Creatures of Habit
The big events get all the press. The A-List actors make all the money. The biggest hits get all the radio play. The coolest gadgets get all the sales. We latch onto mainstream products because they make a big impact. These phenomena follow Long Tail distributions, where 20% of the causes lead to 80% of the effects. The Long Tail is hard at work in our IT departments too, governing the popularity of our technology functions. (A technology function is any unit of software that could be exposed as a service through SOA.) Each function carries a certain level of demand. Plotting the demand profiles of all IT functions and ranking them from most used to least used would produce a Long Tail graph similar to Figure 1, where the most used 20% account for roughly 80% of the total market for IT functions. These are IT’s star performers and are being tracked down and converted into services by SOA practitioners everywhere. The other 80% lurk in the shadows of IT’s portfolio and are generally overlooked as candidates for SOA. When it comes to deciding which business needs to focus on for SOA, we instinctively draw a clear line in the sand. We grab onto the most recognizable 20% and forget the rest. Why such divided interest?

We see opportunity in catering to the needs of the masses. We see risk in catering to the needs of the few. As SOA practitioners, our goal is to maximize the value of IT by instituting sweeping architectural change. Our job is to unwind decades’ worth of accidental architecture and convert IT from a necessary liability into a strategic asset, and do it all before our competitors do. No pressure. Naturally, when the pressure is on we focus on what we know, what we are most comfortable with. We are most comfortable building for maximum reuse.

The Mythology of Reuse
IT historically caters to the business masses. We own the ERP systems, the intranets, the corporate Web sites, the enterprise portals, the business intelligence platforms, the data warehouses, the e-mail infrastructure, the LDAP forests, the B2B feeds. We build million-dollar solutions and roll them out to hundreds, even thousands, of users. We go big because we have to. IT is the technology utility provider for the enterprise. Because our IT departments have limited resources, we must be selective about the projects we undertake. Traditionally, we select the ones with the biggest impact, the broadest reach, the highest reuse to satisfy as large a contingent of customers as possible. We can’t afford to occupy ourselves with providing obscure products in low demand any more than the corner convenience store can afford to stock every brand of toothpaste, cereal, and beer. IT has been in the business of providing mainstream products to a mainstream audience for decades. When it comes to SOA, we instinctively follow the same strategy. We strive for SOA solutions that offer the greatest value. Unfortunately, the way we typically valuate SOA solutions is flawed.

The most popular method of determining SOA value involves examining reuse. “Do that which maximizes service reuse” seems to be the mantra of the SOA industry today. Indeed, in my own experience, reusability has been the primary, if not exclusive, yardstick with which SOA success is measured. SOA success seems to be synonymous with high service reuse. This philosophy implies that reusable services are more valuable to the business than non-reusable ones, and bases ROI on the number of consumers a service has. More consumers means higher reuse means greater value. In reality, the link between service reuse and business value is a myth.

We tend to believe that services will be reused if designed well, but we have little direct control over how reusable a service is. Sure, we can take care to design our services in ways that let them be leveraged by more than one consumer, but this just ensures that the service is capable of being reused, not that it will be reused. Reuse is really a function of demand. Naturally, the higher the demand the higher a service’s reuse potential. The point is that a service is not reusable because it was designed a certain way. Rather, a service is reusable because it exposes a business function that is already in high demand.

We also have a tendency to believe that reuse leads to business alignment and agility. Service-enabling IT’s most widely used capabilities should bring IT and business closer together and change-proof most of the enterprise, right? Actually, mainstream services benefit IT more than they benefit the business. IT is likely to respond more quickly in times of change, but the benefits of that agility seldom extend out into the lines of business. Suppose, for example, that an IT shop anticipates replacing an antiquated ERP system. It prepares by abstracting its widely used reporting functions through service interfaces and retooling existing business intelligence systems to use those services. Then at some point in the future executive management sends down the order to swap in a new system. If everything goes well, users of the BI tools (and the tools themselves) will have no knowledge of the swap. The transition will be seamless and transparent to consumers of ERP reporting functions. The success would showcase IT’s agility and alignment with corporate directives, but wouldn’t benefit lines of business directly. There’s a difference between providing services that improve business practices and providing services that prevent disruption to business practices. Service-enabling mainstream functions protect lines of business from change (an IT virtue) but don’t make lines of business more competitive (a business virtue).

Mainstream service opportunities are also relatively rare. They are extremely important, to be sure, but they are far outnumbered by niche opportunities. To provide only highly reusable services is to turn a cold shoulder to 80% of the company’s business needs. Who will be there when those needs change? What’s more, niche services benefit business users directly since it’s usually line-of-business personnel who request them. Niche services solve very focused business needs and, by definition, aren’t very reusable; however, niche needs exist en masse and come directly from the trenches – and that’s where business battles are won and lost.

As long as the goal of SOA is to improve business agility, the needs of our companies’ niche markets shouldn’t be ignored. Focusing on popular solutions and maximum service reuse is a natural IT response to the challenge of building enterprise agility. But an SOA built on the premise of maximizing service reuse focuses too much energy on internal, corporate needs. To deliver on the true promise of SOA, our strategies must also push specialized services out to the very edges of our businesses where they can be leveraged when needed. Services do not have to be reusable to be valuable. They just need to be available. This is Long Tail SOA.

Mainstream Versus Long Tail SOA
Reuse is only one factor in the SOA value equation. Now that the myths about service reuse have been revealed we can open our minds to other considerations — namely demand for services and the nature of change. These forces influence SOA implementation and can act for or against SOA goals depending on the strategy. What may weaken a mainstream approach may strengthen a Long Tail approach, and vice versa. Remembering that mainstream SOA focuses on providing popular services for a few large business “markets” and Long Tail SOA on providing specialized services for many small markets; let’s examine how each approach is affected by service demand and change.