Financing Hyper-Virality in the Clouds

This article probes the implications of cloud computing for financing very rapidly distributed internet-based services and products. It contains rough, inadequately researched thoughts, sparked from discussions at the recent CloudCamp Scotland. Read More

WeeWorld

WeeMee. WeeWorld is a teen-orientated social network, best known for their customized avatars, “WeeMees”. WeeWorld has evolved into an eclectic mix of community, casual games, and virtual goods. Steve Young, creative director, spoke to a small group in Edinburgh. Steve discussed the motivations and behaviour of WeeWorld’s users, and explored the challenges of working with 2D WeeMees, particularly as they move into WeeWorld’s new virtual (synchronous) world.

Users

WeeWorld’s core market are teenagers, mostly in North America. Average age 16 (minimum 13, although younger users may simply lie about their age). 60% are female. The dominant market segment was characterised as “spoilt rich kids” – typically those with their own computers. Of the 23 million registered users, about a million visit the WeeWorld site each month, and 80,000 login each day.

Usage differs from other teen social networks, such as Gaia Online: Only 6% of logged-in users visit the site’s forums, while 80% alter their WeeMee. Teen worlds are evidently not generic.

WeeMees (from the Glaswegian, “little me”) can be placed within personalised 2D rooms (in the style of “cardboard theatre”), used as characters within casual games, or rendered as avatars in a new virtual world called, simply enough, “World”. WeeMees are also used on third party websites and services, including messenger services, such as AIM or Live. Initial ideas for WeeMees had resulted in a lot of avatars simply being copied. APIs now provide some control over how WeeMees are reused.

Users’ main aim is “to gather as many friends as possible”. And to chat in a variant of the English language that even JeffK would find almost unintelligible: $iNG-UL?

Virtual Goods

WeeMees can be customized for free: Body, clothes and accessories. However users can also buy “Points”, which can be spent on specific items.

Points can be purchased via PayPal transactions or pre-paid cards, which are sold in US stores. Kids tend to regard these mechanisms like free credit cards: They are not seen as real money.

People pay for “uniqueness”. However, items need not be complex: The most popular item sold is a simple Alice band.

The most fascinating revelation was that the introduction of the new synchronous (virtual) world doubled the sales of virtual goods. This “World” is not even out of beta testing yet. “World” places WeeMees in the same interactive space as one another. This contrasts to the other areas of the site, where WeeMees are not competing for space. I think that implies the more an avatar needs to stand out from the crowd, the more virtual “Bling” is worth to that avatar’s owner.

WeeWorld is keen to avoid its Points being traded as a virtual currency. Money can only be converted into Points, not back again.

Design

The key to WeeWorld’s success is “immersion”. The key to its revenue is “engagement”. These concepts guide development.

Although WeeMees are cartoon-like (in the style associated with South Park), customizations still need to reflect what people would wear in “real life”. For example, T-shirts branding needs to be subtle – a small logo on part of the garment.

The goal for user-generated content (customizations of WeeMees and rooms) is to make it hard for the user to create something that looks bad. For example, MySpace customisations can (and in my opinion, sadly often do) look terrible.

WeeWorld has adjusted to match conservative US culture. The cannabis plants created in early experiments are long gone. There are no alcoholic drinks. Negotiations with Walmart even forced WeeWorld to disable the customization of boob (brest) size.

The development of “World” posed an interest problem: How should WeeMees move? All the artwork and customizations had been designed for static display, without movement animations. The World uses embedded Flash objects to display information to users, so the amount of data transferred about other users’ movements needs to be minimal.

The solution was to make WeeMees hop. Users can also select a trajectory and fire their WeeMees in a particular direction. Navigating World’s 2D platform-ed environment is quite cereal, but strangely fun!

Development

Social networks are becoming more like virtual worlds, while virtual worlds are becoming more like social networks. WeeWorld is trying to steer a path down the middle. Like all the businesses involved, they are still “feeling their way”, finding out what works.

Development time-scales for WeeWorld (and similar products) are very short. Steve was somewhat frustrated that development of the “World” had taken a whole quarter (3 months). The contrast to video-game style virtual worlds is stark: Those typically take 3 years to construct.

WeeWorld use a Scrum/agile development process (which suits the constantly evolving product). Casual games (a commonly requested feature) are often out-sourced to other developers.

The ability to develop content quickly makes it very easy for good ideas to be copied by competitors. For example, Zwinky might seem remarkably similar…

Map of World of Warcraft Online Communities

Michael Zenke’s MMO Blogipelago map [via Tobold], based on the famous xkcd map of online communities, inspired me to create a map for World of Warcraft (WoW) online communities:

Map of World of Warcraft Online Communities.

Glider MMOwned Emupedia MaNGOS Project Baron Soosdon Olibith Oxhorn WoWJutsu Highlander's Profession Leveling Guides Cash Creating Guide (Advert) Zygor Guides (Advert) Brian Kopp (Advert) Team iDemise (Advert) Wowhead Thottbot Allakhazam Nihilum World of Raids MMO-Champion Roguespot Warlock's Den WoW Trading Card Game Elitist Jerks El's Extreme Anglin' WarcraftPets.com Petopia WoW Insider WoWWiki WoW Radio WoWAce Inc Gamers Curse Ten Ton Hammer Stratics Warcry Gamespot IGN GameSpy BlizzPlanet
Map of World of Warcraft Online Communities.

The article below explains the logic behind the map. Read More

Dave McClure on Social Networking and Web 2.0

Dave McClure addressed a Edinburgh Entrepreneurship Club/Edinburgh-Stanford Link event on 29 January 2008. He outlined some of the advantages of “Web 2.0”, talked extensively on the use of real-time metrics to evolve web services, developed a history of social networking websites, and highlighted the interesting aspects of Facebook. This article summarises Dave’s talk, with some additional commentary from myself.

Advantages of Web 2.0

Web 2.0 is characterised by the:

  • low cost of acquiring large numbers of users,
  • ability to generate revenue through advertising/e-commerce,
  • use of online metrics as feedback loops in product development,
  • sustainable long term profitability (at least for some).

Dave McClure did not actually try and define the term, which was probably wise. Generally the term is applied to websites and services where users collaborate or share content.

Web 2.0 has a number of advantages (although it could be argued that some of these apply to earlier iterations of the internet too):

  • APIs – the ability to act as a web-based service, rather than just a “website”.
  • PC-like interface, albeit still 5 years behind contemporary PC interfaces.
  • RSS feeds (for data sharing) and widgets (user interfaces embedded elsewhere).
  • Use of email mailing lists for retaining traffic. While email certainly isn’t a “web 2.0” technology, his argument is that email is increasingly overlooked as a means of retaining website visitors.
  • Groups of people acting as a trusted filter for information over the internet.
  • Tags (to give information structure) and ratings (to make better content stand out).
  • Real-time measurement systems rapidly giving feedback. Key is the immediacy of the information, and the ability to evolve the web service to reflect that.
  • Ability to make money from advertising, leads and e-commerce. While true since about 1995, the web user-base is now far larger, so the potential to leverage revenue also greater.

Metrics for Startups

I believe the ability to very accurately analyse website usage, implement changes, and then analyse the results, is a key advantage of web-based services. It is an advantage often overlooked by information technology professionals and programmers. I’m not sure why – possibly because web service developers:

  • don’t appreciate how hard/expensive gathering equivalent information is in other sectors of the economy, or
  • are scared to make changes in case they loose business, and/or believe their initial perception of what “works” to be optimum, or
  • just lack the pre-requite analytical curiosity to investigate?

Or perhaps Web 2.0 just isn’t mature enough yet for developers to have to worry too much about optimisation: A new concept for a site will probably either fail horribly or generate super-normal profits. The sector isn’t yet competing on very tight margins, where subtle optimisation can make or break profitability. Of course, optimisation of websites can deliver substantial changes in user behaviour. For example, I have found that a relatively subtle change to the position of an advert can alter the revenue generated by over 20%.

Dave McClure developed the AARRR model. AARRR segments the five stages of building a profitable user-base for a website:

  1. Acquisition – gaining new users from channels such as search or advertising.
  2. Activation – users’ first experience of the site: do they progress beyond the “landing page” they first see?
  3. Retention – do users come back?
  4. Referral – do users invite their friends to visit?
  5. Revenue – do all those users create a revenue stream?

For each stage, the site operator should analyse at least one metric. The table below gives some possible metrics for each stage, with a sample target conversion ratio (the proportion that reach that stage).

Category User Status (Test) Conversion Target %
Acquisition Visit Site – or landing page or external widget 100%
Doesn’t Abandon: Views 2+ pages, stays 10+ seconds, 2+ clicks 70%
Activation Happy 1st Visit: Views x pages, stays y seconds, z clicks 30%
Email/Blog/RSS/Widget Signup – anything that could lead to a repeat visit 5%
Account Signup – includes profile data 2%
Retention Email or RSS leading to clickthrough 3%
Repeat Visitor: 3+ visits in first 30 days 2%
Referral Refer 1+ users who visit the site 2%
Refer 1+ users who activate 1%
Revenue User generates minimum revenue 2%
User generates break-even revenue 1%

These metrics become critical to the design of the product. Poor activation conversion ratio? Work on the landing page(s): Guess at an improvement, test it out on the site, analyse the feedback, and iterate improvements. Gradually you’ll optimise performance of the site.

I find this attempt to structure analysis and relate it back to core business performance, very interesting. However, the sample metrics can be improved on a lot, depending on the nature of the site. For example, to track virality (referral), I might watch the monthly number of del.icio.us adds, or monitor the number of new links posted on forums (Google’s Webmaster tools allow that). Tracking users all the way through the tree from arrival to revenue generation needs to done pragmatically where revenue is generated from very infrequent “big-ticket” sales: With minimal day-to-day data, it can take a long time to determine whether a change genuinely has improved long-term revenue, or whether natural fluctuations in day-to-day earnings just contrived to make it a “good day/week/month”.

Now I know this approach works, but why it works is less clear. We might like to think that we are genuinely improving the user experience, and maybe we are. However, it could be argued that merely the act of change is perceived by users as an improvement – a variation of the Hawthorne effect. The counter argument to the Hawthorne effect can be seen on sites with low proportions of repeat visitors: The majority of those experiencing the improvement will not know what was implemented before.

History of Social Networking

Dave McClure’s interpretation of the timeline of the development of social networking sites is as interesting for what it includes, as for what it omits: No Geocities; no usenet; no forums; no MUDs… The following timeline shows key services in chronological order, except without dates – all the services shown were created within the last ten years:

  • Email lists (Yahoo Groups)
  • 1.0 Social Networks (Friendster) – these early network established the importance of up-time (service reliability) and the ability of users to manipulate pages.
  • Blogs – links between weblogs acting as networks.
  • Photos and video (Flickr, YouTube) – created a sense of community, and allowed tagging/grouping of content.
  • 2.0 Social Networks (LinkedIn)
  • Feeds and shared social information (Upcoming.com event planner)
  • Applications and widgets – the ability to embed data about a user’s friends in applications is probably “the most powerful change on the internet in the last ten years”.
  • Hosted platforms (OpenSocial, Facebook) – most services are likely to allow 3rd-party developers to provide applications on their platforms.
  • Vertical communities (Ning) – ultimately this may develop such that a service like Facebook acts as a repository for a user’s online identity, while specific groups of people gather on other networks.
  • Availability of information – a single sign-on, with automatic data transfer between services.

The future may be “Social Prediction Networks”. This is a variation on the theme of using trusted networks to filter content: Instead of Blogging meets Search, I characterise Social Prediction Networks as Digg meets Facebook. Shrewd observers will note Facebook has already implemented Digg-like features, while simultaneously topic-specific, community-orientated Digg-clones are being launched. People gather into interest groups around a topic, and then through use of tagging and rating, the community filters content. The system effectively predicts what other people in the group will find useful. This may be an optimum approach for groups above the Dunbar number (or an equivalent number representing the maximum number of people a person can form stable relationships with).

Interesting Aspects of Facebook

Three were discussed:

  1. Social graph (friend list) – email and SMS (mobile phone) service providers have rich data on the frequency of communication between people, yet aren’t using this information to form social networks. Dave noted that two major email service providers, Yahoo and AOL, are currently struggling to thrive – this could be an avenue for their future development.
  2. Shared social activity streams – knowledge of what your friends think is important. Friends are more likely to influence you than people you do not know.
  3. API/Platform – dynamic behaviour and links across your social network.

Further Observations

Will growth in social networks continue? Yes – the friend list adds value to the content.

Will others compete? Probably, as a “long-tail” of networks, likely topic-specific.

Can social networks be monetarized better? Currently social networking services generate far less revenue than search services. The challenge for social networking sites is to move towards the wealthy territory of search services. At the same time, search services are moving towards becoming more like social networking sites.

How can traditional companies engage with social networking sites? Social networking sites work best for sales where a product has a strong aspect of peer pressure in the decision to buy. The most important advice is not to create a copy of a website: Instead provide less complex content that uses social networks to draw users to a website.

Applications for social networks tend to be over-complicated, normally because programmers attempt to implement functions found in software they have previously written for other platforms or websites. Generally the successful applications are very simple. Some developers have opted to break complex applications into a series of smaller applications, and use the virality of social networking sites to build traffic for one application from another.

Social network applications are exceptionally viral. They can gain users very rapidly, yet also loose users just as fast. Much of this virality comes from feeds, which typically alert friends when a user installs an application. Within a few years the feed is likely to be based on actual usage of an application.

Facebook now allows applications to be added to “fan pages” (or product pages) – so individual users need not now be forced to install an application to use it.

Those using email lists for retention are best to focus on the title of the email, and not the content. Merely make it easy to find a URL in the content. The key decision for the reader is whether to open the email. What the email says is almost irrelevant – they’ve already decided to visit the site based on the title.