Peter Yared's Musings

Facebook is indisputably a platform since it provides a set of API's on which developers can deliver applications for a broad set of users. Facebook also offers a set of bundled applications, such as Pictures, Events, and Status, which are clobbering sites like Flickr, Evite, and Twitter, since these are inherently social applications - "Judy was just tagged in the photo Jim's Birthday Party!" gets a lot more attention than someone's Flickr feed.

We can learn from history here: Windows was one of the first platforms to both court developers and bundle applications, and Microsoft progressively added more and more applications, such as anti-virus, browser, email, fax/scan, movie editing, etc., and completely obliterated entire industries.

Companies like Slide and RockYou have a majority of their page impressions from delivering horizontal applications such as SuperWall, FunWall, and Top Friends. Facebook's upcoming profile redesign is going to include a Wall redesign which will likely include the popular features of SuperWall and FunWall, and the new Friend list feature will eventually make their way into Friend views.

So while you can get some short term play out of delivering a horizontal application for a platform, it is a fact of life that sooner rather than later you are going to get steamrollered, just like Norton and Netscape were on Windows. Focusing on the long tail of vertical applications, which on Facebook is apps like Compare People and Texas Hold'em Poker, is a more viable bet for longer term application growth and monetization.

Database requirements for websites have become simpler and simpler. Finding a product of a certain type or the invoices for an account is very straightforward and is in fact hierarchical and does not require dynamic relationships. It is an ironic twist that the hierarchical databases of yesteryear like IMS are more suited to today's applications than relational databases.

The main feature of a relational database is relationships - "joining" data across different tables of data. But no one in their right mind does complicated joins of data on a production web database because it is so slow, even when using features like views. MySQL, which has completely dominated the web database market for years, only recently has accomplished feature parity to commercial databases like Oracle.

When people need to do complicated data queries, they copy the data from the production databases to what used to be called a "data warehouse" but is now in fact just another database and run reports off of that database instead of the production database.

Both Amazon EC2 and Facebook Platfrom, which are attracting legions of next-generation developers have recently added simplified database offerings that are essentially hierarchical databases. These databases will meet the needs of most web applications and widgets respectively and do nothing else. Amazon's database is even called "SimpleDB" - for web apps, simple does the trick.

Commercial open source has offered a revolutionary business model for enterprise software. Rather than exorbitant high prices, long sales cycles, impossible to install software, systems engineers, and vendor lock-in, commercial open source software is self-service. By the time a customer pays for it, they have already identified a problem, found the open source software as a possible solution, evaluated it, deployed it, and rolled it out to their organization. Commercial open source companies save in recruiting costs by hiring developers who have already found their project and are proven contributors.

The cost of hiring, marketing, distribution, and sales has gone to near-zero, and those savings are passed on to the customer. Traditional, bloated proprietary software can not compete with this model at all, where even their yearly maintenance cost is more than the cost of the commercial open source equivalent.

There have been three commercial open source successes to date: Red Hat, JBoss, and MySQL. By success I mean companies with a lot of customers, a lot of revenue, and a high liquidity event. There are a slew of commercial open source companies with significant traction that are following in their footsteps, including Alfresco, SugarCRM, and Mulesource. All of these companies are all remarkably similar in three critical ways:

- Direct Replacement for a Proprietary Equivalent
- Easy-to-Install Server
- GPL License

Direct Replacement for a Proprietary Equivalent

The successful commercial open source companies all offer direct replacements for existing proprietary software. UNIX operating systems, J2EE application servers, and databases are very quantifiable pieces of software with well known features and benchmarks. The open source up-and-comers that offer content management, CRM, and ESBs are almost as quantifiable and replace well-known, established proprietary software.

The downside to this requirement is that the successful commercial open source projects are frankly not that innovative technically. Startups can innovate on either technology or business model, and open source is clearly a business model innovation. While there are some edge features in Linux/JBoss/MySQL that are not available in Solaris/WebLogic/Oracle, overall these are not products that represent a technical revolution. Before you flame me on this one, contrast the difference between the software that is getting replaced by open source and their predecessors: Solaris/WebLogic/Oracle/etc. were huge technical leaps over VMS/PowerBuilder/Collanet/etc.

We really learned the lesson on this at ActiveGrid, where we put a ton of energy into innovating on top of the LAMP stack, but we were not a direct replacement for an existing product. It is a lot easier to sell JBoss or lightweight Java into a WebLogic shop where they can compare apples to apples, than selling a new way of doing things like the LAMP stack.

Easy-to-Install Server

Servers are traditionally very unwieldy pieces of software that require a lot of planning, expertise, and vendor systems engineers to get installed and working. Marten Mickos told me years ago that the key to MySQL's success was their 15 minute install policy - that users should be able to install the software and be up and running within 15 minutes. Taking what used to be complicated, cumbersome server software and making it a lot easier to install and administer is now a hallmark of successful commercial open source software.

There have been a few open source successes on the desktop like Eclipse and perhaps OpenOffice, but note that no one is making significant money on these, so they are not commercial open source successes. Firefox is bankrolled by Google as a hedge against Microsoft. Software that installs on the desktop is held to a very high standard vs. server software, where installers are usually a bit more sophisticated and forgiving.

GPL License

After the first set of commercial open source successes were all GPL licensed, the next set of companies all switched their licenses to GPL from licenses like MPL+Attribution. The GPL requires that changes to the application be shared, but creative attempts to force the user's work product to be shared have been abandoned, (like MySQL's GPL PHP driver that would force people to either buy the commercial version or share their PHP app). Enterprises are now comfortable with using GPL software and understand they will not have to share their work product, just changes to the open source software itself. Most enterprises are already running Linux, JBoss, or MySQL somewhere, so it is easy to get new software approved that is using the same license. If the license is different, the advocate in the organization who is trying to bring in the open source software has to go through legal review with their compliance office, etc., which is a pain, so they are likely to either shelve the solution for a while or just try using something else.

Getting software into an enterprise has evolved over time. In the 80's and early 90's, enterprises had to pay for every bit. In the late 90's and early 2000's, companies like NetDynamics where I worked gave tooling away for free and charged for servers. Now companies give servers away for free with the GPL license and charge for tooling.

Conclusions

At the beginning of a market, there is a lot of debate as to the right approach. When a market matures and there are successful products, it is hard to argue with their success. Clearly the path to commercial open source success is to target proprietary server software with known features and benchmarks, create a self-service and easy-to-install equivalent, and offer it under the GPL license with a commercially licensed option. There are now numerous companies following this recipe in every niche of enterprise server software.

What's Next?

Software-as-a-Service is a business model that is using open source server software to eclipse open source itself. Rather than deal with deploying and maintaining software at all, it is increasingly appealing to businesses to simply outsource a particular application to a vendor and access it through a web browser. Most of the innovation I am seeing nowadays is in this category, which is taking on both desktop and server software.

This is a crosspost of an article I wrote for BusinessWeek for their special report CEO Guide to Widgets.

Widgets: The Future of Online Ads

Look to the evolution of television advertising to understand the necessity of widgets in today's online world

The sheer volume of Web sites has grown so overwhelming that an increasing number of consumers—not just those in their 20s—are adopting multipurpose tools to help them manage and personalize the vast amount of data thrown at them every day. The mainstream adoption of online social networks such as Facebook and MySpace and personalized home pages such as iGoogle and Netvibes reflects attempts by consumers to make the Web more manageable. This new mindset, not surprisingly, also holds for the way in which the audience is willing to engage with ads.

Advertisers, many of which have only just begun shoveling more of their marketing dollars into promotional banners and boxes on Web sites, had better heed this shift sooner than later. Some Web trends prove to be fads, of course. But when hordes of consumers start shifting from one technology to another, it's a good bet "old" marketing strategies will grow less effective. (Or so I say, which is why I started a new company called iWidgets last year to help advertisers develop campaigns that embrace the new mentality.)

Learn from TV

If this shift in consumer behavior in response to information overload sounds familiar, it should. We've seen much the same with television. When there were only three major networks, people would walk up to the TV, choose a channel, and watch it for a while. When cable television emerged, people would flip through 50 to 75 channels to see what was on. But with the advent of digital cable and satellite TV, which made hundreds of channels available, many overwhelmed viewers started looking for ways to manage and filter the content. This created a market for digital video recorders, such as those from TiVo (TIVO), and on-demand programming. This in turn changed the game for marketing, as users were suddenly choosing which ads, if any, they'd watch.

This same phenomenon is now taking hold online. Instead of visiting more and more Web sites, users are shifting to the TiVos of the Internet. Consumers are using RSS aggregators to display news and blog postings from numerous sources in one place. They're using personal portals such as iGoogle and Netvibes to view customized collections of stock quotes, local weather updates, and relevant news headlines at a glance. They're using social networks including Facebook and MySpace to share pictures and "what's up" with friends, family, and even business associates.

A Behavioral Shift

As with TV, this behavioral shift is having a significant impact on the online advertising industry, a multibillion-dollar business dominated by Google (GOOG). Until recently, the whole thesis of online marketing has been that ads need to be propagated across thousands of Web sites so that users see them, click on them, and visit advertisers' sites. But nowadays, there are several problems with this approach.

One is that users are not as willing to leave the site they're using to visit an advertiser's site. Although the economic downturn may be playing a role, it's worth noting that Google and Yahoo! (YHOO) both recently reported drop-offs in their click-through rates (BusinessWeek.com, 2/27/08). The two companies have come up with a myriad of explanations, but the reason is actually very simple: Web users are setting a higher bar for advertisers to lure them to another site.

In fact, many users are starting to ignore ads altogether, a phenomenon known as "banner-ad blindness." At the same time, with advertisers shifting more and more of their marketing dollars online, the cost of an "impression"—placing an ad on a site—keeps rising. The net result is that online advertisers are now paying more to get less.

Beyond Online Billboards

So how should advertisers adjust to this new situation? First, let's look at what they're buying when they place an online ad. They are buying a piece of real estate on a Web page that's essentially no more than a link to another Web site. But that rectangle of space can also be exploited to function as a site within a site. Instead of just replicating a billboard on the side of the highway in a bid to get users to click and go somewhere else—something they hate to do—the ad should actually let them do something right then and there. Right in that little box, without switching sites, users should be able to browse products, customize orders, and eventually make a purchase. This concept of a little Web site embedded within another Web site is called a widget.

Widget ads are just the first step. Instead of simply trying to build brand awareness, marketers now have the ability to reach out to customers with useful features to enhance their personalized pages on social networks. An athletic gear company could offer an application that lets a group of running buddies track how their times and distances compare. An airline could offer a "come visit me" application that displays the latest fares for a trip between the hometowns of viewers and Web page owners. Even better, an airline might offer a widget that lets users track their frequent-flier miles and search for award trips right from their iGoogle pages. These types of applications can be useful, engaging, and viral—when you see one on your friend's profile, you are likely to install it as well.

Advertisers and consumers have played a game of cat and mouse for years. As we've learned from the ongoing upheaval in the entertainment industry, online advertisers have to adapt their approach to match consumer behavior, or risk irrelevance. They need to move beyond glorified billboards and deliver useful, engaging applications people will want to use and share.

Peter Yared is a serial entrepreneur and the founder and CEO of iWidgets, a San Francisco startup helping brands deliver next-generation Internet advertising.

AutoByTel just commissioned a study on people's search habits, which found that 72% of Americans are experiencing "Search Engine Fatigue". They search but they are not finding what they are looking for. We in the industry are quite familiar with search fatigue. As I have posted before, most of the searching I do is vertical search from the Firefox toolbar with plugins like Wikipedia, Yelp, IMDB, Whois, etc.

Search ranking based on counting the number of inbound links made a lot of sense when inbound links actually meant something. Today, people have gamed the system so much that inbound links mean nothing. If you don't believe this statement, search for something that nobody cares to make money off of like Bridge Construction and the first few links actually are relevant and there are relevant targeted ads off to the right! It the mindblowing Google experience we all had a few years ago when we set Google to be our homepage.

Now search for something like Futon Filling or iPod Connectivity. These types of searches used to point to the content that people thought was most relevant. Now it points to the content that has the most cross linking by smarmy advertisers. Global search has made a gradual transition over the years from something useful to something useless, yet people are still repeating the patterns that used to work and wondering what is going on.

Is link-based page ranking dead? Maybe not, Jimmy Wales announced last year that he is building a search engine based off of the number of links from Wikipedia, which are well qualified links. Google keeps trying to stem the tide of bad links, but eventually they will have to change their algorithm to reflect the reality that most links suck. In the meantime, I think it is safe to say that global search is pretty lame, and I am going to stick with vertical search for most of my searching.

I got some good natured ribbing when I posted nine months ago that Google would soon exceed Microsoft's market cap on the rational that the market would soon value the dominant Internet company as worth more than the dominant PC company. At the time, Google was worth $150B (64 P/E) and Microsoft was worth twice as much at $300B (26 P/E).

Google is now worth $201B and Microsoft is worth $284B, so Google is now worth 71% as much as Microsoft. If Google's P/E had held at 64 it would now be worth more than Microsoft. At Google's current P/E of 55, it only needs to increase its annual revenue by a few billion to equal Microsoft's market cap.

That would put GOOG at $915/sh, although it is likely that some type of market correction will lower the value of both companies. It is also interesting to note that Google is now worth more than CSCO, the dominant networking company, whose market cap is $192B (27 P/E).

Two interesting DHTML library announcements on the same day:

JQuery UI 1.0 has finally been released! JQuery is best known as a phenomenally efficient way to update and manipulate DOMs from JavaScript. It is unbelievable how much JavaScript code one line of JQuery can replace. JQuery has numerous user interface plugins, the most important of which have been upgraded and delivered as the new JQuery UI library. My new project has an extensive amount of UI intensive features like drag-and-drop, resize, and DOM manipulation, and I went through several JavaScript libraries until Carnet Williams at ChipIn told me about JQuery's interface plugins. I have been incredibly pleased with JQuery UI, in particular its tight codebase and its very clean way of separating the DOM from JavaScript DOM enhancements.

My old company ActiveGrid has announced it has acquired the TurboAJAX Group and is going to support the Dojo toolkit. I think that this is great news, Dojo has a ton of great widgets and features, but it can be hard to learn and extend. Having a company officially support Dojo is an important step in increasing its adoption.

It's great to see the DHTML JavaScript market maturing. The better the libraries get, the more great web apps we all get to use!

# posted by Peter Yared @ 1:21 PM 0 comments