Saturday, February 07, 2009

Delivering software as a service

Delivering software as a service

A new delivery method is shaking the software industry’s foundations. Traditional vendors should take heed.

The online delivery of software—sometimes labeled software as a service—has been a long-standing dream for some vendors and CIOs. The concept is simple and attractive: rather than buying a software license for an application such as enterprise resource planning (ERP) or customer relationship management (CRM) and installing this software on individual machines, a business signs up to use the application hosted by the company that develops and sells the software, giving the buyer more flexibility to switch vendors and perhaps fewer headaches in maintaining the software. For many years, traditional software vendors (those who sell licensed and packaged software, often along with a maintenance contract) have been able to overlook a rising crop of competitors that offer software as a service, as the latter have struggled to develop truly competitive services. It’s now time for traditional companies to pay attention, for they risk losing their privileged position to attackers that offer applications in this new way.

The complacence of traditional vendors is easily understood in light of the record: the first generation of online software delivery, in the late 1990s, failed to meet the reliability and quality standards demanded by business users. But the new delivery method appears to be taking off. While it won’t replace existing licenses and in-house custom-developed code overnight, an IDC report1 projects that 10 percent of the market for enterprise software will migrate to a pure software-as-a-service model by 2009. Our analyses suggest that software as a service is a growing priority for CIOs and venture capital investors.

Our review of venture capital investments shows that companies whose main business is delivering software as a service saw their revenues rise from $295 million in 2002 to $485 million in 2005, an 18 percent increase. On the buyer’s side, our fall 2006 survey of senior IT executives indicated a dramatic jump in the number of companies considering software-as-a-service applications during 2007.2

With software as a service, a customer contracts to use an application, such as ERP or CRM, hosted by a third party, rather than buying a software license and installing the application on its own machines. Just as consumers can check e-mail or use mapping programs with their Web browsers, so too can enterprise customers access business applications over the Internet.

Several factors are spurring the growth. New software design and delivery models allow many more instances of an application to run at once in a common environment, so providers can now share one application cost effectively across hundreds of companies—a vast improvement on the old client-server model. Bandwidth costs continue to drop, making it affordable for companies to purchase the level of connectivity that allows online applications to perform gracefully. Perhaps most important, many customers are eager for the shift, as they’re frustrated by the traditional cycle of buying a software license, paying for a maintenance contract, and then having to go through time-consuming and expensive upgrades. Many customers believe they would have more control over the relationship if they simply paid monthly fees that could be switched to another vendor if the first failed to perform (see sidebar “How CIOs can get maximum value from software as a service”). And finally, the successes of early leaders, such as and WebEx, have demonstrated the viability and value proposition of this model.

Market indicators suggest that investors share the enthusiasm of vendors and CIOs. Our index of companies whose main business is software delivered as a service3 outperformed the overall software company index (excluding Microsoft) by more than 13 percent from January 2002 to December 2006.

Although software-as-a-service vendors are less profitable than some traditional software vendors today, this gap is primarily caused by a lack of scale. We expect the economics of online delivery to change as the model gains wider acceptance. Large software companies (excluding Microsoft) typically have operating margins of around 25 percent. However, the margins of companies with revenues below $1.2 billion a year hover around 14 percent—close to the 13 percent margin of the average software-as-a-service vendor (Exhibit 1). A few service vendors already have much higher margins (WebEx, at 26 percent, and Digital Insight, at 19 percent) because they’ve been able to achieve scale and a leading position in their niches. Other leaders, such as (which provides on-demand CRM and sales force automation tools) and ADP (the world’s largest automated check processor) have also gained mainstream uptake among midsize and large companies.

Traditional software vendors across many industries, we believe, will find their privileged position threatened unless they move aggressively to serve their customers by making software applications available online. To be successful, vendors will have to understand the rules of the new game, especially how customer service and sales models differ, and adjust accordingly. They’ll need to grasp why software is moving to this model and how the new economics affect customers, intermediaries, and providers—not only software developers, but also IT and business process outsourcing providers. They will have to make changes in their own organizations by ramping up their ability to deliver software from large data centers and by developing new ways to sell to and service their customers. The sooner software companies embrace the change, the better access they’ll have to top talent and new markets and the better off they are likely to be in the long term.

Emerging potential

The first wave of adoption for software as a service has been under way for several years. Companies are eager to acquire the technology for human-resources applications such as CRM and payroll and for collaboration tools that aren’t mission critical, involve relatively low data security and privacy concerns, have a distributed user base, and require little integration with on-premise applications and little customization. In addition, several industry-specific applications are gaining popularity in large industries such as financial services, health care, and retail, as well as in smaller ones such as automotive retailing, law, and real estate.

The next wave of applications seems likely to involve transactions between buyers and suppliers, including procurement, logistics, and supply chain management. As customers grow increasingly comfortable with the concept, a third wave of applications more critical to business, such as hosted environments for software development, may also start to be delivered in this way. Enterprise customers and those in small and midsize businesses are likely to adopt applications at different rates. For example, some smaller businesses are already embracing the online delivery of financial applications, whereas large enterprises may never do so because of the breadth and depth of functionality required or because of concerns about data security (Exhibit 2).

All three waves mostly aim to replicate the functionality of applications that have been sold as packaged software and hosted on the customer’s site. The next frontier—we might call it software as a service 2.0—will include new classes of applications which are actually better suited for online delivery and seamlessly integrate with on-premise applications. Consider spam- or virus-protection applications, which are superior to e-mail filters because they stop junk e-mail or harmful viruses before they enter a company’s fire wall. Other well-suited applications are those that perform some kind of data reconciliation, like software that reconciles differing SKUs between suppliers and retailers. Such products will increasingly rely on repositories in “the cloud.”

What software vendors must do

Software developers that shift from a traditional licensed model to software as a service will need to work hard to retain existing customers. Companies that have purchased long-term contracts for updates and maintenance, for example, will want guarantees of favorable subscription pricing under the new model. Others may be so comfortable with their current setup that they will switch only if the software vendor makes it painless, perhaps even free. Additional issues to consider during the transition from packaged software to software as a service include R&D and customer support—not all customers will switch at the same moment, so vendors may need to run dual operations in these parts of the business.

Under the new model, the economics of the software and IT services industries will also change. Vendors will have to adapt their financial and operating models or risk losing their privileged positions (see sidebar “The threat to BPO vendors”).

Adjust the revenue model

Software developers that deliver software as a service experience higher sales and marketing costs relative to earnings than traditional vendors do. There are at least two reasons. First, a subscription model for software produces lower revenues during the growth phase, since payments are spread over a period rather than made immediately in a one-off license sale. Sales expenses in both models are expensed as incurred, however, leading to a higher ratio of costs to earnings for the service model. Second, a higher percentage of a service vendor’s customers are small and midsize businesses: more effort is required to reach them compared with the large-enterprise customers that make up more of the customer base of traditional software companies. Over time, however, as subscription revenue streams become steadier, the service model may look better, since the incremental marketing and operating costs of adding additional subscribers and of offering new services and applications for existing customers are minimal.

The cost structure for online software as a service differs in other ways too. For example, software-as-a-service vendors have lower R&D costs than traditional software companies do because the former don’t have to develop and maintain multiple versions of a product to run on different platforms. While they tend to have lower customer support costs (partly because they often limit customer service to self-help options rather than telephone support) and do not incur any product distribution costs (such as CD creation, packaging, and manuals), software-as-a-service vendors have higher costs for delivery than their traditional counterparts do because of the costs associated with hosting, bringing customers and users onboard, and managing the application and data center environments. Traditional software companies may also depend more on revenues from consulting services, which can often amount to two to four times the license fees. Consulting opportunities appear to be fewer in the software-as-a-service realm, where implementation is less challenging.

Build a platform

As in other radical shifts that have affected the software industry over the past two decades (PCs, client-server computing, and the Internet, for example), players who respond quickly to emerging trends will often be best placed. A first step, as noted earlier, is to understand which applications will migrate when and to position one’s own offerings early in the uptake. For example, Siebel lost precious time by initially dismissing the idea of software as a service, allowing to steal a lead in the online delivery of CRM applications. Tracking the receptiveness of customers to the service delivery model is critical, so vendors will need to build organizational capabilities to monitor and shape demand.

For traditional vendors, a broad platform that allows them to supplement and enhance their applications is essential, as is deciding which platform to join. Most large software companies are adept at building these developer platforms in the licensed model (Microsoft’s Windows platform being the best known). Platforms for software as a service are similar in that they require vendors to establish and evangelize a set of application-programming interfaces (APIs) and standards for data exchange among applications. But platform partners in the software-as-a-service model can also build onto the back-end infrastructure of applications, where the billing, metering, provisioning, and advertising functions may reside. This approach can reduce development costs for partners, but it also may increase those costs for the platform provider.

A second key difference between the models is that version upgrades tend to occur more often and in smaller chunks with software as a service than with the traditional approach. A platform vendor typically releases a major revision once every three to five years, and the company and its partners go to great lengths to make sure that the new version works with older machines and applications and that any upgrades from partners don’t break the platform. The smaller but more frequent upgrades in the software-as-a-service model allow platform providers and partners to ensure compatibility continually with an ongoing investment in smaller increments.

In the migration to the online delivery model, partnerships are likely to change. Some competitors may become partners if their interests align on a common platform. Just as likely, some intermediaries (especially resellers and systems integrators) may find themselves cut out of a direct relationship with the developer and the customer unless they can find a way to add value. The compensation structure, for both internal sales and channel partners, will need to change. Specifically, commissions will have to be based on ongoing customer usage and revenue rather than on the sale of large up-front licenses.

Improve internal capabilities

In the new delivery model, the sales and marketing function should follow the lead of those software-as-a-service vendors that are targeting the business side of the customer organization. One reason for this emphasis is that many of the early applications for online delivery—sales force automation, human resources, CRM—are pure business applications. Once they have demonstrated their value, IT managers may be more willing to adopt this mode for applications within the IT realm, including e-mail, security, and storage. Even CIOs who support the concept of online delivery may be concerned about the security and reliability of this new approach and so may hesitate to change if the current system isn’t broken.

The biggest capability gap for software companies embracing the new model is in the operational and customer service skills necessary to deliver software online. The operational challenge is to host the software rather than shrink-wrap and ship it. Companies will have to develop capabilities to handle massive data center operations, systems and network monitoring, and billing.

Software vendors will also need to change their attitudes toward customer service. Complaints frequently involve incidents that are mission critical, thus demanding immediate attention. Systems outages and connectivity problems can affect all customers at once, implying large fluctuations in incident volumes, so first-line user support must be able to handle unpredictable spikes in inquiries. In our experience, moreover, software-as-a-service customers rely heavily on online tools, knowledge bases, and forums and as such represent a savvy and demanding user group.

R&D capabilities will also need to change. Smaller but more frequent online upgrades will demand new attitudes on deadlines and quality. While deadlines for upgrades can be flexible, the online release of a new version to the entire user base will have to be bug free. Also, the R&D department will need to expand its focus from the product and its features to the entire delivery model, including tools for serving the software, billing the customer, and aggregating different software services. Last but not least, to ensure a smooth end-user experience during releases R&D units must develop a customer service mind-set and collaborate with those parts of the company that manage sales and delivery. This shift will require new processes, such as joint release planning, that resemble the way application-development and infrastructure groups operate in traditional IT organizations.

With most software vendors planning to offer their wares online in the next few years, a time of rapid industry change is approaching. Software industry executives will need to gauge the receptiveness of their customers to this shift and work out what services to offer and to which customer segments. They’ll need to build up their internal capabilities for service delivery and to manage major shifts in R&D, business models, partnerships, and compensation models for sales teams and channel partners. They’ll also need to monitor the rise of online platforms—and not just from established leaders (such as IBM, Microsoft, Oracle, and SAP) but also from pure plays like’s AppExchange. And finally, they’ll have to build the management processes and organizational structures to manage two distinct but potentially mutually cannibalizing businesses (traditional software and software as a service) with different business models and requiring remarkably different sets of capabilities. Although this challenge is formidable, software executives who continue to put it off risk being left behind. 


About the Authors

Abhijit Dubey is an associate principal in McKinsey’s San Francisco office, and Dilip Wagle is a principal in the Seattle office.

The authors wish to acknowledge the contributions of Aparna Bhatnagar and Chandrasekar V to the underlying assertions of this article.


1 Worldwide and US Software as a Service 2005–2009 Forecast and Analysis: Adoption for the Alternative Delivery Model Continues, IDC, March 2005.

2 Janaki Akella, Kishore Kanakamedala, and Roger P. Roberts, “What’s on CIO agendas in 2007: A McKinsey Survey,” The McKinsey Quarterly, Web exclusive, January 2007.

3 The index included the performance of the following companies from January 2002 through December 2006: Concur Technologies, Digital Insight, Digital River, RightNow Technologies,, Taleo, Ultimate Software, WebEx, WebSideStory, and Workstream.


Google’s view on the future of business: An interview with CEO Eric Schmidt

Google’s view on the future of business: An interview with CEO Eric Schmidt

How the Internet will change the nature of competition, innovation, and company operations.

Few would dispute that Google sits at the center of the Internet. As the leader in search, Google is now the Internet’s premier brand and the planet’s most potent free service. Managing that commanding position falls largely to seasoned technology executive Eric Schmidt, who in 2001 was tapped for the CEO post by Google founders Sergey Brin and Larry Page.

In his years at the company, Schmidt has delivered steady growth while expanding Google’s reach. By anticipating the ways in which people would expand their use of Internet applications, Schmidt has introduced new products from the popular Web-based e-mail service Gmail (Google Mail in Germany and the UK) to the recently unveiled G1 mobile phone. And as Google’s audience and influence have increased, so too has its appeal among advertisers worldwide.

Making all this happen depends on Google’s ability to attract and engage top talent. The organization that Schmidt has helped shape depends on collaborative projects and free flows of information that encourage employees to share ideas. Staffers devote 20 percent of their work time to special projects of their own design, an inventive and effective policy that is at the core of its innovation efforts.

Not many executives have a better vantage point on the changing technological landscape than Schmidt. He recently took time out to discuss his views with McKinsey director James Manyika. Schmidt sees more powerful digital assistants arising from cloud computing, markets morphing at an ever faster pace, and plenty of space for human creativity if organizations are willing to carve out a place for it.

The Quarterly: The Internet has radically changed the world. What are the kinds of developments you see ahead?

Eric Schmidt: When people have infinitely powerful personal devices, connected to infinitely fast networks and servers with lots and lots of content, what will they do? There will be a new kind of application and it will be personal. It will run on the equivalent of your mobile phone. It will know where you are via GPS, and you will use it as your personal and social assistant. It will know who your friends are and when they show up near you. It will remind you of their birthdays. It will entertain you. It will warn you of impending threats and it will keep you up to date. It will use all of that computing power that’s in the cloud, as we call it.

So, for example, when you go to the store this device helps you decide what to buy at the best price with the best delivery. When you go to school it will help you learn, since this device knows far more than you ever will. So this vision of nearly infinite computing power, network power, and these powerful devices is the basis of the next generation of computing.

The Quarterly: Armed with all this technology, what happens to how people live and work in the world?

Eric Schmidt: There's such an explosion of content, and yet there’s so little understanding of it. So, I think the gap between what computers do—which is very high volume analytical and replication work, and the things that humans can do, which are essentially insightful—is a large gap. In our lifetimes, we will not see that gap close very much. Corporations will change the way they sell products to people who are increasingly computer assisted. But ultimately, we still run the world.

The harsh message is that everything will happen much faster. Every product cycle, every information cycle, every bubble, will happen faster, because of network effects, where everybody is connected and talking to each other. So there's every reason to believe that those who are really stressed out by the rate of change now will be even more stressed out.

However, there's a new generation who are growing up with this as the normal pace of their lives. They will develop the social norms. As leaders they'll figure out how they want to organize their world, when you and I are sitting around watching them from our retirement.

The Quarterly: Will the Internet bring down barriers, making markets more democratic?

Eric Schmidt: I would like to tell you that the Internet has created such a level playing field that the long tail1 is absolutely the place to be—that there's so much differentiation, there’s so much diversity, so many new voices. Unfortunately, that’s not the case. What really happens is something called a power law, with the property that a small number of things are very highly concentrated and most other things have relatively little volume. Virtually all of the new network markets follow this law.

So, while the tail is very interesting, the vast majority of revenue remains in the head. And this is a lesson that businesses have to learn. While you can have a long tail strategy, you better have a head, because that's where all the revenue is.

And, in fact, it's probable that the Internet will lead to larger blockbusters and more concentration of brands. Which, again, doesn’t make sense to most people, because it’s a larger distribution medium. But when you get everybody together they still like to have one superstar. It's no longer a US superstar, it's a global superstar. So that means global brands, global businesses, global sports figures, global celebrities, global scandals, global politicians.

So, we love the long tail, but we make most of our revenue in the head, because of the math of the power law. And you need both, by the way. You need the head and the tail to make the model work.

The Quarterly: So how do companies make money in these markets?

Eric Schmidt: Free is a better price than cheap. And this simple principle has been lost on many a business person. There are business models that involve free with adjacent revenue sources. And, in fact, free is a viable model with branding [advantages], [charges for] service, and other things. But it’s a different business model from what most of us are used to.

A rule of economics is that for manufacturing and mature businesses, eventually the price of the good goes to the marginal cost of its production and distribution. Well, in the digital world, for digital goods, the marginal cost of distribution and manufacture is effectively zero or near zero. So, certainly, for that category of goods, it's reasonable to expect that the free model with ancillary branding and revenue opportunities is probably a very good thing.

The Quarterly: What’s an example of how an industry might adapt to these changes?

Eric Schmidt: Obviously, for things which have some physical cost of production, you'll be losing money in a million units at a time unless you come up with some offsetting revenue. Telephony is a classic example. Most of the costs for telephony physical infrastructure are sunk costs. The cost of operating is not that great—mostly billing and so forth. So you could imagine a situation where telephony went from being billed by the minute to being billed by the purchase of the phone. You buy the phone, and covered in the cost of the phone is a part of that infrastructure. And then you could use the phone forever.

People have to accept that, at least in the digital world, the cost of transmission and distribution, is not going to go up. It's on its way down. The people who build physical devices that connect to [transmission and distribution] will eventually morph their models into more of the prepay model, because it will be more consumer efficient.

The Quarterly: What kinds of management changes are needed to cope with all this?

Eric Schmidt: The old saying is no one knows you're a dog on the Internet, that you can't tell the size of the organization, and so suddenly the Internet levels playing fields in many ways—distribution, branding, money, and access.

But it has a lot of other implications for the way corporations operate. They can't be as controlling. They have to let information out. They have to listen to customers, because customers are talking to them. And if they don't, their competitor will. So there's a long list of reasons why a more transparent company is a better organization.

There are many business models predicated on control. My favorite example is movie distribution windows. As a consumer, I want to watch the movie whenever I want, and on whatever medium I want. But the whole economic structure of the movie business, up until recently, was organized around distribution in a certain format, at a certain price, and then wait a while. But in the new world people won’t wait. A good example was the delay of the Harry Potter movie. The fans were fanatical, writing letters and calling private cell phones to overturn the delay. The industry has a fan base that they need to spend time thinking about.

There's a lot of evidence that groups make better decisions than individuals. Especially when the groups are selected to be among the smartest and most interesting people. The wisdom of crowds argument is that you can operate a company by consensus, which is, indeed, how Google operates.

The Quarterly: So, how do you do it?

Eric Schmidt: You need two things. You have to have somebody who enforces a deadline. In a corporation the role of a leader is often not to force the outcome, but to force execution. Literally, by having a deadline. Either by having a real crisis or creating a crisis. And a good managerial strategy is “let's create a crisis this week to get everybody through this knot hole.”

And the second thing is that you have to have dissent. If you don't have dissent then you have a king. And the new model of governance is very much counter to that. What I try to do in meetings is to find the people who have not spoken, who often are the ones who are afraid to speak out, but have a dissenting opinion. I get them to say what they really think and that promotes discussion, and the right thing happens. So open models, beyond input from outside, also have to be inside the corporation.

Encouraging this is an art, not a science. Because in traditional companies, the big offices, the corner offices, the regal bathrooms, and everybody dressed up in suits cause people to be afraid to speak out. But the best ideas typically don't come from executives. And, unfortunately, the executives don't agree with me on that.

The Quarterly: What types of people will succeed in this environment?

Eric Schmidt: I would venture to say that [in some organizations] many people still show up at 9:00 and leave at 5:30 with a half an hour break. And their productivity is defined by how much they can get done in the eight hours. And I would venture today that things haven’t changed very much in 50 years. A counter example would be people who are trying to make money in the blogosphere. Bloggers, for example, can't sleep. It's so competitive that if they go to sleep, and a story breaks all the clicks from their blog go to the others. So bloggers end up sort of disasters on a human scale because they can't deal with the fact that they can't even sleep, it's such a tense environment.

There's a spectrum in between, and, as an executive you have to think about where you are in that spectrum—both personally and as an organization. For senior executives, it's probably the case that balance is no longer possible. I would love to have balance in my life except that the world is a global stage and, when I'm sleeping, there's a crisis in some country, and I still haven’t figured out how not to sleep. So the fact is that you're going to select executives who like the rush of the intensity. They're going to be drawn to the sense of a crisis. The sense of speed. And they are the ones, ultimately, who are going to rise to the top.

The Quarterly: Could you tell us about how Google innovates?

Eric Schmidt: Innovation always has been driven by a person or a small team that has the luxury of thinking of a new idea and pursuing it. There are no counter examples. It was true 100 years ago and it'll be true for the next 100 years. Innovation is something that comes when you're not under the gun. So it's important that, even if you don't have balance in your life, you have some time for reflection. So that you could say, "Well, maybe I'm not working on the right thing." Or, "maybe I should have this new idea." The creative parts of one's mind are not on schedule.

So, in our case, we try to encourage [innovation] with things like 20 percent time, and the small technology teams, which are undirected. We try to encourage real thinking out of the box. We also try to look for small companies that we can acquire. Because, often, it's small companies which have the great new ideas. They have gotten started with them but can't really complete them.

Google's objective is to be a systematic innovator at scale. Scale means more than one. And innovator means things which really cause you to go, "Wow." And systematic means that we can systemize the approach—we can actually get our groups to innovate. We don't necessarily know this month which one [will succeed]. But we know it's portfolio theory. We have enough groups that a few [innovations] will pop up. And, of course, we also cull the ones that are not very successful. We push them to try to do something different, or retarget—or, in fact, get canceled. Although that's relatively rare.

The Quarterly: Is there a type of organization that has an edge when it comes to fostering innovation?

Eric Schmidt: Executives always want to simplify their lives. So they have three divisions, and four products, and all the marketing and so on. That may work in some businesses. But, for most businesses, due to the nature of technology, they’re going to become more complex. They will have more products and more variance. And it's part of [maintaining a] competitive barrier to have resilient, scalable, differentiated, and global products.

Which means they can't be built by two people anymore. So, in our case, while we recognize that innovation comes from small teams and we organize that way, we also encourage them to talk to each other.

One of the things that we've tried very hard to avoid at Google is the sort of divisional structure and the business unit structure that prevents collaboration across units. It’s difficult. So, I understand why people want to build business units, and have their presidents. But by doing that you cut down the informal ties that, in an open culture, drive so much collaboration. If people in the organization understand the values of the company, they should be able to self organize to work on the most interesting problems. And if they haven't, or are not able to do that, you haven't talked to them about what's important. You haven't built a shared value culture.

The Quarterly: What are some of dangers you see as the Internet continues to develop?

Eric Schmidt: There are a number of initiatives to try to build essentially a global standard for the Net. Given the history of war and global politics, it's highly unlikely that we'll see a single regime, for example, for copyright law, or for what content is appropriate, or for the penalties for inappropriate content, or for all of the issues that people face in the online world. Today, the way we solve this problem is to use per-country domains. So a domain that's per country is seen as different, such as the US domain, which is called dot com.

There are likely to be legal and political challenges to this over the years, and I think the next one will come soon. The Internet has withstood them so far. The most important thing in these situations is to have a large number of lawyers. The reason is that the laws have become so complicated that, to operate globally, every large corporation I know of has to have a lawyer who understands Brazilian law, one who understands Turkish law, and one who understands the European court. In the case of information, and in particular cultural information, there are widespread differences in what's legal and what's not. The Internet [response is] that people are subject to the local laws.

It would be a tragedy if the Internet, because of these issues, became balkanized at a physical level. It would be a tragedy if every country built, sort of, a police state around its Internet. It's much, much better to use other approaches to make sure that what's legal in one country and illegal in another country does not go from the legal country to the illegal country without appropriate supervision. 

About the Author

James Manyika is a director in McKinsey’s San Francisco office.


1A phrase coined by Chris Anderson, editor-in-chief of Wired magazine, in an October 2004 article of the same name. The theory of the long tail states that as the costs of production and distribution online fall, niche products and services can be as economic as mainstream ones.