jump to navigation

Three ways to build an online media business to $50m in revenue February 26, 2007

Posted by jeremyliew in advertising, Consumer internet, Digital Media, Entrepreneur, Internet, start-up, startups, user generated content, VC, Venture Capital, web 2.0.

As a venture capitalist, I’m interested in investing in companies that could be big one day, that could get to at least $50m in revenue.

Here are three ways to get to $50m in revenue as an online media business; indulge me in some math:

1. Be a site with a broad reach (say general social networking, communications, news). At large scale, without a great deal of targeting possible, a startup’s “run of site” or “run of network” advertising might be able to get to the $1 RPM range (Revenue per thousand impressions, including CPM, CPC, and CPA models). To get to $50m in revenue you would need 50 billion pageviews in a year, or just over 4 billion per month. According to Comscore, Bebo had the 10th most Pageviews in the US in Janurary 1007, with 3.4bn, so you would need to be bigger than that.

2. Be a site with demographic targeting (say a Latino portal, or a sports site (targeted at men) or a social network targeted at baby boomers). Although in TV and in magazines, demographic targeting can generate double digit CPMs, online at scale, RPMs tend to be in the low single digit range. Lets assume a $5 RPM. To get to $50m in revenue you would need 10 billion pageviews in a year, or just over 800 million per month. According to Comscore, Microsoft had the 22nd most Pageviews in the US in January 2007, with 792 million, so you would need to be bigger than that. [Microsoft isn’t a demographically targeted site – i just use it as a comparison point for overall traffic size.]

3. Be a site with endemic advertising opportunities (say a site about movies that movie studios will want to advertise on, or a site about cars that auto manufacturers will want to advertise on, or a site about travel that hotels and airlines and online travel agencies will want to advertise on). If you have a highly targeted audience that is interested in buying a specific product, you can command RPM’s well into the double digits. Lets assume a $20 RPM. To get to $50m in revenue you would need 2.5 billion pageviews in a year, or just over 200 million per month. According to Comscore, Adelphia.com had the 125th most Pageviews in the US in January 2007, with 198 million, so you would need to be bigger than that. [Adelphia isn’t an endemically targeted site – i just use it as a comparison point for overall traffic size.]

Admittedly, all these Comscore #s are US only, and all businesses will have international traffic as well, but the principle still holds.

Which do you think is easiest?

UPDATE: If you liked this post you will likely like my prior post on why new forms of advertising are hard

UPDATE II: To all new visitors, if you like what you read, subscribe to the Lightspeed Venture Partners blog RSS feed. Its at the bottom of the Right Hand Side column. We post 2-3 times per week on topics including consumer internet, web 2.0, lead gen, ecommerce, startups and venture capital.

UPDATE III: I’ve posted more on the difficulties in building a media business to $50m in revenues here.

The pros and cons of “Stealth mode” February 22, 2007

Posted by jeremyliew in start-up, startups, VC, Venture Capital.

I recently heard about a company that has been in “Stealth mode” for four years. That is certainly on the extreme range, but I’ve noticed an increasing number of companies in stealth mode, as can be seen by seaching on Craig’s list job ads. Indeed, one of my portfolio companies, Tippit, is in (partial) stealth.

This got me to thinking about the pros and cons of stealth mode. Typically, companies want to be in stealth so that they can emerge “fully formed” into a new category with a big enough head start over potential competitors that they can’t be easily caught.

However, there are a lot of drawbacks to being in stealth mode as well.

    Hiring can be very difficult. This can be mitigated if the founding team can hire enough people that they know directly to fully staff out the initial organization. Otherwise, without the personal trust in an existing founder/employee/investor, it can be hard to pique the interest of a highly qualified candidate with many other easily understood options. Getting them to sign an NDA before you tell them what the company does is hard given the plethora of options that good people have right now.

    Business development can be difficult.
    As I’ve posted about in the past, in the early stages of a new consumer technology category, distribution is critical to success. Getting these distribution deals lined up can be difficult while in stealth mode. Again, this can be mitigated if the founding team already has existing relationships into potential distribution partners. Personal trust can overcome uncertainty about the benefits of a deal, at least enough to take an initial meeting.

    Less serendipity. It’s been said that it is better to be lucky than smart. Startups often benefit from serendipity; when people who know the company introduce them to potential employees, partners, customers, or tip them off on competitive threats, opportunities or new markets. The less people who know the company, the less opportunity there is for serendipity.

Given these drawbacks, when does it makes sense to be in stealth? I think some of the following criteria need to be met for the tradeoffs to be worthwhile:

    The idea is truly novel. If the idea is an extension of existing platforms, you’re likely to see a lot of competitors regardless of what you do. “High concept” movies are movies that can be summarized in one sentence, often referencing another well known movie or book e.g. “Superfriends in 19th Century London“, “Schindler’s list in Rwanda“, “Jaws in space“, or “Heart of Darkness during the Vietnam War“. If you have a “High Concept” startup (e.g. “Facebook for China“, “Digg for Games“, “Flickr for Video” or “Myspace for Baby Boomers“) then you’ll likely have a lot of other people trying to do the same thing, and being in stealth mode doesn’t help you much. (NB none of these companies are in stealth mode.)

    Technology development is lengthy but not difficult. If the technology problems are difficult, then that alone should be enough to be a barrier to entry. Execution is always a lot harder than just ‘having a great idea’. If the technology development isn’t lengthy, then being in stealth mode won’t give you much of an advantage once the product gets launched anyway. Hillcrest Labs took this approach while developing their next generation of TV remote controls.

    Sources of proprietary advantage are undervalued. Some companies depend on taking information that is publically available but hard to get and making it searchable and available online. Others might be buying undervalued assets and monetizing them better. In both cases there is real advantage to keeping your strategy secret until you have locked in your advantages and to keep other potential bidders for the asset in the dark.

    Your entry into this space will cause people to follow you by virtue of who you are.
    Apple certainly felt this way about their iPhone. Some Venture Capital firms also prefer to keep their investments in stealth because they feel that their investment in an industry is enough to cause a flood of competitors to enter that industry. Sometimes this is true, sometimes the time is right for an idea to happen. Entrepreneurs will need to decide if their Venture Firm’s relationships will be sufficient to counter the increased difficulty in increased hiring and business development.

    I’d love to see comments from readers who are or have been in Stealth

Adweek on why all pageviews are not created equal February 20, 2007

Posted by jeremyliew in advertising, Consumer internet, Internet, social networks, user generated content, web 2.0.
1 comment so far

Through Mashable, found a link to an excellent article at Adweek that talks about why the value of a pageview at web 1.0 companies like Yahoo! are higher than the value of pageviews at social network companies like Myspace and Facebook.

The short of it is that audience engagement matters as well as pure volume of pageviews. If users are not spending much time on pages, they are not in the mood to click ads. This is compounded for performance based advertisers as multiple explosure to the same audience eventually leads to dramatically lower click through rates. But don’t let this summary dissuade you from reading the whole thing.

New forms of advertising are hard February 19, 2007

Posted by jeremyliew in advertising, Consumer internet, Digital Media, Internet, Lead gen, Search, start-up, startups, web 2.0.

I’ve seen a few startups recently that are relying on launching a new form of advertising as their business model. These can include product placements, sponsorships of various flavors, new forms of local advertising, interactive out-of-home advertising, and lots of variations of mobile advertising. This is a hard business. If successful, it can be very, very successful (e.g. Overture/Google with sponsored links in search) but entrepreneurs often underestimate how long it will take for revenues to ramp.

To understand how new forms of advertising get adopted, you need to understand how advertising is bought today. In most instances, ad agencies control the ad budgets for the largest advertisers in the world. Within those ad agencies, one of the functions is media buying. A media buyer’s role is to optimize reach (and sometimes quality of audience) for their client across all possible advertising channels. The problem with new forms of advertising is that they are often not represented in the media buyers’ spreadsheets and models. And if it’s not in the model, it doesn’t get allocated any ad spend.

Startups sometimes get traction with a new form of advertising because there are always some forward thinking advertiers who are willing to experiment. This early traction is often a customized program negotiated with an advertiser that is friendly with the startup through personal relationships. However, crossing over from a “business development” focused model (where each new deal is custom crafted) to an “ad sales” focused model (where standardized products are sold off of a rate card) is the key to massive scalability of revenues. To do this you need to get into the media buying model; you need to sell a standardized product.

For internet companies, that usually means that you need to get the IAB (Internet Advertising Bureau) to issue a new “Standard” ad unit, in much the same way that the IAB issued its first set of “voluntary guidelines” that set up 8 standard banner ad units in 1996, a massive reduction from the over 150 ad sizes that were in use at the time. This standardization greatly eases logistical complexity for both advertiers and media companies.

The process of creating a new standard can be quite a lengthy one. It usually involves a coalition of both media companies and advertisers coming together and negotiating the key elements of the standard. The composition of the IAB board is usually dominated by larger online media companies and it can be hard for a startup to have much influence on this decision making process. It can often be easier to align youself with the interests of a larger media company and let them carry the water up the hill, rather than trying to do it independantly as a startup. If you’re Dogster, you’ll have less success pushing a new standard for “sponsored profiles” than MySpace/FIM or AIMpages/AOL. So making sure that your sponsored profiles packages contain the same elements as those of the big guys will make your life easier as they take this new ad unit through the standardization process

The alternative approach is to make sure that your new form of advertising so closely parallels an existing standard ad unit that it can be considered within the existing bucket. 30 second online video ads (same format as TV),online leads (similar to phone leads) and new variations of CPC advertising (similar to search) have all been “close enough” to an existing ad unit that they have been able to tap existing ad budgets and grow quickly.

In either case, when building business plans on the assumption of the adoption of new ad units, make sure you give yourself enough time in your plans for the market to be created before it can grow to scale.

Useable Health Vertical Search February 16, 2007

Posted by ravimhatre in Consumer internet, Entrepreneur, Internet, Search, start-up, web 2.0.

There was a TechCrunch post today regarding a new search service from Healthline called Symptom Search which attempts to provide an information service suggesting common illnesses related to symptoms that a user is experiencing.

Symptom Search is a great idea. The challenge is to acheive comprehensiveness such that typical symptoms are accurately and completely mapped (in relevant order) to all possible diseases and vice versa.

Just for fun, I tried a couple of not uncommon symptom searches related to actual problems people I know have experienced recently. For example, one person I know experienced chronic tendonitis in reaction to being treated with Cipro. Unfortunately there were no results in Symptom Search that relate these two subjects.

I also had a recent experience with someone who experienced sudden hearing loss as the result of an ear infection however the system was not able to correlate the two and didn’t serve any relevant information.

Health information search is a technically challenging problem and one I believe requires a search metaphor and deep technology to address in a way that is meaningfull to the consumer. See my previous Lightpseed blog post about Vertical Search as a way to better address these types of research oriented searches.

Also, try the following two searches on the Kosmix Health Search portal (full disclosure, Lightpseed is an investor).



I typed in simple search requests such as “Sudden Hearing Loss” and “Cipro Tendonitis” and got back a wealth of topics and articles that linked these symptoms to their underlying causes and also to potential treatments and doctors.

Health Vertical Search is a challenging problem but one, if well solved, that could yield substantial consumer benefits as well as create a company of significant value.

Great analysis of digital goods/gifting February 14, 2007

Posted by jeremyliew in Consumer internet, Ecommerce, Internet, social networks, startups, user generated content, viral, viral marketing, web 2.0.

Great post from Dana Boyd about the new digital gifts that Facebook released recently. After ranting randomly about rampant commercialism, she gets to the good stuff; analyzing gift giving dynamics (self worth, scarcity, levels of value, reciprocity, importance of timing) and how Facebook’s release measures up to that (spoiler – not very well). Well worth reading.

Hotornot.com were probably the first in the US to experiment with digital gifting, James Hong also blogged about it recently in context of the LiveJournal release of digital flowers.

Update: Read James Hong’s reactions to Dana’s post here

Wikis vs blogs vs user generated databases: which to use and when? February 13, 2007

Posted by jeremyliew in Consumer internet, Internet, Uncategorized, user generated content, web 2.0.

I’ve been thinking a bit about the many different flavors of user generated content models; wikis, blogs, discussion boards, etc, and trying to come up with some sort of framework about what works best for which purpose. It seems that non-fiction user generated content can be categorized into three different types:

    1) “One right answer” content (e.g. dictionary definitions, lyrics to a song, plot summaries etc)

    2) “Many right answer” content (e.g. recipes. I have one recipe for meatballs. You have another. Both produce meatballs if followed, so both are “right”. Yet your meatballs may be juicy and suculent while mine may just suck.)

    3) “No right answer” content (e.g. Answers to question like “Who was the greatest shooting guard of all time?” or “What should we do in Iraq?”. These are conversations that can be debated without ever getting to a “right answer”).

These seem to match up best with three models of organizing user generated content:

    1) Wikis. Anyone can contribute and edit content, so only content that is generally agreed to remains in place. (Not quite the same thing as being “right”, but certainly a proxy). The contributing authors’ names are de-emphasized because identity of the contributor is not important when dealing with some canoncial truth. Wikipedia is of course the best example, but others include wikiXbox360 (a Wetpaint powered site) [game cheats are a great example of “one right answer” content], Battlestar Wiki, the Wikia sites, and many more.

    2) User generated databases, enabling rating and reviewing. Allrecipes.com allows users to submit, rate and review recipes, so that my horrible meatball recipe gets buried and your wonderful meatball recipe bubbles to the top. The same principles apply to Youtube, and to the reviews from sites like Yelp, Flixster and Amazon. Reviews somehow seem to behave more like “many right answer” than “no right answer”, perhaps because they are not meant to be part of the give and take of an argument. But because the identity of the reviewer, recipe contributor, or videographer is important (as an indicator of quality, or based on history, as an indicator of your liklihood to approve), the contributing author’s name is central to these sorts of databases.

    3) Message boards and blogs. Although some claim that the web is a conversation, its only primarily blogs and message boards that carry out this promise. It’s here that you see call and response behaviour carried out, both within a message board or blog (through comments discussion) or across blogs as posts spur counter posts. In this richness of debate, once again, identity of the author is critical, as is the case in an offline conversation. “Who said that?”, or “Where did you read that” is a common response to a quoted opinion. With no objective measurement of truth or value (unlike “one right answer” and “no right answer” models), rating and ranking makes less sense and each comment stands on its own and can be responded to independently.

    Even within a single topic, there is room for all these models to coexist. For instance, in travel, Wikitravel provides the objective facts about travel destinations in wiki format. Yahoo!’s Trip Planner and TripAdvisor use ratable user generated databases to the “many right answer” problems of sample itineraries and hotel reviews. Lonely Planet’s Thorn tree and GridSkipper lead with Q&A style message boards and thought provoking blog posts to start conversations. Each have their place.

    When there is a mis-match between model and content type, there can be bad results. The best example of this was the LA Times’ experiment with a wiki editorial a couple of years ago. Mixing a “one right answer” model with “no right answer” content led to chaos!

    If I could, I’d make this post a wiki as I think there is “one right answer” to this model for user generated content, and I’m not sure that I have it right yet. As the next best thing, I’d welcome comments.

    Update: Some good comment discussion below from folks like Rich Barton (Founding CEO of Expedia, Zillow), Gil Penchia (CEO of Wikia) and others. If you’re reading this in RSS, its worth clicking through for the comments

Understanding the users of social networks February 11, 2007

Posted by jeremyliew in Consumer internet, Digital Media, Internet, social networks, user generated content, web 2.0.

New York Magazine has a great article in its current edition where it attempts to understand the psyche of the users of social networks, especially the younger users. Although as investors and entrepreneurs, we do our best to get into the heads of these users, many of us are a generation (or at least half a generation!) too old to be of this demogrpahic. The article won’t have much new to say to regular readers of Dana Boyd and Clay Shirky, both of whom are quoted in the article, but its worth reading none the less.

The author, Emily Nussbaum, has been tracking this theme for several years and wrote a good story in the NYTimes a couple of years ago on teen blogging that is also a good read.

Nussbaum talks about younger internet users’ willingness to reveal their private lives on the Internet (blogging, posting photos, sharing “private” details on myspace or facebook) as causing the first true generation gap since rock and roll 50 years ago. She satirizes the older generation’s “disgusted, dismissive squawk” as going something like this:

Kids today. They have no sense of shame. They have no sense of privacy. They are show-offs, fame whores, pornographic little loons who post their diaries, their phone numebrs, their stupid poetry- for God’s sake, their dirty photos! – online. They have virtual friends instead of real ones. they talk in illiterate instant messages. They are interested only in attention – and yet they have zero attention span, flitting like hummingbirds from one virtula stage to another.

Nussbaum highlights three changes in the mindset of younger internet users who grew up as Digital Natives:

They think of themselves as having an audience

They have accepted that in a world of cellphone cameras and security videos and Google, there is no guarantee of privacy anyway. Whereas once only celebrities and politicians had to parse every sentence, these users have learned to modulate their voices to address their constant audience. They have learned the coping skills of Paris Hilton: enjoy the attention instead of fighting it and do your own publicity before someone does it for you. One telling quote “why go to a partiy if you’re not going to get your picture taken?”

They have archived their adolescence

Unlike older generations, whose memories of their youth are largely centered around yellowing old photos in shoeboxes and the occasional artifact – a trophy or an award, this generation has a detailed history of their lives stored digitally. This archive is open to anyone – potential employers, spouses, friend – but most importantly it’s open to its creator, who can dip back into her past with ease. Its certainly open to online predators and pedophiles too – but worrying about that is akin to saying “How can you move to New York? You’ll get mugged!” in the minds of these users who have grown up learning to deal with porn popups and IMs from dodgy strangers.

Their skin is thicker than yours

Public life is fun. It’s creative. It’s where your friends are. It’s theatre, but it’s also community. But the flip side of all of this is that you also open yourself up to shaming, to snark, to anonymous flaming. These internet users can feel embarassment and absorb it without shutting down. Quoting another of Nussbaum’s interviewees “Well, I guess you have to be sort of honored that someone takes the time to write about you, good or bad”.

The key takeaway for me is that user generated content on social networks (whether away messages on AIM or posts on the Wall at Facebook or restaurant reviews at Yelp) is as much about performance (for an audience) as it is about communication. Self expression is not about what I think, its about what I want you to think about what I think.

In an offline world, my self expression is limited to choosing from a predefined set in a confined space: which bumper sticker will I put on my car, which poster will I put on my dorm room wall, which colored charity band will I wear on my wrist. In each case, I’m carefully choosing my “self expression” to make sure that it casts me in the light in which I want to be seen. Online, my self expression is unbounded. As is often the case, online adoption occurs fastest when it mirrors existing offline behaviour and improves it, and we definitely see this in the area of social networks.

So while I highly recommend reading the article, and agree with a lot of its observations, I ultimately don’t think that this generation is so hard to understand. They’re just like us, only moreso.

What to do if you are a platform with two-sided network effects February 9, 2007

Posted by jeremyliew in Consumer internet, Digital Media, Ecommerce, Entrepreneur, Internet, start-up, startups, VC, Venture Capital, web 2.0.

I subscribe to Harvard Business Review but rarely read it – the long articles intimidate me! However, a friend of mine recently pointed me to a fantastic article in the October 2006 edition entitled “Strategies for Two-Sided Markets” by Thomas Eisenmann, Geoffrey Parker and Marshall W. Van Alstyne that is well worth reading.

The article addresses a common phenomona in technology, where a platform brings together two groups of users, each attracted to the other group. Examples might include Ebay (bringing together buyers and sellers), Monster (bringing together job seekers and employers), Youtube (bringing together video posters and video watchers), even Gaming Consoles (bringing together game developers and players). The article discusses all such platforms, but I found the most interesting cases to be when members of both groups want to be part of the platform where there are the most members of the other group, or “cross-side positive network effects” in the parlance of the article.

The article answers two critical questions to owners of such platforms:

Who do you charge?

In a two sided market, you can charge one side or the other, or both. Typically platforms end up charging one side (the Money side) and subsidizing the other.The article suggests that there are six guidelines in making this decision, some more obvious than others. I’ll highlight three of them here:

    Subsidize the more price sensitive side. E.g. Adobe gives away its PDF reader for free but charges for Acrobat.

    Subsidize the side that is more sensitive to quality (i.e. charge the side that has to PROVIDE quality – and let suppliers use their willingness to pay as a signal of quality). E.g. console game players (PS3, Xbox 360, Wii) demand high quality. Developers pay a high fixed cost to deliver quality so they need a lot of players to make their business models work. They are willing to pay a high royalty and adhere to strict licensing terms to reach those big audiences.

    If strong “same-side negative network effects” exist on one side (e.g. if one side would prefer not to see too many others on the same side, such as competitive suppliers), charge that side and possibly limit the number of available slots. E.g. Autobytel gives zip code exclusivity to dealers in a given territory and charges dearly for that limitation

Proprietary or shared platform?

Often in these sorts of markets only one platform will survive because both groups want to be on the platform where there are the largest numbers of the other group. Competitive platforms need to make a “bet the company” decision – to fight to be the winning platform, or to cooperate and share a single platform. The article says that there is typically a single winner if the following conditions apply:

    Multi-homing costs are high for at least one user side – i.e. it is expensive to support multiple platforms. This is true in the case of online auctions (if I only have one antique cuckoo clock, it can only list it in one place) and not true in the case of job sites (its easy to post my resume on both Hotjobs and Monster).

    Neither side’s users have special needs. If the market can be segmented then different platforms can co-exist, each serving a niche.

High definition DVDs looks like a market that fits these criteria. Blu ray and HD DVD are battling it out to be the winning platform. Typically the winner of such a war of attrition will have (i) Cost or differentiation advantages (longer play times, better image quality) (ii) Pre-existing relationships with prospective users (movie studios) (iii) Reputation for past success and (iv) Deep pockets. Often, because of the confidence of the executive teams in their products and the winner-take-all nature of these industries, competitors will choose to fight. However, cooperating can also bring benefits including a greater overall market size (when standards take longer to emerge many users delay commiting to either platform) and lower overall marketing costs.

The article also addresses questions such as when to be a first mover (and more importantly, when not to be a first mover), and how to deal with envelopment from adjacent platforms (relevant to those facing Microsoft’s “Embrace, Extend, Destroy” strategy). If your company is a platform that faces these “cross-side positive network effects”, I highly recommend reading “the article

A play about success in consumer technology, in three acts February 6, 2007

Posted by jeremyliew in Consumer internet, Ecommerce, Lead gen, Search, social networks, startups, user generated content, viral, viral marketing, web 2.0, widgets.

I’ve previously posted on the importance of distribution during the initial phase of a startups life. To be more accurate, I think that distribution is the most important factor for a consumer facing company competing in a new category.

However, I think that this is just the first act of a three act play, with a different factor being critical to success in each act.


To summarize the earlier post; early in a category’s lifecycle, users don’t recognize that they have a particular need/problem. They don’t recognize that a category exists and so there is no demand pull. If even your own mother doesn’t know what it is you do; ESPECIALLY if your own mother doesn’t know what it is you do, then you likely face this problem!

Having the best product is neither necessary nor sufficient. Having a decent product is good enough.

You need to get to users as they won’t get to you. Hence the importance of distribution. Read the original post for flavors of distribution and how to get them, plus examples.


Over time, categories become established in the minds of consumers. In the case of online travel agencies it took about 3-5 years. In the case of user generated video, it took only 12-18 months. Consumers start to understand who the competitors are in an industry. If switching costs are low (as they are in both online travel agencies and user generated video), users often sample the offerings from multiple competitors. At this stage, assuming that you have done enough to get into the consideration set, product and user experience matters a lot. Distribution has become the ante, and the companies that win will win on the best product.

Having the best product means much more than having the most checks in a feature comparison matrix. It goes far beyond the technology. It can mean having the best prices, the best selection/range, the best customer service or the best community. In the early days of online travel Travelocity won on distribution through its deals with AOL and Yahoo! But once the category became established in people’s minds, Expedia slowly took market share from Travelocity on the back of a better overall user experience, one important factor being better pricing (mostly hotel pricing). [Update: Note comments below from Rich Barton, founding CEO of Expedia, on his view of why Expedia overtook Travelocity.]

Similarly, many user generated video sites claim better features than Youtube, yet Youtube never lost its lead because it had the biggest range of content (and the biggest audience for people looking to upload videos). As technologists we can fall into the trap of defining “best product” too narrowly, but our customers are not technologists and they look at the whole experience.


As a category continues to mature, it becomes harder to maintain product differentiation. There are some exceptions; when there are positive network effects an early leader like Ebay or Youtube can often hold their leads. But if the advantages built in Act Two stem from technology, process or supply advantages, these often get whittled away as competitors copy, innovate and partner to make up lost ground. At this stage of a category’s evolution, branding is the most important factor and product has become the ante once again. (Note that I distinguish here between branding and marketing. Online marketing that is more direct response (CPC or CPA) in nature is really more a form of distribution.)

Google is sometimes presented as the canonical example of the best technology winning over time. However, as I mention in the post on distribution, Google‘s traffic only really started to climb after its distribution deals with Y! and AOL. It really did pull away from the other search engines on the basis of its better product during Act Two. But today, its not at all clear that its search results are that much better than anyone elses.

When I was at AOL a couple of years ago, we used to test search relevance from multiple engines by taking the results from all the major search engines, stripping all branding and UI, and showing the lists to users who scored the quality of the search returns. The results were surprising – all the search engines were very close to each other in relevance, with variations as to who was “best” from month to month and search term to search term. Interestingly enough, when you put the branding and UI back in, the users always rated
Google as having the best search results.

Google‘s dominant brand is now what enables it to hold and grow its search market share. And in mature categories such as online travel, you see the big players compete purely on branding ads.

In some categories, branding never matters and we never reach Act three. If your users are unlikely to transaction with you more than once (say you sell rowing machines or curio cabinets) then your category will likely never develop beyond Act One. But if you’re in a category with repeat users, whether books, DVD rentals, online auctions, or shoes, branding matters.


Distribution, product and branding, all are critical but at different times. Making sure you focus on the right factor at the right stage in the evolution of your category can help you make sure you’re fighting the next war, not the last one.