Elastic (ESTC): SaaS Transition Kills Open Source Debate, but Too Early

Elastic N.V. (“ESTC”) provides elasticsearch, an extremely popular open source search engine that’s used for search, infrastructure monitoring (including logging), and security. An excellent background is provided here, written by “CMF_Muji”. He did a great job covering the basics of the company, product, and use cases. I will not repeat that effort here.

For the purpose of this article, I will refer to Elastic the company as “ESTC”, as distinct from elasticsearch the product.

Elasticsearch the product is well established. No one questions the product is great, or that it’s a de facto standard (at least in areas like logging); or that it has a long growth runway ahead of it. So let’s get to the point.

Amazon’s Attacks on ESTC

The biggest problem for ESTC investors are threats to its open source business model, as exemplified by Amazon’s recent moves. In this post I will explain the nature of Amazon’s attacks, and why I think this highlights ESTC’s vulnerability to other and future threats.

Amazon's attack has two fronts: 1) Amazon's hosted service (“Amazon Elasticsearch Service”) competes with Elastic's hosted service (“Elasticsearch Service”). This actually has not been a huge problem. Despite the obvious potential for naming confusion, ESTC’s own service is growing nicely. 2) Amazon’s “Open Distro for Elasticsearch”. This stems from Amazon’s complaint that ESTC mixes open source and proprietary code in its default distribution.

Of the two, Amazon’s Open Distro is by far the bigger problem. This article gets to the crux of Amazon’s criticisms.

The issue is a bit of “he said, she said”. Note that ESTC actually let you download an all open source version with no proprietary features. So arguably Amazon fails its attempt to claim moral high ground. Granted, ESTC's OSS version is compiled version and not the source code, and that a developer who want to work with the open source code from Github will have a hard time avoiding the proprietary code. Still, if you’re a developer who just wants to use the ELK stack, you’re not going to fiddle around with the source code in Github, you’ll just download the binary anyways. So I’m not sure this whole debate matters for the average user.

ESTC also says it’s all very clear what’s what, according to the article.

“All of the code for our proprietary features are kept in a separate top-level folder called “x-pack,” to avoid any mixing or confusion. We also include a header on every source file, indicating whether it is licensed under Apache 2.0 or the Elastic License, to prevent any ambiguity”

Non-existent Barrier to Entry

If this debate is purely about open source versus proprietary, then there’s an easy way for ESTC to put it all to rest. They can simply provide a version that only has free open source – the source code, not the binary.

But that’s not the only issue. Amazons Open Distro actually has more functionalities than ESTC’s "all open source" version. It is closer to ESTC’s premium distributions, but all for free.

With the first release, our goal is to address many critical features missing from open source Elasticsearch, such as security, event monitoring and alerting, and SQL support,” Cockcroft wrote.

Will these extra features be enough to sway developers toward Amazon’s version? Who knows? Does it matter?

The point is, this could just be the start. Anytime ESTC comes up with some premium proprietary feature, AMZN can fork it and come up with a free and open source version. Perhaps they’ll even add features that are better than those in ESTC’s premium paid version. It would then be up to ESTC to match those features in price (as in zero, free), and release them under open source Apache 2.0 license. Will they have to do that? I don’t know. The potential for an ugly spiral to the bottom is certainly there.

The underlying problem is ESTC’s “Open Core” model. There is next to no barrier to entry. Source codes for the core functionalities of elasticsearch are on Github for all to see and use. Forget Amazon, an enterprising startup can take those codes, add their own features (open or proprietary) and created their own version. “open source” is more like open season for wannabe competitors.

Adding to that pressure, selling software is not Amazon’s core business, nor Netflix’s, nor Expedia’s. Elastic charges for their service by “nodes” and it gets expensive real quick. Big corporate customers won’t hesitate to hurt ESTC’s business if that gives them an edge in negotiating and pushing down those licensing costs. For ESTC this could be an ugly race to the bottom – everything for free.

Transition to SaaS Makes this Debate Moot

Now let’s go back to CMF_Muji’s article. One of his best insights is that ESTC’s acquisitions all have something in common, which is these are applications built on top of elasticsearch itsef. ESTC can then easily integrate them into its own platform and offer SaaS products.

Note I’m not talking about the so called “SaaS” that shows up as revenue in ESTC’s income statement. ESTC counts Elasticsearch Service as “SaaS”. But that’s really just hosted service, still very much a “build it yourself” approach. It’s more akin to IT infrastructure/platform (see notes for various product lines). What I mean by “SaaS” here is end users getting the application functionalities without having to interact with the underlying infrastructure stack. Think of SaaS like SalesForce, like Workday.

That’s not ESTC, at least not yet. For now they are primarily a vendor of on-premise, self-managed software, with offers to host that on the cloud (but still have customers build their own apps on top). And even that hosting, “Elasticsearch Service”, is still under 20% of revenue.

ESTC is moving in the SaaS direction though. With a series of acquisitions, ESTC is well positioned to shift away from selling infrastructure software/source code and toward being an enterprise application vendor. App Search Service (powered by Swiftype which they acquired), Site Search Service, and the new Enterprise Search Service – these are all closer to what I would consider SaaS products. ESTC also just acquired Endpoint, and will use it to security application space.

CMF_muji correctly points out that this “true” SaaS approach is an upside option for ESTC. I think it’s more than that. It’s a way out of this whole “existential threat to business model” quandary.

This is a step function change in business model – selling application instead of selling enabling infrastructure. Your value proposition is different. The customer’s mindset is different. The SaaS user will focus on fulfilling the business case and ignore the infrastructure. They will not play around with Github and source codes.

At that point the whole debate of open source versus propriety becomes moot.

That is a promising thought, but one that is a long way off. The value proposition, customer set, as well as ESTC’s sales and marketing approach all have to change. This is by no means an easy move.

What I'm doing with the Stock

ESTC is a company with optionality, but also one whose core business model is threatened and will likely have to undergo a big transition (no easy feat). It is also trading at some 13x revenue. 

For that, ESTC goes into the too hard pile.

When would I revisit? I would when actual SaaS revenue (ex Elasticsearch Service) shows a clear pathway toward being half of total revenue. I will also rethink this if stock shows technical support in low $60’s.


Notes: The product set

ESTC’s offerings consist of 1) on premise and self-managed, under which there are free and premium versions, and 2) cloud offerings, which include ESTC managing the infrastructure for you, as well as what I consider “true SaaS” application offerings.

• On premise/self-managed products
• https://www.elastic.co/subscriptions
• Free versions. The default distribution ("Basic") includes open source as well as proprietary features (but still free).
• Paid versions. The Gold and Platinum versions include support as well as advanced features.

• Elastic Cloud. https://www.elastic.co/cloud/
• Elasticsearch Service - this is their hosted service (can choose AWS, GCP)
• Elastic App Search Service
• Elastic Site Search Service
• These are cloud software, but not all are SaaS. ESTC counts all 3 to be SaaS, but I would only count the latter two (App Search and Site Search) as real “SaaS” in the sense of how market uses the word. Elasticsearch Service (for now the bulk of revenue in this segment) is still "build it yourself" software.

• Elastic Cloud Enterprise
• https://www.elastic.co/products/ece
• This is confusing naming because this is very much an on-premise solution. Sure, it can be deployed on public or private clouds, but if someone wanted to deploy on public cloud, they would have just used ESTC or Amazon’s hosted service instead of installing this.
• here's the description: Elastic Cloud Enterprise, or ECE, is the same product that powers the Elastic Cloud hosted offering, available for installation on the hardware and in the environment you choose. ECE can be deployed anywhere - on public or private clouds, virtual machines, or even on bare metal hardware

Framing the Questions for GOOGL

Wall Street Journal has a good article that spells out the various products underlying Google's search advertising platform. It is an excellent five minute primer. Here's the link.

The search advertising business is not one product, but various components that interact to make a market place. The article shows that Google owns and dominates all sides of that market:

• Sell side (for publishers to publish) 
• Publishers use "ad servers". These ad servers has information on what spaces on media properties are selling ad placement. The ad servers provide that supply info to the market place.
• Large publishers typically use its server, commonly known as DoubleClick for Publishers. (DoubleClick for Publishers and AdX are now tied together in one product called Google Ad Manager.)
• Smaller publishers often use Google’s AdSense.
• Mobile app publishers use Google’s AdMob.

• Buy side (provide ad purchasing tools)
• Google Ads (formerly known as “AdWords”) for buyers to bid for search ads placement on Google (its own property)
• DV360 for (for Display and Video) for buyers to bid for video ad placement. This does placement even outside of Google properties.

• Exchanges 
• Googles owns AdX which is the largest exchange with about 50% market share

• Media Properties
• Google search
• Youtube
• Others

• Analytics

There's plenty of firepower for regulators and antitrust guys here! 

For one, how can you have a market where all participants are owned by the same party? There are numerous other issues. Just as an example, the bundling is problematic. Google Ad Manager is a bundle of 1) DoubleClick for Publishers and 2) AdX. Combining a sell side tool with an exchange is not good optics, since the exchange can favor its own publishers.

Another example is analytics. This is an issue because it acts as the "referee" of the effectiveness of other Google search products. I can easily see regulators demand Google separating this out.

In short, plenty of ways regulators can cut them up.


The question then, is what pricing power or otherwise benefits did Google derive from owning all of this?  If this is all broken up, would GOOGL suffer economically? if so, how much? 

I suspect these questions are unanswerable. Google (and its parents Alphabet) is just not very transparent, particularly for a megacap. I cannot tell how much money they make from Ad exchange versus their own property selling ads, versus the various buyside/sell side tools (including ad servers?).  

If anti-trust is going to be the big overhang over GOOGL stock, then not having the transparency to answer these questions is not going to help. 

For now Google does not even break out Youtube. So they have a long way to go. 

Radcom: Creation of New Market Plays to Its Advantages

Executive Summary

• Radcom failed to “cross the chasm” in 2016-2018, but they have worked to further rounded out their product portfolio.
• Rakuten’s greenfield project potentially creates a whole new set of customers and represents a TAM expansion for all NFV vendors.
• This new customer segment is small and new enough that Radcom’s small size is not a disadvantage. It is also a niche where customer requirement plays to Radcom’s strength and negates incumbent’s advantages.

Background on Radcom (RDCM)

Radcom (RDCM) provides software based service assurance for telecom networks. This is software that measures how networks perform from customers’ perspective in real time so network operators can react accordingly.

The stock had a great run up during 2016-2017 after they revealed AT&T as a major client. Back then AT&T was starting out its network function virtualization (NFV) initiative, replacing dedicated hardware with software, with RDCM’s product being one of the winners. But the AT&T project has not translated to sustainable revenue growth, and the disappointment crashed the stock.

At its peak around mid-2018, Radcom was worth almost $250mm in enterprise value. When I bought it recently, (around $8.6/share), it was only worth around $60mm of enterprise value.

So what happened? In short, RDCM failed to “cross the chasm” (in Geoffery Moore’s term, see footnote). Instead of advancing NFV into the mass market, the AT&T project hogged up all of Radcom’s resources and paid the company just enough to break even and sustain itself.

Since then the stock stalled, but Radcom kept moving. Despite the setback, they have continued to develop their product expertise and have rounded out their capabilities. Back in 2017 all they ever talk about is vProbe service assurance. Now a look at company website shows that they also have a network visibility / network packet broker service, as well as a higher level network insights product.

Background on Rakuten

In May 2019, Radcom announced a deal to integrate its products throughout Rakuten’s greenfield mobile network project.

Rakuten is the Amazon of Japan. Despite having no telecom experience, it is trying to use cloud based technologies to building out a mobile network quickly and cheaply. Below is a simplified diagram of Rakuten’s network design, and this article from Cisco article helps provide greater clarity. In short it’s virtualization and software everywhere.

Note that the radio access network will be virtualized (vRAN), while the entire core network runs off a common cloud datacenter platform (“Telco Cloud”). All the network functions are implemented as software (“Virtualized Network Functions”, or “VNF”) on top of this cloud platform. Indeed the vRAN itself is simply a VNF, as the case Radcom’s solutions.

A fundamental tenant of virtualization is taking functions done by hardware and modularize into commodity hardware and software. The benefits go far beyond cost savings from replacing proprietary devices with commoditized servers. Replacing hardware with software also result in energy efficiencies, and in general allows for less bulky apparatus and thus alleviates space constraints, and by extension location constraints. This again lowers cost via speed of deployment and real estate flexibility.

Virtualization itself is not new, but Rakuten takes it to the next level by virtualizing the radio access network itself – (vRAN). By splitting the baseband unit into software modules, vRAN enables “small cells”, a key 5G requirement. In Rakuten’s words:

This innovative approach enables deployment of very lean cell sites, with only antenna and remote radio heads, which in turn maximizes opportunities for successful acquisition of cell sites.

An Nikkei Asian Review article reported that Rakuten’s cost of building base stations are only 10-20% of competitor’s:

Rakuten has the lowest cost of base station installation among Japan's four top wireless carriers, an analysis of plans submitted to the communications ministry shows. The company's cost totals about 8.2 million yen ($76,000) per unit, only 10% to 20% of the level for industry leader NTT Docomo.

Rakuten’s Creates a New Customer Segment - TAM expansion for all NFV Vendors

Rakuten’s implementation of NFV– using “telco cloud” and vRAN - could be adopted by not just other cloud companies, but eventually telecoms and even cable companies as well. That would amounts to a massive TAM expansion for all NFV vendors, and particularly plays to RDCM’s advantage.

The ability to build out mobile networks cheaply and flexibly should be appealing to all:

• Cloud companies (Google and Amazons of the world). These guys can build their own network where telecoms do not have strong coverage, or just for leverage against telecoms. For example Google/Waymo might like to push out autonomous car in certain cities, and they’re not going to wait around for AT&T to set it up. Another example could be Amazon building out drone based delivery in certain areas. Amazon can build its own vRAN based system to direct those drones and use that as bargaining chip to talk down Verizon charges.

• Cable companies. These guys have to get into wireless data. Their video business is dying from cord cutting and over-the-top video competitors. Once mobile data plans become the norm for home usage, it will be game over for the cable guys. As it turns out, they recognize this threat and are trying to offer mobile data services as well. Satellite companies are facing the same threat and are also getting into mobile. The vRAN is one way to deploy mobile networks cheaply.

• Telecom. The vRAN could be a way to accelerate telecom’s move toward NFV.

So Rakuten’s effort can open up the NFV game to not just telecoms, but cloud companies and cable companies, and that means TAM expansion for all NFV players. 

New Customer Segment Plays to Radcom’s Strengths

The implications are even better for Radcom.

In general, small companies should concentrate their resources on where they have the highest probability of winning. That means aiming for market share dominance in a new and small niche, before leveraging that customer reference and industry position into another segment.

Rakuten deal creates a whole new “cloud based mobile operator” customer segment, and that is an easier segment (compared to massive telecoms) for Radcom to win, if only because it is too small for large competitors to focus on.

The elegance and simplicity of Rakuten’s greenfield, software driven build-out also make it easier for RDCM to leverage what they already have and complete the whole product. There is no legacy infrastructure that RDCM has to cater to. Established competitors cannot neutralize RDCM’s product advantage with some software/legacy hardware bundle. No, this market demands a pure software solution and RDCM is it.

This product-market fit was demonstrated in the way Radcom won this new segment – with minimal work and in record time. Indeed, CEO Yaron Ravkaie says Rakuten was one of the fastest sales cycles he has ever been involved in.

RDCM used to be a small fish in a big pond, now it can be a big fish in a small, but expanding pond. 

Conclusion

RDCM has credibility as a vendor of NFV based network assurance, so a TAM expansion of NFV paradigm beyond telecom increases the probability of success for them.

To be clear, Radcom (and NFV/vRAN for that matter) has a long way to go before hitting mainstream adoption, so buying it here is speculative. But the upside is real enough and they have managed well enough in the recent downturn (low cash burn, strong cash position, no debt, no dilution) that I think they can stick around to harvest that upside. 

The Rakuten deal is already better. It is a subscription deal, as opposed to the project based nature of the old AT&T deals. For all these reasons I think there’s a good chance Radcom is worth a lot more than its enterprise value of $60mm (using $8.6/share) 

So I bought RDCM, and I will be watching Rakuten’s initiatives closely over the next few years.

Notes on "Crossing the Chasm"
According to Geoffrey Moore, early on in the technology adoption cycle, innovative companies are immature and need visionary clients to sponsor it. These visionaries purchase an incomplete product in the hope of developing it and turning it into a dramatic competitive advantage. The market then gets excited about the company winning a big name client, extrapolate rosy success, pumping up the stock price.

However it turns out that the visionary isn’t such a great client – they demand various modifications that cater to the specific client, and hog up all the startups resources. In the end the market fails to go from niche to mass market, thus the company couldn’t leverage the initial client into more clients. The company thus falls into the “chasm” between visionary adoption and main stream adoption.

That is the story of Radcom and AT&T in 2016-2018. The stock chart captures the joy and pain.

The strategy to “crossing the chasm” is to find a suitable, small niche you can dominate, and leverage that market leadership into a related niche, capture that and leverage that into another adjacent. Do this until the collection of related niches turns into a mass market.

As my article indicates, this is what I think can happen with Radcom working with Rakuten.

Checking In on Fidelity National Financial (FNF)

FNF is something I have held for years and written about a few times. I’m doing a quick update here.

The reasons I’m holding remain the same: 1) intermediate term defensiveness, 2) long term upside, and 3) Stewart acquisition synergies. For the next year or so it’s a value/dividend play: number one player in oligopoly; strong cash flow generation at solid valuation; great balance sheet; 3%+ dividend yield. In the longer term FNF has upside from demographic tailwind (millennial reaching prime home buying age). FNF is also in the process of acquiring the industry’s number three player, Stewart, and will benefit from acquisition synergies.

Quarterly earnings will fluctuate and are not my focus. The main things I look for in earning calls are: 1) Stewart acquisition progress, 2) financial health, and 3) any disruptive threats in the horizon. Here are the latest.

1) STC acquisition. Not much update on this front. The deal is still stuck on NY regulators and it may come down to divestitures.

2) Financials. Balance sheet remains very strong. FNF debt outstanding was $837mm (compared to investment portfolio of more than 4.6bn). Debt to total capital ratio was 14%.

3) Disruptive Threats. One of the biggest trends in real estate is the emergence of iBuyers. The risks for FNF are 1) they have relationships with real estate agents who recommend FNF’s services, but iBuyers seek to place themselves at the center of transaction, resulting in channel disruption for FNF. 2) Also iBuyers will have better bargaining leverage compared to regular home owners since they hold large portfolios.

This exchange from 1Q19 transcript shows that FNF may have not grasped the full significance of iBuyer. It’s hard to tell, maybe management is simply being careful about what they say to maintain relationship with real estate agents. That said, it’s still early innings in the iBuyer game and a group with strong tech track record like FNF will likely get on top of it.

Mike Nolan
“…I think what we see at least today is that the real estate agent is still at the center of the transaction. And so while there's disruption potentially in brokerage, I don't know that that's disrupted the agent yet. Now that could change of course. But we're really focused on real estate agents, because that's who gives us the transactions and that's why we've made investments in real estate technology and lead gen and things like that because that's where we're going to continue to focus.

Jason Deleeuw
Got it. And what about the iBuyers and – I mean, how are you relating to them? It seems like it's going to continue to be a growing segment of the market. Is there a change in how you're doing that? Or you're reaching out to them or you're already working with some of them? How is that working?

Mike Nolan
Really both. We're reaching out and working with some of them. They're a customer just like anyone else.

They have transactional volume that they can control. And we'd like to perform that title and closing works. So, we're calling on them. We're working with them. In some cases they might be working with our agents. So by extension we're working with them. But they're really just another type of customer from our perspective.

Overall, title insurance is a sleep industry and the latest quarter was business as usual. I continue to hold a 4%-5% position, and would hold even in the unlikely scenario that Stewart acquisition does not come through.