“You can’t always get what you want
But if you try sometimes
You just might find
You get what you need!”
– Rolling Stones, You Can’t Always Get What You Want
Most executives in Silicon Valley take it for granted that selling software is a better business model than selling hardware. In their minds, this goes without saying. The self-evident reasons relate to software’s remarkably high gross margins. With near-zero variable costs, software businesses offer the ultimate in scalability. Software businesses are simultaneously less capital intensive than hardware. This combination of low capital intensity and high gross margins, also leads to better valuations in the public marketplace. What’s not to like?
Despite these seemingly irrefutable advantages, many startups are “choosing” to sell hardware instead of software. “All things being equal,” software may indeed be the better model, but therein lies the issue. There is an old joke where a man is crawling beneath a streetlight looking for his lost keys, despite that he readily admits he lost them at a location further away in the dark. When asked to explain the location of his search, he naively replies “the light’s better”. While the “light” may be better with a software business model, these models are becoming difficult to execute – particularly for a startup. With tactical execution and business model choice highly interdependent, having the “better” business model, but a simultaneous inability to succeed, is a futile exercise.
There is a silver lining. The industry has changed in ways that improve the “business model” elements of selling hardware. The key driver is the standardization and general availability of hardware components, particularly those used in generic Intel-based 1U servers. As a result, the hardware is not a proprietary design, but rather a type of packaging. Think of it as an alternative to a cardboard box. This, combined with the proliferation of Ethernet, TCP/IP, and license-free operating systems such as Linux and BSD, allows a company whose primary competitive advantage is software to deliver that software in a box – a hardware box.
One might ask why a company would choose a hardware model in such a scenario when 1U servers are clearly available from many reputable vendors. Ironically, the generally accepted practice of delivering software designed for open platforms to be installed by the customer at the customer’s location may be up for review. When you examine the details of what it takes to sell, provision, manage, and rely on a particular piece of software, the customer increasingly prefers closed-end hardware to open platform software. On top of this, for a startup, this “software in a box” path is actually more cost effective and easier to execute.
Consider these issues:
1) Development Complexity / Quality Assurance: One might assume that it would be easier for a startup to attempt an “all-software” development plan rather than one that includes hardware. However, when implementing on an open platform, the developer is responsible for ensuring that the code will work properly in every potential execution scenario. When you consider the number of different hardware vendors, OS flavors (and versions), peripheral vendors, peripheral driver versions, etc, you end up with a near-infinite test plan. Choosing a single and static hardware deployment platform dramatically reduces the risk and the costs of software development.
2) Performance: Any software vendor concerned with performance will intrinsically favor a single hardware implementation, as opposed to running on multiple, and potentially unknown configurations. With fewer degrees of freedom, it is possible for the developer to tune the performance of the hardware and the systems software (OS + drivers) to the critical task at hand. In addition, the designer can eliminate code and/or features designed to support tasks that are unnecessary in this implementation, freeing more resources for the specific function. This can also help reduce cost.
3) Security: One of the hottest issues for CIOs today is security, and here again the hardware delivery mechanism has an advantage. A closed box design is much harder to penetrate than an open platform system. Also, by intentionally designing the system to have limited points of entry, the available security holes are minimized. Lastly, the security design can be hard-coded into the system, with no opportunity for removal by the user, or more importantly, by an intruder.
4) Provisioning: When a customer receives a 1U server with a single on/off switch and a single I/O port (Ethernet jack), implementation is fairly straightforward: plug it in, and turn it on. One huge negative result of open platform software is that IT departments have been burdened with the tasks of installing, provisioning, tuning, and trouble-shooting software. More degrees of freedom lead to more ways to screw up the process. Software delivered in a closed box is inherently simpler, and easier to provision.
5) Reliability / Stability / Customer Service: When open platform software fails, there are many help desks one can call, all of which will be eager to place the blame on someone else. With software sold as a single closed server, there is only one point of contact. More importantly, because of the single static implementation, there are dramatically fewer potential issues. This is extremely important for applications that live closer to the network layer where system interruption is unacceptable.
6) Pricing: It is increasingly difficult to establish and defend baseline pricing for software. As before, this is more important if the functionality of the software is closer to the network layer than the application layer. In this space, hardware is typically purchased and software is given away to support the sale. Perhaps it is simply a perception issue – people are more comfortable paying for something they can touch, see, and feel. This is accentuated by the fact that many software applications are available as an open-source in some shape or form. And while your software may be clearly better than the free alternative, free is a tough price point with which to compete.
7) Distribution: One of the key reasons to consider an open platform software approach is to potentially leverage the customer base of the hardware vendors that are already in the market. Admittedly, this worked great for Microsoft with IBM and Veritas with Sun. However, this is no easy task. Hardware vendors have learned from their mistakes in the past and are much less likely to sign favorable distribution deals. Also, hardware vendors are unlikely to consider distributing a product until it is proven in the marketplace. Since one must first establish early traction with direct sales, there is no reason not to adopt a “software in a box” model from the beginning. You can always OEM the software to a hardware vendor at a later date.
One market in which this phenomenon has already begun to play out is the firewall market. Check Point pioneered this space with a software focused, open platform business model. Later, vendors like NetScreen began taking market share with a hardware based delivery platform. In a June 2002 Internet Security report, Bob Lam suggests that over 50% of the 2001 firewall market was hardware based. Moreover, by 2005, Bob predicts the market will be 70% hardware based. Why would hardware vendors gain share going forward? According to Bob, “Check Point appears to be at a disadvantage because of its software based architecture, which some customers view as more difficult to install and less robust than comparable products on the market.” Responding to this threat, Check Point has parted with companies like Nokia to deliver closed box systems based on their software.
Admittedly, this theory applies much more to functional software such as firewalls, web servers, security devices, and storage solutions, than it does to high-end applications. You shouldn’t expect SAP to ship a hardware device anytime soon. Functional software applications are more task-oriented and require less customization than high-level business applications. Additionally, functional applications are much more likely to be judged on criteria such as performance, throughput, reliability, and availability. All of these things favor the static hardware implementation.
The three reasons we initially discussed for favoring the software delivery model were higher margins, reduced capital intensity, and higher valuations. Obviously if you are passing through hardware sales, it will be impossible to mimic 90% software margins – but 70+% margins are not out of the question. In addition, today’s custom manufacturers are more than willing to assemble boxes after the order (a la Dell), and drop ship them to the end customer. This virtually eliminates inventory risk and has a very positive impact on capital intensity. Most importantly, however, Wall Street appears to understand the value of this new approach. NetScreen currently trades at 6.5X forward sales – which is more than you can say for most software companies these days.