Investor Guide - Terminal FCF Margin (Pt.1)
An extended preview of Part 1, where we dive deep into the factors impacting terminal gross margin estimates for high-growth software companies.
Intro
The terminal Free Cash Flow (FCF) margin greatly affects a company's valuation, regardless of whether an investor uses a DCF method or anticipated future FCF multiples. The principle 'cash is king' is pivotal to investors. A company with abundant cash ensures sufficient liquidity, reduces dependency on debt, and provides flexibility for investments. Cash also shields against economic downturns and asset devaluation. Furthermore, as investors we view a strong cash position as a sign of potential reinvestment or dividends and buybacks, boosting capital appreciation and returns.
Predicting a terminal FCF margin is challenging, however. For growth investors, many companies being analyzed may have low or negative FCF margins, making long-term forecasting fraught with estimation risk. Yet, this risk has the potential to generate the most alpha because most of the value for such companies resides at the long-end of the forecast period. For growth investors focused on tech, long-term estimating becomes even riskier because we are often assessing companies in young markets, with unique technological architectures. This makes a comparables approach to terminal forecasting more fraught with estimation risk.
The FCF margin primarily consists of revenue, deducting COGS, S&M, R&D, G&A, and factoring in D&A and SBC, changes in working capital, and capex. Each element is influenced by various factors and such analysis is especially difficult for tech companies due to additional tech-specific variables that investors must take into account.
We have recently constructed a number of frameworks for evaluating and estimating terminal FCF margin for B2B software companies. We’ve created a Terminal Gross Margin Framework, a Terminal Opex, Non-Cash & Capex Framework, and a set of points of consideration for Terminal Working Capital Changes, which collectively cover the high majority of FCF components. The frameworks provide a systematic approach with a scoring system for evaluating terminal gross and FCF margins.
The full details of the Terminal Gross Margin Framework are available in Part 1 of this research series, available at convequity.com. The full details of the Terminal Opex, Non-Cash & Capex and points of considerations for Terminal Working Capital Changes will be published in Part 2 and Part 3 during the next couple of weeks. In this article we will touch on some key points from Part 1.
The Four Factors Influencing Terminal Gross Margin
After a long while of contemplation, discussing the most influencing factors, and much back and forth tweaking the framework, we decided on the following four factors of influence on a software company’s terminal gross margin:
Marginal costs
Talent
Vertical integration
Value proposition
The objective is to evaluate how each of these factors will impact a B2B software company’s gross margin as it grows toward its mature stage. These factors are not mutually exclusive as there is some overlap (i.e., a company needs top talent to achieve vertical integration and a profitable value proposition). Also we are not claiming this is an exhaustive list of terminal gross margin factors. Nonetheless, the factors serve as a useful mind map of sorts, if not a perfect systematic approach, for evaluating terminal gross margin.
The company being analyzed should be in a growth phase with significant scaling potential; otherwise, these factors might already be reflected in the gross margin. Current strengths may not be evident in the existing gross margin due to the company's current scale or strategic direction.
For each terminal gross margin factor there are category labels: Very high, high, Mid-High, Mid, Low-Mid, Low, Very low. The company will be assigned one of these category labels for each factor of terminal gross margin. Each category label translates to an increase in basis points (bps) which is added to the current gross margin. If the company shows promising signs of being able to leverage a factor for gross margin improvement as it scales, it will receive a category label that translates to a high bps increment. This would suggest the factor can become a strategic advantage in the company's journey. Gross margin increments in bps are added together and added to the current gross margin to arrive at an estimate for terminal gross margin.
We won’t delve into the specifics of the scoring system here but we will briefly discuss the four factors. For full details on the scoring system, visit convequity.com.
Marginal Costs
A company with low marginal costs can sell to and onboard a new customer with minimal expenses, leading to low COGS and a high gross margin. Bill Gates noted in a 2000 interview how technology companies benefit from scale economies, making it extremely cost-effective to add new users. He highlighted that for software companies, the marginal costs can be almost nil. While they may approach zero in certain aspects, there remain discernible marginal costs in the software industry, albeit much lower than in other sectors.
While the software industry generally has lower marginal costs, there's still a range among companies. To better understand a software company's marginal costs at full scale, we've devised a structured approach. We've segmented the costs based on: 1) revenue model (license vs SaaS), 2) degrees of multitenancy, 3) information entropy, and 4) platform breadth. Analyzing these factors aids in estimating a company's terminal gross margin.
Revenue Model
A software company’s revenue model has a big impact on marginal costs, both current and terminal, though there are a lot of nuances and it isn’t so clear cut. At a high level, a software company can use a license-based or SaaS-based revenue model. Generally, after deep deliberation it seems as though SaaS-based companies do have lower marginal costs at maximum scale (which might be the common perception among software investors), in large part thanks to the usage of cloud computing and the recurring revenue model. However, there are many dynamics that may not make this the case when comparing two companies, one with a license-based and one with a SaaS-based revenue model.
License-based software comes in varieties like perpetual, subscription, and consumption-based, each influencing marginal costs differently. Revenue recognition policy, either sell-in or sell-through, also affects costs. A perpetual license vendor using sell-in recognizes revenue upon sale to a channel partner, likely resulting in minimal involvement in software installation and lower costs. In contrast, using the sell-through method, a vendor recognizes revenue when the end-customer can fully use the software, implying a more hands-on approach with channel partners and higher marginal costs. Additionally, installation complexity and the customer's IT expertise can dictate the extent of the vendor's involvement in successful onboarding.
Generally, B2B SaaS providers benefit from reduced marginal costs due to their cloud-centric deployment approach. When they acquire a new customer, there's no requirement for software setup on the client devices or on the on-prem servers. This is because the software is already active and hosted in the cloud by the SaaS provider. Users simply access the software via their browsers over the internet. As a result, theoretically, the costs involved in integrating a new customer for SaaS providers are less than those of license-based vendors. It is also interesting to note that a license-based vendor's marginal costs are mostly associated with software installation, whereas the SaaS vendor's marginal costs are associated with the compute, storage, data transfer, and networking required to deliver the software as a service over the Internet.
It may not always be the case that SaaS has lower marginal costs at scale than license-based software, however. It depends on the type of SaaS revenue model and the degree of multitenancy within either model. SaaS-based software can be largely divided into seat-based and consumption-based, and each have different marginal cost dynamics.
Consumption-based SaaS vendors, like Cloudflare for CDN, Twilio for API services, and Snowflake for data analytics, experience a wide range in resource consumption, making a consumption-aligned revenue model most apt. In contrast, seat-based SaaS vendors, including Salesforce, Microsoft, and Adobe, typically experience lower resource consumption, which is both stable and predictable, and the usage difference between power users and average users is smaller. As a result, SaaS providers find it viable to charge based on the number of users. This pricing suits when many employees need software access.
Economically, seat-based vendors generally enjoy lower marginal costs and potential for higher terminal gross margins due to pricing and the benefits from factors like Moore's law. Underutilization versus what is priced for, leads to gross margin accretion, and usually seat prices remain the same of even increase over time, while Moore’s Law dictates resource costs will decline. Meanwhile, due to standards in transparency, consumption-based vendors often pass on cost savings to clients, thus forfeiting gross margin accretion for competitiveness.
As investors contemplating the terminal gross margin of a software company, consider the chosen revenue model and how it may influence marginal costs at maximum scale.
Degrees of Multitenancy
The level of multitenancy varies between SaaS providers and has a significant effect on marginal costs, impacting possible terminal gross margin. Multitenancy, the technology backbone of SaaS, allows users to share computing resources, aiding vendors in reducing costs and scaling quickly. Its evolution includes:
1970s: Clients shared servers.
1980s: Departments shared data centers.
2000s: Companies shared clouds.
2010s: Users shared cloud-based machines, operating systems, and apps.
Different multitenancy levels affect marginal costs:
Shared Everything: Every resource is shared among tenants, maximizing efficiency but demanding careful software design for data privacy.
Shared Infrastructure and Software, Separate Databases: While infrastructure and software are shared, databases remain distinct, enhancing data isolation.
Shared Infrastructure, Separate Software, and Database: Only the infrastructure is communal. This provides a balance of efficiency and isolation but is more resource-intensive.
While optimal efficiency is at level 1, enterprises often lean towards levels 2 or 3 due to stricter data requirements, raising vendors' marginal costs. However, targeting enterprises remains very valuable for SaaS vendors, because the despite the higher costs the vendor can sell its SaaS at a higher multiple of COGS – we discuss this in more detail in the Value Proposition factor.
Leading SaaS companies like Salesforce have combined high multitenancy with enterprise requirements, resulting in substantial margins. In contrast, Microsoft's venture into extensive multitenancy, while catering to enterprises, has seen numerous vulnerabilities arise. Some vendors unable to marry the two often offer reduced multitenancy to enterprises (e.g., self-managed cloud-based software).
In conclusion, SaaS vendors maximizing multitenancy stand a better chance of reducing marginal costs, scaling, and maximizing potential terminal gross margins during the company’s maximum scale.
Information Entropy
Information entropy, or data randomness, significantly impacts a vendor's marginal costs. A simple way to consider entropy is that low entropy data contains less noise, and high entropy data contains more noise. Hence, for the same amount of information (data that is useful and valuable), high entropy data needs more storage and processing than low entropy data.
With that, lower entropy signifies more structured, predictable data, leading to reduced costs. For instance, collaboration SaaS platforms like DropBox and Monday manage structured textual data, hence have low entropy and reduced resource needs.
In contrast, vendors like network security firms deal with high entropy, unstructured data, necessitating greater computational resources and bandwidth. This is compounded when data must be processed in real-time, as such data is inherently high in entropy. Similarly, video conferencing or 3D modeling software, due to the nature of their data, face higher entropy and thus increased costs.
However, effective data compression can counterbalance these costs. If high entropy data can be efficiently structured and compressed, it can result in lowered costs and increased gross margins, though not all vendors are in such a position to do so. To sum up, the type and entropy of data directly shape a vendor's costs, with lower entropy generally leading to reduced costs and higher potential terminal gross margins once the company has reached maximum scale.
Platform Breadth
Platform breadth significantly influences marginal costs. Upsells and cross-sells to current customers are more cost-effective for both SaaS and license-based vendors. For instance, adding a solution for a consumption-based customer or upgrading a seat-based customer often incurs lower costs since existing infrastructures can be utilized. Similarly, adding features to license-based software is cheaper once initial setups are in place.
Broad platforms enhance customer loyalty and retention, as they have more scope to satisfy future emerging customer demands. Furthermore, the more services a vendor provides, the more of their customers' data they hold, which in turn increases vendor stickiness and contributes to higher ARR, NDR, ACV, and revenue, and lower marginal costs. Thus, software vendors with wider platforms can achieve lower marginal costs, especially when existing customers dominate.
To incorporate platform breadth into a vendor's terminal gross margin, investors should first evaluate if the platform is expansive or limited. If it's limited, they should ascertain whether the vendor has previously shown product innovation indicative of potential platform expansion in the future.
Incorporating Marginal Costs into Terminal Gross Margin Estimates
To incorporate all the above into the terminal gross margin estimate, we categorize the vendor against each subfactor, sum up the respective basis point scores, and then add this to the current gross margin. Remember that this framework is to be applied to companies that have not yet reached their potential scale, have ample expected growth in the future, and are likely to leverage these factors, such as favourable marginal costs, as they approach maximum scale. The process entails taking note of such a company's current gross margin, considering each component of marginal costs, and then adjusting the gross margin accordingly.
For full details visit convequity.com.
Talent
The software industry is driven by knowledge and innovation, with intellectual property and problem-solving central to a company's competitive edge. To lead in innovation, software firms highly value top-tier engineering talent. These experts spearhead R&D breakthroughs and ensure efficient product delivery.
A key role they play is reducing marginal costs. They achieve this by curtailing bugs, enhancing user experience, and building a flexible architecture on sturdy software infrastructure, thereby decreasing onboarding and support costs, and boosting gross margins. By introducing important features, engineers also reduce update needs and associated costs, and they creatively use existing technologies to further reduce expenses. Important to note, is that great developers and engineers enable a company to expand into new products and markets, thereby minimizing the need for M&A and the ensuing amortization of acquired technology recorded in COGS.
Not only do elite engineers minimize COGS, they also maximize the value proposition, which is a multiple of COGS. If a software firm has great engineers, and a culture that fosters frugality, they could produce and deliver a product at very low COGS and generate massive value such that S&M can sell the product at 5-10x COGS and still be competitive.
In consumption-based license software, performance often correlates with usage and CPU cores. Engineers aim to produce scalable software that supports rising consumption without compromising integrity. For example, while many software applications are designed for 8 CPU cores, outstanding engineering can support up to 40 cores without redeployment.
For seat-based SaaS, adept engineers maintain a cost-effective and robust architecture even as user count rises dramatically. In consumption-based SaaS, sophisticated design ensures scalability and adherence to service levels amid fluctuating usage. Case in point: Cloudflare maintains web functionality during DDoS attacks, whereas Akamai requires an upgrade for superior DDoS defense.
Incorporating Talent into Terminal Gross Margin Estimates
Generally, the framework submits a high basis point score for software firms that have thus far demonstrated great engineering talent. It is assumed that the current great engineering talent will attract more elite talent, continues to enrich the company culture, and will lead to many more future innovations that maintain or increase gross margin as the company approaches maximum scale.
For full details visit convequity.com.
Vertical Integration
Vertical integration offers companies enhanced control over supply chains, resulting in efficiencies and differentiation. To navigate this for software companies, we have divided the tech stack into four layers - software, data, infrastructure, and silicon.
Software Layer
Tech firms usually debut with unique software applications. As they grow, so does the complexity of their software, strengthening their market position. However, evolving software inevitably accumulates technical debt, hampering a firm's ability to consider vertical integration. Technical debt not only burdens cognitive capacity to explore vertical integration, but it also raises costs, impacting the company's gross margins. This is a good layer to start the vertical integration evaluation and how it can be accretive to gross margin, because if technical debt is already high at the software layer, it doesn’t bode well for future plans to vertically integrate upstream. As outside investors, taking note of new feature release cadence, industry reputation, and the engineering transparency (via blogs etc.), are some ways to make inference as to the level of technical debt of software company.
Data Layer
Most flourishing software companies find off-the-shelf data solutions inadequate for their unique needs. For example, many cybersecurity vendors use SIEMs not tailored for their specific requirements, resulting in higher costs. Companies like Uber, Netflix, and Meta, unable to find market solutions fitting their data needs, built custom systems: Hudi, Iceberg, and PrestoSQL respectively. In cybersecurity, Palo Alto Networks, SentinelOne, and CrowdStrike developed or acquired tailored data layers. Owning a custom data layer can result in significant COGS savings, both from optimizations and by cutting third-party system costs. As an investor, identifying those software companies with plans to vertically integrate with the data layer, could yield significant alpha if you grasp the implications before the market.
Infrastructure Layer
Though SaaS companies enjoy many benefits over license-based peers, competition is pushing some towards vertical integration at the infrastructure layer. There is evidence that there are diminishing marginal returns for large-scale SaaS companies, in large part brought on by the popularity of multicloud operations. These operations, aiming to avoid vendor lock-in, incur costs like data egress fees and complexities of managing multi-cloud security protocols, which inflate the COGS of SaaS vendors.
By shifting some or all of the infrastructure away from cloud over to colocation data centres, not only can this drastically reduce costs but it can also enhance software performance, security, and enables vendors to better meet stringent customer requirements in highly regulated industries.
Thus far, only a handful of cloud-native vendors have transitioned a major part of their operations to self-hosted infrastructure. This is due to the colossal risk and level of engineering talent required to bring such an infrastructure overhaul to fruition without major problems. Yet, market dynamics and investor expectations are likely to influence more companies to follow the examples set by DropBox and Veeva. While cost savings is a leading advantage, there are other merits such as improved performance and security, as well as meeting data isolation and residency needs.
Silicon Layer
Custom silicon is the ultimate goal for companies aiming for complete vertical integration, offering transformative performance and efficiency benefits. By opting for tailored silicon, software businesses can optimize their applications, avoid inefficiencies of generic chips, and enjoy quicker processing at reduced costs. These chips can also consume less energy, leading to fewer expenses on power and cooling.
Having custom silicon means less reliance on external suppliers, granting better control over costs and supply consistency. This self-reliance can also give a company a unique edge in the market through exclusive features, enabling premium pricing. Additionally, a harmonized ecosystem of software and hardware can result in an optimal user experience and potentially command higher prices.
Yet, challenges abound. The financial burden of designing and producing a custom chip can be monumental, with costs for advanced nodes potentially reaching $500 million. Constant redesigns, like Intel's experience, can further escalate these costs. Success in this realm, as seen with Fortinet, demands an extended vision. Short-sighted plans risk being overtaken by technological and market changes, threatening return on investment.
In essence, while the allure of custom silicon is undeniable, it comes with challenges like significant capital requirements, prolonged development phases, increased operational complexities, and potential system-wide disruptions. Ultimately, this is why so few software firms have ventured this deep into upstream integration.
Incorporating Vertical Integration into Terminal Gross Margin Estimates
To incorporate the vertical integration factor, investors ought to consider where the company under examination is right now, and how they may integrate upstream in the future. Such information may be gleaned from announcements by the company’s management.
Value Proposition
The previous factors have mostly focused on the COGS side of gross margin. The value proposition is the other side of gross margin, calculated by price minus COGS. A vendor's success in selling products at high margins hinges on its engineering team's ability to address customer pain points effectively. A versatile platform that caters to most customer needs, with customization options, ensures both customer satisfaction and profitability. Conversely, platforms lacking comprehensive features lead to COGS as the vendor will need to customize for too many customers. They may still try to sell the product at 5-10x COGS, but surely, they would be uncompetitive if they have high COGS. Hence, the value proposition in terms of a COGS multiple is important, but so are the COGS itself. Another note, is that high-value products also attract competitors, so vendors should plan ahead for sales and marketing in anticipation of increased competition.
Incorporating Value Proposition into Terminal Gross Margin Estimates
As our evaluation takes a forward-looking approach, it's crucial to gauge the potential of a vendor's value proposition. How probable is it for a software firm to sustain or enhance their COGS multiple? Which companies are poised to follow in the footsteps of industry giants like Fortinet, Palo Alto Networks, Cloudflare, Snowflake, Datadog, and Zscaler, amplifying their value proposition, preserving or boosting their pricing capability, and bolstering their gross margins? Furthermore, is it feasible for these leading companies to demand even higher multiples of COGS for their services?
A Note on Non-Recurring COGS
Part 1 uses four factors to predict terminal gross margin, but the initial point is crucial. The current gross margin, if affected by recent large M&A activities and related technology amortization, can mislead. If a company regularly acquires, the current gross margin is fine. But for infrequent, major acquisitions, it's best to exclude technology amortization as it is likely a significant influencer on current gross margin. Due to these nuances, investors should adjust as needed.
Conclusion
These four factors influencing terminal gross margin are by no means fully exhaustive but they are helpful for contemplating the potential increase of a software company’s gross margin by the time they have reached maximum scale. In the full report, available at convequity.com, we run this evaluation framework for estimating Cloudflare’s terminal gross margin. We will apply it to other companies in the next few weeks.