Understanding Twilio

Summary

• Articles on Twilio usually fail at describing exactly what the company sells and why customers buy it.
• Understanding the nature of the business has significant consequences for understanding its prospects.
• This article both explains the business and draws the obvious conclusion regarding resulting economics.
• This idea was discussed in more depth with members of my private investing community, Idea Generator. Start your free trial today »

Twilio (NYSE:TWLO) has a poorly understood business model. It's thought of as a disrupting CPaaS (Communications Platform as a Service) that's driving a new paradigm through the provision of programmatic communications, available through an API. As a result of this, Twilio is bundled with other IaaS/PaaS providers like Amazon.com's (AMZN) AWS, Google's (GOOG) Cloud Services, Microsoft's (MSFT) Azure, Salesforce.com (CRM), etc. In this article, I'll show that what Twilio actually sells and its customers actually seek is something much more prosaic, dictated by a very specific issue. This understanding of the company has significant negative implications for its prospects and valuation.

One of the things which struck me the most was how neither other articles nor Twilio itself really fully described what the company “did for a living.” There was this underlying CPaaS (Communications Platform as a Service) “is the future” and “is going to disrupt this and that” feeling, and little thought was actually given about what Twilio sold or why customers bought it.

This lack of caring about the actual business also meant that there was little thought given to:

• Competition and competitive dynamics
• Pricing power

The only time where some thought was given to the nature of Twilio’s business was back in early 2017 when it started losing Uber’s (UBER) business. But even that introspection quickly faded from memory.

In this article, I’m going to do my best to highlight just what Twilio actually sells and what that implies for the company's ongoing business, profitability and growth.

Twilio’s Main Business

When it comes to overall chatter, you’ll see statements to the effect that Twilio is disrupting communications. That it is providing a unified API able to handle all kinds of communications, including voice, texting, video, email and whatever else. That it offers a virtualized communications network.

These are all true, but what ultimately matters is what Twilio actually sells that customers really want to buy. Overwhelmingly, the company’s business is about 2 things:

• Programmable messaging (i.e., SMS). Being able to handle outbound and inbound SMS traffic. A lot can then be built on top of this, like reacting to the content, but that's besides the point.
• Programmable voice (that is, phone calls). Being able to handle outbound and inbound phone call traffic. A lot can then be built on top of this, like reacting to content, but that's besides the point.

Twilio’s earnings PRs usually don’t focus on this reality. We have to go back to the 2017 Investor Day presentation to see just how overwhelming this revenue distribution is (91% corresponds to the business highlighted above):

After the SendGrid acquisition, this will likely “look better,” if email is bundled into Application Services Revenue - however, that's not organic growth. But still, it’s obvious that Twilio’s core business runs around processing SMSs and phone calls.

For sure, one could easily say that it’s like this because those were the first two segments the company launched. However, there is actually a much deeper reason for this.

The reason has to do with how the telecom market is organized. Within a given country, you have many telecom operators; several fixed line operators and several wireless operators. Then, internationally, you have the same thing times the number of countries you want to have a presence in.

Now, as a firm, you could use a single telecom’s API to interface with this huge telecom network, but then you’d have two significant problems:

• Domestically, you’d be charged different, higher, rates when issuing phone calls or SMSs to other domestic operators different from the one you’re tied to.
• Internationally, you’d not only be charged different, higher, rates to issue phone calls or SMSs to other international operators, but these rates would be outlandishly high. These rates would make it economically prohibitive to handle international traffic from your home base. Significant decentralization and cost would be added to the operation.

It's here that Twilio comes in. Since we’re dealing with programmable communications, instead of your API interfacing with a telecom operator, it will interface with Twilio’s virtualized network infrastructure (consisting of servers within a cloud provider, running Twilio communication software).

The company’s virtualized network infrastructure will then interface your calls or SMSs with each relevant telecom operator at a local level. Thus, both domestically and internationally, telecom costs will always be based on local (and optimized) rates (plus, possibly, volume discounts).

There are three consequences here:

• A single (Twilio) API gives your app local telecom access to all telecom operators Twilio is connected to. And it does this at rates based on that telecom operator's local rates.
• This system design disrupts inter-telecom domestic rates. There is a level or arbitrage versus domestic telecom rates.
• Most importantly, this system design disrupts much higher inter-telecom international rates. There is a large arbitrage versus international telecom rates.

For nearly all but the very largest customers, it would be unthinkable and too expensive to go around establishing 1,000+ API connections to each operator in all godforsaken territories to achieve the same result. Hence, the adoption of this kind of API and network structure can be expected to grow like wildfire.

This trend is made even stronger by the emergence of worldwide apps which have to work in countless countries. Think Uber, WhatsApp (FB), etc. It’s not a coincidence that they are customers. The trend is stronger both because we’re talking about many more telecom connections being required, and because the arbitrage on international telecom rates presents a much fatter target.

This is seen in Twilio’s statistics as well (these don’t come often, so again, we have to rely on the 2017 Investor Day presentation).

While most customers are domestic (red highlight is mine)...:

... most usage is international (red highlight is mine):

This, again, is to be expected. After all, connecting to a limited number of domestic telecom operators is much easier than doing the same to the plethora of foreign telecom operators.

We already saw the distribution of revenues between the core business (phone calls and SMSs) and everything else. That distribution flows from the nature of the business and what’s attractive about it (the problem it solves and the attractive economics) to potential customers. Something else also flows from the nature of the business: Twilio charges by usage and not some other way (subscription, etc.).

The reason is simple. Twilio is basically, in its nature, a reseller of phone calls and SMSs. Its reason to be comes both from this reselling being attractive because of the arbitrage it presents (versus telecom operators, especially international telecom operators), as well as the single-API nature of such arbitrage (hard to replicate for an end-customer). This too is borne out by Twilio’s statistics:

Since Twilio is a reseller of this usage, never could the company base its business model on subscriptions. This reality is also evident when it comes to COGS (Cost Of Goods Sold). Twilio overwhelmingly charges for usage, and then it overwhelmingly pays back to the telecom operators for such usage:

Hence, the company's usage price can be thought of as “Telecom operator fee plus markup.” Exactly like you’d expect for a reseller of such service (though enabled by a complex virtualized communications network, able to drop off the traffic on favorable terms, at the local level).

At this point, you’ve probably understood Twilio’s business. Of course, the company is trying hard to diversify away from this underlying nature. For instance, adding other communications abilities (API for WhatsApp, video, authentication, call center building blocks, etc.). But still, its business remains overwhelmingly as described.

There’s something of a structural reason, as well, for the business not to shift. Although it’s sexy to think that Twilio provides a single API for all communications, one needs to understand that the problem existing in the telecom world (thousand-plus operators, thus thousand-plus APIs) does not typically exist elsewhere.

For instance, if you want to access WhatsApp worldwide, you’ll be dealing with a single API. Thus, the attraction isn’t there to access it through Twilio and pay for it. You pay to access a thousand different telecoms through a simple API because otherwise you’d sustain a lot of cost and complexity to access them directly. Remove the “lot of cost and complexity” and the incentive to pay is gone.

Now Here’s The Problem - Competition And Margins

You could say that Twilio has a level of customer lock-in because of developers' familiarity with its API, and it being the most complete, etc.

But the thing is, you need to understand the problem Twilio solves. The problem it solves (and thus the reason its revenues are like they are) is the telecom arbitrage (always local rates) and connection problem (thousand-plus telecoms). That’s the value that customers pay for. Everything else is peanuts. Moreover, this problem is solved on a usage basis, and usage is opex (operating expense) for the customer. Opex is always a target for optimization (indeed, a part of the Twilio's business reason of being is one such optimization, the telecom arbitrage I described).

The thing is, there are others which have the same exact capability to solve the same exact problem for customers. They didn’t arrive at it brilliantly and purposely like Twilio did. But they still got the same underlying capability through another route. I am talking about other companies which developed SIP trunking solutions. SIP trunking is what enables a phone network (PBX - Phone Exchange System) to interface with the internet.

So here’s the thing, lots of VoIP (Voice over IP) companies had to develop an extensive SIP trunking infrastructure. They had to do this because they needed to take an internet inbound or outbound call and be able to pass it along to a physical (non-VoIP enabled) phone. They were servicing regular enterprise and individual customers, not app-centric enterprises. They weren’t prescient to think about this whole “unified API” thing, but they nonetheless put the infrastructure in place to provide it!

They had to put in this infrastructure because one of the main premises of VoIP was that it would be cheaper (it would run at no marginal cost for 2 internet phones, or pay telecom rates for an internet call originating or ending on a regular phone). VoIP could only become competitive if, when going through regular telecom networks, it didn't end up suffering from long-distance rates. As a result, the VoIP SIP trunking networks had to have a local presence in a similar way to Twilio's deployment.

Given this, now that the VoIP companies see the communications API segment exploding, they can enter it easily. And the thing is, there are lots of VoIP companies with SIP trunk capability. So, there are a lot of potential competitors.

Now remember, Twilio’s business model is basically reselling phone calls + SMSs at a markup. These companies are ultimately similar, but they need to enter the market for this undifferentiated product. So, how will they enter it? You guessed it, price.

Of note, before you think "this is technically hard for the VoIP companies," it's not - they were already able to do the core functions which Twilio performs that are valuable to customers. They already had to interface with hundreds or thousands of telco APIs/PBXs. Developing one more customer-facing API would never be much trouble.

This brings us to Twilio Q1 2017’s announcement that Uber would be relying less on the company. At the time, it seemed that Uber was going to replicate Twilio’s infrastructure (an undertaking in size, but not a major undertaking in technical terms). Even Twilio’s CEO commented that it was the nature of the business for something like this to happen (Q1 2017 earnings call):

As a result Uber grew as a percentage of total revenue in each quarter throughout 2016, ending the year at 17% of our revenue in Q4. In the first quarter of 2017, Uber accounted for roughly 12% of total revenue and as Lee will outline in a moment, we expect their contribution to decline further this year. Uber continues to grow rapidly, but they’re changing the way they utilized and consumed communication services.

Previously, they used our platform to support most of their used cases in majority of their operating territories. Now, they’re optimizing by used case and by geography, resulting in a more active multi sourcing program. In addition, they plan to move communications for some of their used cases inner. We believe that Uber will remain an important customer for us going forward, but their tremendous growth and the resulting magnitude of their communication spend has resulted in this change and approach.

While it’s not uncommon for large companies with mission critical used cases to dual source key technologies, communication included. There are very few companies in the world with the global scale and engineering promise to take on a project of this magnitude. Most companies are unwilling or unable to navigate the complexities of integrating into multiple vendors and optimizing on a geographic level for each specific used case, much less taking on the risk inherent in doing so.

Notice that emphasis is put on Uber in-sourcing the same functions. But it also talks about a multi-sourcing program. As it turns out, yes, Uber multi-sourced. For instance, Vonage (VG) Nexmo lists it as a customer.

Vonage is a VoIP company. It’s a straight example of what I described above. VoIP companies by and large had a similar infrastructure setup dedicated to a different use case. Now, seeing the segment growth, they make this infrastructure serve the same use case: reselling phone calls and SMSs to programmatic customers.

Given this, the main business Twilio is in is intrinsically competitive. There has been no gross margin expansion over time. It has been stuck around 55% forever. Moreover, although Twilio just bought a ~$160 million/year 79% gross margin company (SendGrid), it still kept its overall gross margin guidance around the mid-50s. This is obviously strange, since just keeping both gross margins (Twilio and SendGrid’s) stable would yield a 57.7% gross margin for the aggregate during 2019. Hence, Twilio likely expects continued gross margin pressure.

In my view, the company’s nature as a phone call + SMS usage reseller means that the business will be intrinsically margin-challenged going forward (and increasingly so). Diversification towards higher-margin areas hasn’t yet proven to be able to combat the fact that the main business is structurally bad.

Another Problem - Cash Flow

There’s yet another problem with Twilio - the company being billed as CPaaS and having extremely high revenue growth has led to it carrying amazing valuation multiples. It now trades for around 18x forward 2019 revenues. The crazy valuations are common to other IaaS/PaaS/SaaS companies.

Yet, Twilio is hardly like most other SaaS or PaaS companies. These companies usually sell subscription contracts ahead of service delivery. We’ve already seen that Twilio is different and basically bills for usage.

The typical PaaS/SaaS company generates significant cash flow due to upfront billing together with service growth. This leads to an ever-increasing deferred revenue/customer deposit balance. Twilio doesn’t work like that. There’s hardly any upfront billing, and deferred revenue balances are minimal. Hence, Twilio structurally generates little operating cash flow. In fact, if we strip out stock-based compensation, the firm routinely posts large negative cash flows.

Is There A Saving Grace?

From what I described above, it’s easy to understand that Twilio’s business is structurally bad and challenged. Competition in the space will be intense and margins will be minimal long term. That’s similar to what happened to VoIP players.

However, there’s little doubt that programmable communications, specifically phone calls and SMSs, will still provide a significant volume tailwind, possibly for years to come. Hence, Twilio has significant ease in posting large revenue increases and attracting new customers. This means it can be seen as a growth company for years to come.

It’s possible that a large slice of the investing public will easily ignore the structurally bad economics of the space, as well as valuation, as long as this optical growth persists (unless margins collapse in a very obvious fashion, I’d say).

A Small Aside

Twilio publishes a “Dollar-Based Net Expansion Rate.” This is commonly over 100%. 130%. 147% most recently.

This metric is badly understood. Lots of commentary refers to this rate as if it represents a 147% growth rate. It does not - 147% represents a 47% growth rate. This is due to the calculation method:

The Dollar-Based Net Expansion Rate is the quotient obtained by dividing the revenue generated from that cohort in a quarter, by the revenue generated from that same cohort in the corresponding quarter in the prior year.

A simple quotient will mean that stable revenues show up as 100%.

47% is still a very good growth rate, but I’d like to clear this confusion up.

Conclusion

I take three conclusions from this exercise:

• Twilio’s main business is structurally bad, economically. It basically consists of reselling phone calls and SMSs, through a scheme (network virtualization, local feeding into the telco grid) which undercuts telecom operators as well as presents a unified API for dealing with numerous telecom operators. While this presents an attractive arbitrage versus telecom operators, this is also open to other companies doing exactly the same.
• VoIP companies are uniquely positioned to replicate the same business. Hence, they’re doing it. An example is Vonage taking away Uber business. This CPaaS business will be eroded until only small markups and margins are left. Thus, Twilio’s main business gross margins are unlikely to expand meaningfully (in % terms).
• However, Twilio is likely to keep on showing extreme revenue growth for the foreseeable future. In spite of the bad economics, the business model lends itself to see significant volume growth, due to the underlying telecom arbitrage as well as the emergence of multinational apps requiring the communication capabilities this approach delivers.

Overall, I expect Twilio’s bad economics to result in a (very) bad outcome for investors in the long term, unless the company is able to launch some other significant business to replace its main business. However, short term Twilio will look like an explosive growth company, and this can easily attract growth investors at literally any valuation level.

Disclosure: I/we have no positions in any stocks mentioned, but may initiate a short position in TWLO over the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.