Marc Andreessen recently wrote a useful post about web platforms. It’s a timely collection of thoughts that has spurred conversation amongst executives at several large technology companies, including Yahoo and Google, according to my sources. There’s a lot of ambiguity around this whole issue of platforms so I’m glad that Marc has taken the lead in throwing down a detailed perspective.

Part of the ambiguity stems from the fact that the word is totally hackneyed and over-played. What everyone seems to agree on is that in the distant future there will only be a few big IT companies – all of them with platforms at the core. Here we mean to say that current borders between layers of the technology stack – network, data center, web services, applications, etc. will fuse together to form platforms that challenge traditional lines. Understanding the nature of these platform opportunities, therefore, becomes an issue of survival. The issue has reach beyond traditional web companies, which are the focus of Marc’s ideas, affecting a spectrum of companies ranging from Cisco to IBM to Google.

There’s a lot to parse through in The Three Types of Platforms You Can Meet on the Internet. You might read it first before reading some of my commentary. I’ll focus on the main point, which is the introduction of a new idea – the Level 3 platform.

First some review. Marc defines Level 1 platforms as those available via “access API’s” along the lines of Adsense or PayPal syndicated on a third-party site. Level 2 platforms, by contrast, involve deeper integration along the lines of Firefox extensions or Facebook apps. I think he’s pointing to the deeper level of integration as a way to create better, richer user experiences. Good point.

Another great point is the retrospective on AOL – Marc says the company had a chance to do a “Facebook” when it was driving the online world but lost that chance by choosing to remain a closed system. That drove developers and third-party content providers to the web. In a different scenario, maybe they would have standardized around the AOL environment if the company had made access to its user base and network open and accessible. Quite possibly, we would have experienced a very different web and AOL would have remained as central to the medium as Google is today. But that’s just an interesting aside to consider as Marc explores what it means to be a web platform.

While I appreciate the logical difference between Level and 2 platform, I’m not sure juxtaposing them tells us much. Another way to characterize the difference between Level 1 and 2 is to think about inbound and outbound platforms and integration. Whether a platform delivers or receives services is really a matter of context. It’s very specific to the application and I’m not sure we can attribute causality between success/growth and the choice of Level 1 or 2 positioning. It just depends on what makes the most sense. PayPal and Adsense are necessarily Level 1 platforms because they’re not destination sites and make sense as syndicated services. Facebook makes sense as a Level 2 because it’s a destination that can be enhanced by letting third-parties in the door. You can’t just pick to one strategy over the other because it stems from where you play in the value chain.

Let’s move on to the main point, that a new type of platform (Level 3) would provide deeper integration between 3^rd party apps and the IT stack of the provider controlling the end-user experience. By IT stack we mean the plumbing – security, application performance, network performance, etc. To clarify with an example – imagine Facebook saying to the companies making apps for its platform: “just worry about the business logic – the code that is specific to your application – we’ll bring the rest of the IT stack to make your app perform, scale, optimize.” Right now, Facebook provides access to its users but obviously does not integrate with the plumbing of the third-party.

It’s a big vision and perhaps a representative forecast of how the platform business will look in the future. But it seems sub-optimal for an applications (web) company to bite off the entire scope of what is described as a Level 3 play. The better alternative for a web company like Facebook may be to address in stages or collaborate with specialized infrastructure providers to preserve focus and flexibility. In fact, that’s probably a separate opportunity all together.

Don’t get me wrong. I do believe in IT stack coalescence, especially at the lower levels. Another way to describe this is Paas or Naas (the Network as a Service or Platform as a Service). Paas or Naas companies would provide the IT environment for applications businesses. These are humongous market opportunities. The most well positioned companies are Cisco, IBM, Google and maybe even Amazon. They will work with application companies large and small. As I pointed out, some of them may even have sizable application businesses via separate divisions. This is clearly where Google is headed. I wouldn’t be surprised if Amazon’s agile management positions the company for this state of the future as well. Whereas Paas or Naas companies will compete on economies of scale, application companies compete on other dimensions. In the case of Facebook, it’s network effects which are the basis of its platform. In other cases, it’s brand or just-plain traction. Keep in mind, not all dimensions are alike – networks effects are always preferred though not always native to a business model.

Back to the Level 3 platform….there’s an implicit assumption that integration with lower parts of this IT stack is a strategic consideration that can make Level 3 platform more effective positioning than Level 2 platforms (which don’t integrate at these levels). The proposed idea seems to force application and IT platforms (two separate things which I’ll speak about below) into one view but most trends suggest declining dependencies between the two (“loose-coupling” of these layers). It feels risky to focus on two platforms during a market entry stage/process.

I think the operational models/alternatives need to reconcile with (3) big picture considerations:

=> There’s not yet significant evidence that the extra value-add of Level 3 over Level 2 is a pain-point for developers (and therefore, a pass-through issue for end-users). I recognize this market is still early, so there will be examples of performance and scalability bottlenecks that impact user experiences as suggested in the post. That said, isn’t the real constraint for developers conceptualizing and designing useful products that gain traction with consumers as opposed to the ability to secure reliable and cheap IT services to deliver these products? Specialized companies are working hard to make the infrastructure a commodity business.

=> The circumstances under which deeper integration matter are 2^nd order factors (meaning if the platform provider cannot bring 1^st order value to the table — customer footprint, large user networks, marketplaces, robust social graph – there’s no new/additional problem to solve). In other words, without addressing other issues first, the additional contribution of Level 3 capabilities don’t influence a company’s ability to compete/win in the market. In the case of social media, which is still early, there’s a lot more value to create for developers by bringing (for example) a composite profile to the table than seamless infrastructure and application delivery services.

=> There are several approaches/solutions to the unique value-add of Level 3 over Level 2 that seem amenable to specialized efforts by the likes of IBM, Amazon and others on the infrastructure side. These best-of-breed approaches to “IT infrastructure services” might rival the attempt of many application providers (e.g., Facebook) to do the same; therefore, concentrating effort on the aspects of the platform specific to social media may be more prudent use of organizational resources. Once market leadership is established, “infrastructure services” might be a nice up-sell possibility to the installed base of developers. Note: Salesforce.com followed this progression. They are now offering PaaS [Platform as a Service] but that’s years after they established the basic “add-in” model. Here they use “platform” to mean IT platform.

It’s important to clarify some of the terms being used in the post. To put some more descriptive concepts into the mix:

=>> Application Platform: Development environment that enables 3^rd party services to be integrated within context of an end-user application

=>> IT Platform: Run-time environment for managed application delivery by a source other than the developer

As the post explains, Level 2 is a sub-set of Level 3 where:

* Level 2 = Application Platform

* Level 3 = Application Platform + IT Platform

The question is whether moving to Level 3 is a 1 + 1 = 3 situation. The answer maybe yes but there several ways of getting to the same end-point. On the one hand you might have a situation like Photobucket into MySpace, where a 3^rd party app is central to the user experience and worth integrating into the core (hence the acquisition). Bringing Photobucket in-house has certainly reduced service-level risk. On the other hand, there are specialized efforts by IBM, Akamai and others to reduce the operational risk inherent in these types of collaborations. Absent functional dependencies between the application and IT layers, the job of connecting these layers at desired levels of performance and service-levels becomes a commodity and amenable to outsourcing. Whether this theory works in practice is an unknown operational risk [though the emergence of entire sectors of enterprise software from database to app server are proof points].

This comment doesn’t imply that the IT platform is not a good market opportunity; it may be a nice piece of extension business but clearly one that is secondary if the goal it to create a social media business. As stated earlier, it may be a very large but distinct business opportunity in and of itself.

Overall, the framework starts out pretty strong in contrasting the Level 1 vs. Level 2 (outbound versus inbound platforms) but loses strength on the key new idea. The confusion stems from mixing different concepts that don’t necessarily hold together when thinking about it from both a business and technical perspective. Here, we’re referring to the conceptual leap between Level 2 and 3.

But, read this post if you have not already. It will make you think about a number of inter-dependent issues.

Major kudos go out to Marc because he’s put his finger on another super-strategic issue in IT convergence. It’s one that has the potential to re-define the landscape. The basis for sizable platform businesses at the application and IT levels are already beginning to emerge. I’ll be writing more about considerations that may drive the playing field at different levels of the technology stack.

• The Web is the Platform
• The Social Network Operating System
• Andreessen vs. Cisco: Web 2.0 platform competition heats up
• IBM Confirms: Google Poses NO Enterprise Threat

Filed under: Google, Web Apps, Innovation, Enterprise Technology

Email It   Digg This!   Slashdot It!

Tags: web platforms, syndication, Facebook, Google, Cisco, IBM, Amazon,