On Aligned Cost Structures, Switching Costs, Distribution and Valuations in Enterprise Software

I’ve been getting a lot of questions recently about public and private enterprise software companies and if these valuations are sustainable. Many of these questions seem to stem from a general fear of switching costs, misunderstanding of product advantages, and high relative valuation multiples compared to other stocks. This post will be my attempt at unpacking what I think are 4 key aspects to understand when evaluating enterprise software companies.

Aligned Cost Structures – Works Like Magic

Aligned cost structures brings up the idea of top-down software vs bottoms-up (“product-led”) software companies. For case studies, I will use ServiceNow and Twilio to demonstrate both types of products respectively.

ServiceNow’s original product IT Service Management is a system of record for IT events (everything from security incidents, customer support, service tickets, etc.). Let’s take a look at ServiceNow’s key metrics to understand how they align their cost structure.

Unfortunately, ServiceNow does not show full average ACV but given the large enterprises and contracts they are booking, we can assume that the ACVs are fairly high for enterprise software contracts. Per the above graph, Service now in Q4’18 had an average G2K ACV of $1.7M, 24% growth YoY.

ServiceNow’s renewal rate is consistently in the high 90% range. Said another way, gross churn is between 1-3% and net retention is probably around 105-110% (depending on the size of the lands).

Here we can see ServiceNow’s average contract terms are close to 3 years. These are huge multi-year contracts that ServiceNow is selling to large enterprises that typically renew/expand as evidenced by the strong retention metrics.

Given that ServiceNow has strong retention, large multi-year deals, and high ACVs, we would expect to see significant sales & marketing costs as a percentage of revenue. The reason for this is the company can spend a larger upfront amount on acquiring customers due to the long and growing LTV expected from the customer. In many cases, LTV for systems of records could be even greater than analysts can imagine given how sticky these products are and the size of these contracts.

The key here though is even at the scale that ServiceNow is currently at, due to the dynamics of the customers they’re pursuing and the size of the deals, it still makes sense for ServiceNow to be spending nearly 50% of revenue on S&M. At steady state, we can expect that ServiceNow will still have a large component of S&M spending as that incremental customer warrants a large upfront CAC.

This is an aligned cost structure. At meaningful scale, ServiceNow is still investing 46% of its revenues into S&M due to its ability to secure long-term, high ACV deals with customers. Many research analysts from other industries balk at seeing these numbers when evaluating SaaS companies. It actually makes complete sense when compared to the company’s average ACV, retention, and contract lengths.

Let’s look at Twilio now, a company that has a product-led business model. Twilio develops a suite of communication API products targeting SMBs and large enterprises.

At the end of 3rd quarter 2017, Twilio had revenue of $100M or run-rate revenue of $400M. Using this amount divided by total customers gets us to an average ACV (imprecise for sure but we’re working with what we’ve got). This gives us an average ACV of $8.6k. As we saw before, ServiceNow’s ACV amongst G2K customers (again not quite apples to apples) was $1.7M.

Twilio also has very high revenue retention and this has grown as the company has penetrated more enterprise accounts. That being said, looking back over the past few years, we can see that Twilio had retention rates of 95% or 5% gross churn. This is compared to the 1-3% gross churn that ServiceNow sees.

Net retention has historically been close to 140% if not higher. This means customers are expanding by large amounts after the initial contract. I guesstimate ServiceNow has 105-110% net retention.

Finally we can look at Twilio’s S&M as a percent of revenue. Twilio spends close to half as much as ServiceNow on S&M. Put another way, Twilio spends significantly less on CAC as a % of revenue.

Taking a step back, is one model better than the other? Not really. They are just different. ServiceNow sells huge multi-year contracts that they expect to exhibit very little churn and to continue using the product for decades potentially. The average ACVs are much higher than Twilio. As such, ServiceNow can also spend more upfront on CAC to acquire these customers since they have long-term visibility.

Twilio on the other hand has much smaller initial ACVs, higher churn, and smaller absolute LTVs. For this reason, Twilio spends much less on upfront CAC to acquire customers and relies on product-led viral adoption to drive growth. Additionally, Twilio sales teams are able to get detailed usage statistics on the customers within larger organizations that are using the product and capital efficiently upsell these customers to larger enterprise software contracts; increasing ACVs, LTVs, and probably decreasing churn.

Both of these companies have performed extraordinarily well, rapidly outpacing the Nasdaq over the last 2-3 years. They both have very different business models but have done a great job of aligning their cost structures to their revenue opportunities. They are being rewarded by the market for this alignment.

The Misconception of “Modern” Switching Costs

Switching costs are a wonderful thing. Microsoft is perhaps the best example of this where the Windows OS for the PC got initial adoption by being easy for users to complete basic tasks. This led to more third-party developers building applications for Windows which in turn led to more users and improved the OS experience further. This network effect made it such that Windows was able to have a long-term sustainable competitive advantage leading to literally billions of dollars of free cash flow that the company could use to reinvest back into growing the business and return to shareholders.

The predominant example of traditional switching costs in “modern” enterprise software is systems of records. Salesforce’s customer relationship management (CRM) platform is a great example of this. Customers store contact data, emails, tasks, notes, etc all in the CRM product. This data is the “record of truth” where data is carefully protected to make sure all interactions with external contacts throughout the whole company are logged. Because this system contains all of this data, third party developers and other software products create APIs or easy to use pipelines to call the data in the CRM and use it for analysis, sales tools, expense management, etc. Other companies can basically only increase adoption in their product if they make it easy for customers of Salesforce to utilize the data stored within the product. This is an extremely powerful position to be in as the switching costs increase with every new product that is used on top of the CRM’s data. This concept is important to understand as it will tie into the discussion on aligned cost structures below.

In comes the wave of product-led startups. Companies like Dropbox, Docusign, Smartsheet, Elastic, MongoDB, Slack, Zoom etc. These companies are all trading at or projected to trade at 10x-30x LTM revenue multiples. When evaluating switching costs, in many ways these companies look like they don’t have any in the traditional sense. For example, in videoconferencing, many people think there are low switching costs. You can swap GoToMeeting for Google Hangouts for Microsoft Skype for Zoom. In fact, we’ve probably used all of the above for different meetings. The issue with thinking of switching costs in the traditional sense is this does not account for changing paradigms.

In Product-Led companies, typically small teams, individuals, or business units can all buy the product separately from their overall company’s choice. The reason for this is individual teams are increasingly getting more budget to use the productivity tools, developer tools, or other software that they need. The buzzword for this is the “consumerization of IT” where everyone can bring their own software to work. This is not true in all cases but generally is an overall shift in budget and usage trends. In this new paradigm, commodity products such as videoconferencing just need to work. The way they differentiate is always working (or working better than other products), making it easy to use, and having cheaper pricing than the ROI they’re delivering. Hence, the switching cost is actually different with these types of software. Users are choosing to buy the software because they actually like using the product compared to other offerings. This switching cost is actually rather large if not intuitive. Once a tool is engrained in workflows it is hard to rip out. It’s not because of the technical challenge of replacing it. More that if it performs the task well and makes it easy to use, then there’s no reason for a consumer who opted in to using it to stop using it. A new competitor would either have to come in much cheaper (while still having the same availability metrics) or be 10x better to replace the existing product.

My favorite example of this is Wunderlist. Wunderlist is a simple to-do list app that is now owned by Microsoft. I use the freemium version (but would certainly pay a small monthly subscription to use it) and know there’s tons of productivity tools out there that have this incorporated along with Kanban boards, grids, and additional functionality. That being said, Wunderlist just works. So I have no need to replace it. This is the same thing with product-led companies. They make adoption easy which also may make replacement easier but the switching cost is actually stronger since the user is opting in to continue using the product.

Another example is Excel vs Google Docs. Google Docs is cheaper and works just as well for simple tasks. However, for harder tasks or modeling heavy tasks, Google does not have as easy functionality as Excel. Thus, the switching cost is actually pretty big. Now this could be classic disruption where Docs can move upmarket as functionality gets better over time, but at the same time Excel is not that much more expensive to make this a tradeoff. So as a user, I actively opt in to Excel over Docs and hence there’s a switching cost.

Distribution Matters

In enterprise software, distribution is more of a differentiator than product…Actually, I just wanted to make a provocative statement, they both basically work hand in hand, but distribution really matters.

One of the all-time great examples of distribution is Salesforce. When Salesforce first came to market, they had a product that revolutionized the enterprise software space, they essentially offered a managed service that was accessed through subscription fees rather than an on-premise offering that had upfront license and hardware costs. The user interface and experience were in some cases described as nominally better than Oracle/Siebel System’s offering but the deployment model was revolutionary. However, for the most part product does not sell itself, even product-led software companies still have to figure out distribution (i.e. freemium offerings, user meetups, product advocacy, etc.).

Salesforce started out by doing some aggressive gorilla marketing tactics. They had a launch party where they created an enterprise software “Hell” to show how bad current software experiences were. They used to do city tours to show Salesforce’s product off and eventually turned this into Dreamforce the annual conference. The reason for all of this showy marketing was to establish a brand as the underdog upending the traditional software experience and saving users everywhere. It was the beginning of their distribution channel.

Fast forward to today, Salesforce’s product arguably isn’t the best CRM out there. Its release cycles are about once a quarter while many competitors have release cycles once a week. However, it now has an amazing distribution pipe.

There are literally consulting businesses built around helping customers set up their own customized Salesforce. System integrators around the world help large organizations implement Salesforce’s CRM to get the professional services work. Salesforce has leveraged this immense channel partner ecosystem to then acquire new products, most recently Mulesoft, and push Mulesoft’s integration platform to its existing customers through its established direct salesforce and channel partners. When other people are working on your behalf to get you customers, that’s a healthy distribution channel. For this reason, Salesforce has been able to continue to grow rapidly even as it surpasses $10B in annual revenue.

Another example is Atlassian. Unlike Salesforce, Atlassian does not have much of a direct salesforce, but what it does have is a group of passionate users that helps spread word of mouth adoption. The company has enabled numerous integrations for various other software products to engrain Atlassian products in the customer’s workflow. On top of that, it has enabled a group of system integrators to start businesses advising on the best way to leverage the Atlassian suite. User conferences and meetups educate future users creating a flywheel effect that keeps feeding Atlassian’s customer growth. Once this distribution channel was established, Atlassian acquired products like Bitbucket and Trello to flow through their established distribution pipes and feed more cross-sell/upsell. It is a different way of building up a distribution channel compared to Salesforce but just as effective.

All of this is why distribution is so important. An effective distribution channel is hard to set up and takes time and experimentation to implement, similar to product/market fit. However, luckily there are many case studies of successful companies out there leveraging different methods to arrive at the same place.

Sometimes the best way to secure distribution is not actually the most scalable. In many cases, especially when a company is innovating in a new space or building a product that requires significant integrations with other systems, professional services can be a way to kickstart a distribution channel. Services can help software companies establish beachhead customers who then become references for future customers. Professional services can also help establish long-term customers (although there will short term cash burn) that produce cash flow to help sustain growth in the future. As the company scales, this services component can feed consultants and channel partners who have staff that are “on the beach” but can increase their utilization and profit by actually helping implement the product into customers. Paul Graham has famously said do things that don’t scale. This applies to every stage of growth in an enterprise software company’s lifecycle. For more on the potential benefits of professional services, I highly recommend reading this article by Martin Casado of a16z.

The Intriguing Aspects of Valuations

Valuation is an art that is sometimes confused as a science. The earlier the stage of the company, the more valuation is based on an ownership model compared to dollars invested and relative to the size of the TAM. As the company grows, valuations get based more on comparable companies whether through precedent acquisitions or market trading multiples. All of this however is art not science.

The valuation of a business at scale is the present value of its future cash flows. In some cases, this is not an obvious thing. Instagram was bought for $1B by Facebook with no revenue. However, currently there’s estimates by investment banks that Instagram is worth $150B and producing prodigious amounts of free cash flow already. The same goes for Google’s $660M acquisition of DeepMind. Again, I would assume the company had zero revenue at acquisition and maybe still now. However, DeepMind has helped optimize Google’s datacenter spend and energy spend. The amount of money that will save Google over the long-term probably drastically outweighs DeepMind’s acquisition price and continued operational costs.

In enterprise software, valuations are mostly quoted as revenue multiples. Companies are said to be valued at 15x NTM revenue or 10x NTM ARR. These again are proxies for eventual free cash flow generation. However, they’re necessary because building a discounted cash flow analysis early on in many of these company’s lives would have so many assumptions on it that the analysis would effectively be useless.

So what are the proxies commonly used for valuation? They include revenue growth, gross margins, LTV/CAC ratios, S&M efficiency, churn, upsell, runway, TAM, market share, etc. An exact same business with the exact same metrics will be valued more highly if it has 2 years of runway versus 1 year as the longer runway allows more time for growth and eventual higher free cash flow generation at steady state. For example, paying 50x NTM ARR for ServiceNow when the company had $2M ARR would have actually turned out to be a cheap valuation as the company ended up gaining leading market share in a massive TAM. Again though, investors have to balance risk/reward and the risk in paying such a high multiple is that growth does not catch up to that value in the cash runway allotted and the company is unable to receive future financing to pay for future growth. Bessemer Venture Partners recently published a great framework for valuing SaaS companies in its 2019 State of the Cloud report (picture below).

Given that valuation is dependent on future free cash flow generation, this is where business models, competitive advantages, network effects, switching costs, distribution come into play. Companies that have figured out a more capital efficient repeatable go-to-market in a large TAM will be valued at higher multiples or absolute valuations in comparison to other companies simply because at steady state, the expectation is that company can create more free cash flow as market share of the TAM grows.

For more resources on valuation, I highly recommend reading anything by Michael Mauboussin, especially this article here: Making Sense of Multiples or anything by Bill Gurley on his Above the Crowd blog, especially this article: All Revenue is Not Created Equal.

Sorry for the length here! Just had a lot of thoughts that I've been meaning to write about for awhile so figured I'd put it all together in one post. Feedback is more than welcome! Please feel free to share thoughts, comments, arguments with what I've written above so we can all learn more.