Software as a service

The software industry is less than thirty years old and in that short a time has undergone major changes in technology and business models. So whether you're an existing software company or a new venture it would be wise to understand the variety of business models open to you. This article simplifies the entire industry into seven basic models. The models range from the traditional model (Model 1) which has dominated business software and created modern day giants like SAP and Oracle, to the outsourcing model (Model 3) that powers IBM and EDS, to the newest software-as-a-service players like Webex, Concur, Taleo, Salesforce.com (Model 6) , to the modern consumer Internet companies like Google and eBay (Model 7).

In this article we will outline the key points of the seven models. As you will discover some companies operate in only one model; others have products that operate in multiple models. Below is a chart that characterizes each of the models. To add some concreteness to the examples, we’ll start from the perspective of the traditional enterprise software model.

Model One: Traditional

This is the traditional software business model, mastered by many companies, including Oracle and SAP. The software is licensed in perpetuity with a one-time price, in our example, of $4,000 per user. Subsequently, the customer is charged some percentage; say 20 percent, for the update/upgrade rights as well as maintenance, sometimes referred to as the software support fee. In our case study, this translates to about $65 per user per month, or around $800 per user per year. Many would point out that even traditional software companies have large recurring revenue or subscription businesses[4].

In Model One the business consumer then buys hardware, and hires or redeploys people to manage the software. It's here that the spending really goes up. The Gartner Group has estimated that the end user can spend up to four times the cost of their software license per year to manage these applications. This cost multiplier is not limited to complex business software. Microsoft has said , “the initial purchase is usually only 5% of the total cost of owning and maintaining a program.”[1].  Why does it cost so much?

Amy Mizorus Konary, at International Data Corporation (IDC), and one of the earliest industry analysts to follow the change to software delivered as a service, cites five major areas that cost companies millions of dollars annually. "CIOs and their departments really focus on five key aspects of the management and maintenance of their computer and software systems: availability, performance, security, problems, and change management." The implications of this are substantial. First, CIOs' budgets are dominated by spending on managing software, leaving little money on the table for new programs and projects. In our example, while the price of the software may be $4,000 per user, the customer will end up spending $1,300 per user per month to manage the security, availability, performance and upgrades to the software.

Model One has been a powerful business model. Webex founder Subrah Iyar lectured at Stanford in 2006, and said "I would much rather have built a software business on the traditional model—it's a much better business model for the software company."  It is particularly good since, with the large up-front purchase of the software, sales and marketing expense, which typically can run thirty to forty percent of revenue, can be funded. The license business will typically be break-even, and the profit will be realized from the software support business. Software support or maintenance typically consists of break-fix; access to usage information and the rights to the continued development of new software. Anyone who buys packaged software should appreciate they are not only buying what the software does today, but also the continued refinement and enhancement of that software for the future. It is, of course, in the software company's interest to move customers forward on new releases so as to not incur great expense in supporting multiple releases of the software. In the early stages of any software company, this is mutually beneficial, as the customer needs and wants future functionality and, in many ways, that's why the consumer chooses to buy versus build.

Model Two: Open Source

The free software movement began in the late 80s. By 1998, a group of programmers came up with open source as a replacement for free software, creating the Open Source Initiative (OSI) to promote the new term. Under the name "open source," an in-flux of for-profit dollars has changed the landscape of free software remarkably. Most major free software projects now have corporate backing of one form or another, and many of these companies have become quite adept at working with the community that participates in the projects, and the users who report bugs and suggest new features.

In May 2005, IBM made an acquisition of Gluecode Software, a privately held company specializing in integration software. The purchase price was estimated at $100 million. A year later, in June 2006, Red Hat, the world's best-known Linux distributor, announced the acquisition of JBoss for a reported $420 million. In early 2008 Sun Microsystems surprised everyone with their $1B acquisition of MySQL.

All of this has not been lost on the investment community. Over the past many years, over $500 million has been invested in a variety of open source software companies, including, but not limited to: MySQL, Greenplum, SugarCRM, Qlusters, XenSource, JasperSoft, Black Duck, Zimbra, ActiveGrid, SpikeSource, Astaro, Scali, SWsoft, Zend, GroundWork, Alfresco, JBoss, Collax, Univa, EnterpriseDB, Optaros, Trolltech, Funambol, Pentaho, OpenLogic, ActiveState, SIPphone, rPath, Centeris, Continuent, Hyperic, Sourcefire, WSO2, Palamida, MuleSource, Openbravo, Melodeo, and Compiere.

Why? The open source model (much like Google) has reduced the entry cost to “try it” to zero. MySQL estimates there have been over one hundred million installations and over 60,000 downloads per day. JasperSoft says there have been over five million downloads and over 80,000 corporate deployments in 200+ countries of their Jasper Reports. SugarCRM is now showing a Google map with the heartbeat from hundreds of computers in Europe running SugarCRM software—and SugarCRM has a minimal sales force in Europe.

While having people download, install, and potentially use your software is far better than having them download a brochure, open source software companies have struggled with the business model. Traditional support and maintenance, which has been the lifeblood of Model One software companies, has been difficult to monetize. Simply because with the source open to the community, the idea that you have to go back to the source of the software (the producer) for code fixes is no longer the case. Traditional bronze, silver, and gold support has also been difficult, since there has been little value-add other than changing the hours you will answer the phone and how quickly you’ll respond to a request. Some open source companies, like SugarCRM, are experimenting with providing both appliances and an on-demand solution as a way of providing a higher degree of service than the traditional software support. Some open source companies are depending on indemnification services. Some are providing some functionality in a paid-for subscription service not available in their free version. However, other than for Red Hat with their subscription business to Red Hat Network, revenue has been elusive, but with lots of money and smart people, it may not be for long.

Model Three: Outsourcing

Recognizing that the cost to manage software systems is much larger than the purchase price of the software, companies have emerged who are eager to take on the challenge. EDS and IBM Global Services have each built large, multi-billion dollar businesses outsourcing the management of the software and systems. As with Model One the software is still licensed on a perpetual basis for the $4,000 per user; support fees are still paid to the software company, but now Model Three companies take on the management of that software for less than the $1,300 per user per month the customer would be spending. Model Three companies manage the software either at the client’s site (@Client) or at their own locations (@Home). As much of the cost of the software management is buried in the cost of people, a new group of outsourcers, such as Infosys, have emerged, using offshore resources in India to lower the overall cost. Groups are now beginning to emerge in China and Eastern Europe engaging in the same business model.

Outsourcing can grow big companies and high value companies. Today, EDS is over $20 billion in revenue. IBM Global Services generates over 50 percent of IBM’s revenue, which, in 2007, was $50 billion. Tiny Infosys with over $4 billion in revenue, trades at more than five times revenue, a multiple most software companies would be glad to have. These companies continue to grow and take on higher value functions, including entire business processes. ADP is probably the granddaddy of this business model.

While this model is good, there are two key challenges. First, over time, massive differences in labor rates are not sustainable. Second, the dependence of system availability on people, with the inherent risk of human error, is also not sustainable. IBM has recognized that they could grow their Global Services business by a factor of two, bringing them to over $100 billion in revenue. While the opportunity is there, they would erode their margins. Fortunately, or unfortunately, they are already in the low cost countries – Brazil, Russia, India, and China—so unless there are highly skilled resources in Antarctica waiting to be discovered, IBM will need to invest in technology. Jim Spohrer has been chartered with leading an initiative inside IBM called Service Science. Spohrer has said, “One grand research challenge for us is how to scale up services and how to invest to get year-over-year improvements.” Computer science has benefited from Moore’s Law (where the capabilities of computer chips double about every 18 months). Could a Moore’s Law of service science be possible?[3]

Model Four: Hybrid

For most traditional software companies, Model Four represents a way for existing software companies to deliver their software as an on demand service. This business model leverages what an existing software company already has: products, distribution channel, and an installed base. In Model Four the traditional licensing model is retained. In our example, the software is still licensed for $4,000 per user; the support fee of 20 percent or $65 per user per month is also maintained. The big difference is to give the user the choice of having the software company service their own software. This is the model Oracle pursued beginning in 1999. Customers ranged from middle market to Fortune 500 companies and spanned the range of applications from financials to HR, manufacturing, and CRM and supply chain. From the consumer's perspective, the choice was between servicing the software themselves for around $1,300 per user per month, having a Model Three company provide the service for less than $1,300, or having Oracle service their own software for $150 per user per month. We'll come back to answer the question as to why was it so inexpensive, but first let's make sure you're clear on the model.

In Model Four, the purchase of the software and support remain the same. What's new is the customer is given an option to have the availability, security, performance, and changes managed for an additional fee. Customers were given the flexibility to customize and integrate the applications thru the CEMLI model. A CEMLI stands for Configurations, Extensions, Modification, Localization, and Integrations/Interfaces Framework and represents over 20+ classes of reasons why a customer might add software that Oracle did not author.  By being able to categorize and therefore assess a cost or price to managing these CEMLIs the issue of customizations became a business issue, not a technical or religions argument. This is of course particularly important as one deals with heavier weight business processes, e.g., process manufacturing vs. lightweight processes (we all do it the same way no matter what kind of business we're in) like email or web conferencing.

Model Four also allowed customers to choose when changes would be made to the application and when upgrades to major releases would occur. While there was a great deal of flexibility versus what you will find in either Model Six or Model Seven there was also a great degree of standardization. This started from the hardware and moved up to the software and finally to standardized upgrade, performance management or disaster recovery processes.

Which gets us back to — how was it possible to offer the service at significantly lower prices than in Model Three? And, could this business be operated at "product level" margins? The quick answer is, yes, product level margins could be achieved because of a high degree of standardization. Standardization allows for specialization—say, for example, performance management. Specialization also allows for repetition, the key to quality and skill in any area, whether managing computers or doing heart surgery. With repetition based on specialization, ultimately computers could be put to work automating the key business processes in the business of servicing software. A simple example is upgrades. Upgrades from one application release to another can be done hundreds of times in a year by specialized teams, and increasing parts of the upgrade process were automated. Automation ultimately results in both lower cost and higher quality service.

The advantage of Model Four is that the software does not have to be re-engineered. And, from a sales perspective, since the traditional license model is preserved, you can use your existing sales channel. The model is not cannibalistic. In fact, since the sales teams were compensated not only on the sale of the license but also the first year of the service, the average selling price nearly doubled for on-demand applications. In addition, the ability to offer a complete solution from one vendor was often a significant competitive advantage against traditional software companies who had to go to partners for the solution.

Finally, by having a model that split the software license from the service, the entire installed base of customers who had already bought the software was open for business. In fact, there were many situations where there were no traditional software sales persons assigned to customer accounts that had bought a lot of software, because they had bought all the software they could eat—an interesting paradox.

Of course, any model is not without its downsides. Alternate channels of software distribution do not welcome this model. They have often times made a living off of selling "boxes and bodies" and see the vendors' entry into this space as competitive. The second downside is, this model can only take you so far from a cost perspective. The cost, not price, of providing this service is in the neighborhood of $50 per user per month, which compares favorably with doing it yourself or having an outsourcer provide the service, which is probably closer to $500 per user per month. But if someone could provide 80 percent of the function of your product at $5 per user per month, then you would no longer stay in Model Four. This is of course the story of Siebel versus salesforce.com.

Model Five: Hybrid+

Many people see the on-demand model as a change in the way software is purchased—in particular, the elimination of the upfront license fee. While Model Four allows for the delivery model to change, it is easy to see you are not far from moving the business model as well. In our example, if one were to assume a 36-month time to pay off the $4,000 per user license fee, this would equate to $110 per user per month. Then adding the $65 per user per month for support and the $150 to manage the software results in an “all in” price of $325 per user per month. Of course, the revenue that is recognized in the first year is less than in Model One, but midstream in the second year this begins to change. By the third year, this model is resulting in much larger revenue stream than was true in the traditional model. The choice of moving from Model Four to Model Five is strictly a business choice, not a technology choice.

Callidus Software, a provider of sales performance management solutions, has adopted this business model. Rather than split the license from the service as we have seen in Model Four they are providing one price for the entire service. As in Model Four, they can provide much of the flexibility that customers have grown accustomed to, such as deciding when to upgrade the application. Of course, in order to maintain margins and provide a high quality, repeatable service, they are standardizing as much as possible.. As with Model Four, Model Five enables the software company to leverage their existing product line and their existing distribution channels and is a very suitable option for any existing software company.

Model Six: Software as a Service

As we discussed briefly in Model Four, the cost to manage the software can be driven down through standardization, repetition and automation, but there is a floor. Architecting for delivery of the software as a service, and only a service, promises to take out another order of magnitude of cost for both the software company and the user of the software. Most, if not all, of the new business software companies started in the late 90s have adopted this model. As a business model, even MMORPG (Massively Multi-player Online Role Playing Games) like World of Warcraft have adopted this model.

Software engineered for Model Six will have dramatically lower cost structures than software as a service offered by traditional software companies.  From 2000-2008 at least twelve business software companies, who all deliver their software as a service, and have all made public offerings. They include: WebEx, Salesforce.com, Concur, Taleo, Kenexa, Omniture, DealerTrack, RightNow Technologies, Vocus, Successfactors and Netsuite. This is compared to zero application software companies in Model One. Today, the combined revenues of this group is over $1.5B, and the market capitalization is over $15B. WebEx, the oldest company in the group, was acquired this year by Cisco for $3.2B, ten years after their founding.

Model Six is not without its challenges. While it is clearly a better model for the consumer of the software, one can debate whether it's a better model for the software company. Unfortunately, rather than being able to close million-dollar deals for software licenses in Model One, and then being able to plow that cash back into sales and marketing, companies in Model Six have to find a way to finance the early acquisition of customers. This has typically been done in the private and public markets. If you analyze all of these companies it's taken a lot of money  to establish the distribution channels required to achieve their meteoric growth. Furthermore, as in Model Four and Model Five the existing channels, are at best ambivalent to the Model Six players. With no easy distribution channel, many companies in Model Six reach $10 million in revenue and have no ability to grow larger.

Model Seven: Internet

We already introduced eBay as an example of the differences in cost structures of software as a service versus traditional software. Model Seven encompasses all of the companies in the consumer Internet. Amazon.com, Facebook, Google, Yahoo and eBay are all software companies.

Model Seven is distinctive in a few ways. First, these companies have chosen unique and often indirect ways to monetize their intellectual property, the software. Whether it's eBay charging by transaction, Google by the ad, or Amazon.com by embedding it in the sales transaction, each of the companies in Model Seven has adopted business models that are not simply a matter of counting the number of users. This is not to say there is no cost of sales and marketing. You only need to look at eBay's or Google's financials to realize that there is a healthy spend on sales and marketing. The primary difference is, the person paying for the software is not always the user of the software.

The second big difference is that each of these solutions is highly specialized. You don't buy books, auction cars or find your perfect mate on Google. Google, eBay, and Amazon.com have all been very focused on a single function. Why? Again, it comes to cost. In the traditional world, if I built some interesting software in Sydney, Australia, and I wanted to expand to Melbourne, I had to hire a sales/support engineer and a salesperson. My cost of distribution meant that I needed to extract the greatest degree of value out of each of my customers. That model resulted in the classic Model One software company "going horizontal"—meaning they began to offer more and more parts of the solution and the expected full integration. 

If Google on day one had tried to allow you to search, buy a book and auction a Pez dispenser the application would have been far too difficult to use. Instead specialization leads to greater simplicity and far lower cost. Back of the envelop arithmetic would argue Google's cost to deliver the service is around $0.50 per user per month.  There are other examples of highly specialized services such as the American Airline's SABRE system and arguably today most of the specialized applications that are delivered as a service are internal systems of major corporations.  As the world has moved to being a service business every company is in the information business.

As successful Model Seven companies have focused on delivering these services at a low cost point and at scale they have invested in many of the software management processes that are at the core of the cost of Models Four Five and Six. Anyone delivering software as an on demand service would be wise to learn from their experiences. As an example consider the eBay processes that allow them to make 10-20 feature changes every two weeks (our of 100s in development) and the level of technology they've developed to be able to manage these changes reliably and at scale[2].

Summary

While it may not be clear to you which model to choose, one thing is clear—there are choices to be made.. You can choose to sell Model One and Model Four, which is what Oracle did. You can choose to use Model One and Model Five, which is what Callidus Software did. You can choose to move completely to Model Six, which is what Concur did. For those of you who have not been a student of Concur you should be. They are the only company which has made this type of transition while a public company. 

If you're a startup it's clear you'll start out in Model Six, which is what Successfactors, Netsuite, Taleo and Constant Contact did. Finally, you can choose to acquire Model Six companies, and have some of your product line in Model One, some in Model Four and some in Model Six, which is what Blackbaud, did.

Whatever you choose to do you’ll need to think about how the choices affect the entire company; since unlike the change to client/server, this shift is much more than technology.  R&D, operations, sales, marketing, finance, product management and HR are all affected by the choices you make.