments, we had “data architects and data admins” but the Hadoop world calls for teams of “data scientists” which are in short supply and very expensive.
Cloudera has focused on this market opportunity by filling in some of the technology gaps and making the Hadoop platform easier to adopt for enterprises. For the full Cloudera story you can refer to our write-up: Cloudera Wants to be the King of Modern Enterprise Data.
MongoDB is the easiest, most flexible type of database out there. The most striking feature is flexibility. All database systems rely on a structure or “schema” as it’s called. This has two major drawbacks - 1) if you don’t know exactly what data you will want to have and 2) if it is likely to change or vary a great deal. These databases are simple, scalable and allow for fast iteration.
MongoDB, The Market & Competition
MongoDB sits in a niche of the database market. However, that market is $45B and growing. With $124M in revenues (LTM) the company can grow substantially in their niche. The average customer spends on the order of $40K on MongoDB. The company points out that 300 customers have reached $100K ARR. That compares to tens of millions those same customers spend on RDB technologies. Existing customers alone could spend 10x on MongoDB and it would still be a niche.
The direct competition for MongoDB is very limited and takes the form of “projects” like CouchDB. Since document database systems represent less than 5% of the total market it’s hard to justify many commercial systems. According to CrunchBase MongoDB has raised a total of $453M in capital – that’s quite a bit invested and probably enough to ensure that MongoDB is the “commercial leader” in document database technologies much as the $1B invested in Cloudera was enough to cement them as the commercial leader when it comes to Hadoop.
More concerning for MongoDB is the “cloud suite” approach offered by companies like Amazon and Google. Amazon’s AWS offers hosted MongoDB service. However, they also offer their own “DynamoDB” and other third-party options. There is a notion that “MongoDB is great for prototyping but when you deploy you want to use other technology.” Although not as simple and flexible as MongoDB the fact that Amazon has options is an important factor in considering just how much “headroom” MongoDB has for growth.
To combat the AWS approach MongoDB has launched their own hosted service called “Atlas” and they continue to support traditional “on-premise” solutions which neither Amazon nor Google offer. This is still a key enterprise requirement and is likely to remain so.
One key difference with respect to MongoDB is that it is open source. Even though companies pay MongoDB for the commercial version of the software they can take comfort in knowing they have full access to the underlying technology. Anyone can download the software and do everything on their own if they want to.
As we cover in the business discuss the additional challenge for MongoDB is that they must spend substantial money on sales and marketing. The company is making progress on that front – thanks in some small part to Atlas. For the six months ended June 2017, the operating loss was $45M on $67M of revenue versus the same loss on $45M in revenue for the first half of 2016.
Business Model, Valuation & Conclusion
MongoDB is following the “classic” enterprise SaaS model with an open source twist. It’s also called the “land and expand” marketing and sales approach.
The result is a classic subscription model where the 90% of annual revenues are recurring and contract values expand as customers grow and develop more applications with MongoDB.
The cohort analysis suggests that the upfront investments pay off but there’s a significant lag and it’s not known how these other cohorts will perform over time. We’re using the “typical” target model to form a valuation argument for MDB.
While our QuickIV argues for $50/share we note that the market so far hasn’t given these valuations. We revised our IV for CLDR based on recent results and it suggests a $32 price versus the current $15.
Our same model for MDB yields an IV of $35. Based on the current market this might end up being $18-20/share. The proposed pricing of $20/share seems a bit high given the risk/reward.
This IPO should have been filed at more of a discount to prevailing valuations. CLDR and HDP are trading at 6-7x sales and are more established. MDP at the mid-point is already 7x sales.