jump to navigation

Using Prosper and Mechanical Turk to figure out if people who are shifty look shifty March 19, 2009

Posted by jeremyliew in user generated content, web 2.0.

The Economist has another fascinating article about face. Specifically, about physiognomy – the idea that the way you look is a reflection of your character.

In particular, it describes research done by at Rice University to see if people could identify people who were bad credit risks by the way they look. They looked at 6,821 loan applications on Prosper. They asked 25 Mechanical Turk workers to assess each of the potential borrowers’s likelihood to repay a $100 loan. Here is what they found:

Their first finding was that the assessments of trustworthiness, and of likelihood to repay a loan, that were made by Mechanical Turk workers did indeed correlate with potential borrowers’ credit ratings based on their credit history. That continued to be so when the other variables, from beauty to race to obesity, were controlled for statistically. Shifty physiognomy, it seems, is independent of these things.

That shiftiness was also recognised by those whose money was actually at stake. People flagged as untrustworthy by the Mechanical Turks were less likely than others to be offered a loan at all. To have the same chance of getting one as those deemed most trustworthy they were required to pay an interest rate that was, on average, 1.82 percentage points higher, even when the effects of historical creditworthiness were statistically eliminated.

So it takes two web 2.0 services to tell you that many people who look shifty are indeed shifty.

Slides from Games 2.0 presentation at Web 2.0 Expo April 25, 2008

Posted by jeremyliew in gaming, social games, social gaming, web 2.0.

Some people have asked for the introductory slides from my talk today at Web 2.0 Expo about Games 2.0. These slides were just for scene setting and the bulk of the time was spent in panel discussion and case studies for the various elements outlined in the presentation. None the less, here it is.



Ben Vandgrift liveblogged the session. Since there is no way to link to just this post on his blog, I am quoting it here without any changes or edits (although note that my name is spelled Jeremy LIEW!)

{w2e fri session 1: games 2.0 (@11:56)}

jeremy lieu @ lightspeed venture partners

we’re talking about the future of game development and game types with web 2.0.


lots of things have changed in the web—mostly the variability of cost. these changes are tricking to the web industry.

aaa games can cost upwards of $30M. marketing driven by print and media advertising. building is about levels, and monetization is through box sales.

changes parallel the web industry.

small agile teams can get in-browser games out quickly.

(really want this slide .. don’t want to record it here.)

web games are multiplayer—not ai driven. no level design. &c.

some numbers:

halo3 $30M, yielded $700M.

powerchallenge < $8m. 1M players.

friends for sale < $1m. 7m players.


$30M marketing dollars, 10M copies sold, 90% through retail.

the rest—facebook.

aaa game sales declines over time.

social games are backward—they grow over time.

one of the critical issues for multiplayer games online is asynchronous play.

online gaming revenue: $3.8B 2006, $5.3B in 2007, more in 2008, and growing.


seqi chen (serious business)

johann christiansen (power challenge)

shervin pishevar (social games network)

mark pincus (zynger)

warbook—developed in two months by two people.

there have been improvements in play since launch, since it’s server hosted. just launched a sequel.

johann started in 2001, started text-based, added graphics over time.

how import are graphics, given their expense?

not important in management games.

launching early with a text-based game: profitable from the first year. management games develop a loyal user base.

variablized content?

mark: launched scramble as a live game, competition in real time. maxed out at 20K daily active users. spent three months reworking the game to be asynchronous. several tries before it reached a tipping point. thought asynchronous was a bad idea initially.

they come back to an asynchronous play because it’s your turn. there is a social obligation. the live game, though, represents 1/3 of daily users.

seqi: not turn based, but is asynchronous. (friends for sale). success about creating a game where social relationships are embedded into the game.

there is some value in unstructured asynchronicity.

johann: the social aspects of games are critical.

shervin: (gaming graph v. social graph). the emotional connection between people who know each other is quite valuable.

(think about a word scramble type game with continuous scoring.)

a very high accept rate on invitation is key to viral growth.

ghost racer: example of something too hard, yielding a high dropoff rate. ideally, a game should be easy, spreadable, with incentives for invitation.

six or seven months to 1M daily actives for popular web games. the important metrics are how things are being used. example: buying drinks for someone at the table at texas holdem. 250,000 drinks bought a day, for this little side feature.

(some discussion about funding of ventures.)

?: how to do something additional to viral.

mark: games are not naturally viral. other mechanisms have to be employed, and repeat usage has to be focused on. the ad rates you can buy traffic from are ridiculously low. networks from zynga are open to any new game developers.

shervin: ad-sharing is critical to cross-promotion. gaming bars, who’s playing what. one-click to go play. an independent channel from facebook. combined reach through SGN is 100M users.

seqi: (re crosspromotion) since they have one app, not so much cross-promotion. forced virality v. virality via game mechanic.

?: what are people willing to pay for?

digital goods, personal branding

?: real-money trading? digital goods?

shervin: a discussion on selling warbook gold. freegifts = largest virtual goods e-tailer.

mark: there is a direct line relationship between engagement and monetization.

?: social gaming and mobile phones?

they’re on the way.

?: $ for 1000 players / day? rate?

difficult answer. 4% monetized. few people getting north of 10%.

?: demographic information?

ffs: 60% female. 50% 20-25. 4th largest norway, big in saudi arabia.

power challenge: 90% male, 18-19 avg age.

texttwirl, etc: 60% women

warbook: 70% male

scramble: female and u.s.

texas holdem: make and foreign.

?:how important are incentives, like hint points?

word scavenger hunt.

Speaking on Games 2.0 at Web 2.0 Expo on Friday April 20, 2008

Posted by jeremyliew in gaming, social media, web 2.0.
Tags: , , ,

Apologies for having gone dark on the blog for the last two weeks – I was out on vacation. However, I’m back in the saddle now, in time for the Web 2.0 expo this week. I’ll be around the conference at Moscone West much of the week, and in particular am speaking on “Games 2.0” on Friday at 11am.

I’ll give some of my thoughts on how the games industry is undergoing the same forces of change that led to web 2.0, and discuss this with an illustrious panel of entrepreneurs working in this area: Siqi Chen of Serious Business (Friends for Sale), Mark Pincus of Zynga Games, Shervin Pishevar of Social Games Network and Johan Christensen of Power Challenge.

It should be a great session, both for people in the games industry, and for those in the web industry. There is much to learn from game mechanics that are more broadly applicable to social media sites. I hope to see you there!

Social Media: Culture = f(UI) December 19, 2007

Posted by jeremyliew in culture, facebook, game mechanics, interaction, Internet, myspace, social media, social networks, UI, web 2.0.

Lightspeed hosted a summit for portfolio companies and friends of the firm in the fall, focused on consumer internet user acquisition. One of the panels was about building community on a social media site, and on that panel Angelo Sotira (CEO and founder of deviantART) noted that for social media sites, culture is a function of UI. (deviantART is the leading community for artists and their fans on the web, and is an Alexa top 100 site. [Disclosure: my wife did some consulting for deviantArt.])

I was reminded strongly of this when reading Judith Donath‘s paper on Signals in Social Supernets that was published in the special theme issue of JCMC on social network sites guest edited by dana boyd and Nicole Ellison:

Variation in the design of SNSs promotes the development of different cultures (Donath & boyd, 2004; Lampe, Ellison, & Steinfield, 2007; Lenhart & Madden, 2007b). On a site where creating a link involves little cost, users may amass thousands of “friends,” but an observer has no way of knowing which, if any, of these links represent a relationship between people who care about or even know each other (boyd, 2006; Fono & Raynes-Goldie, 2007). On Orkut, for instance, one simply clicks on a profile to request a connection, and being connected provides no special access or information.

On sites with higher costs for creating a link, the observer has reason to believe that the links represent genuine relationships. Members of aSmallWorld are careful to request connections only with others whom they are sure wish to be linked to them, since they can be banished for having a few link requests declined (Price, 2006). On LiveJournal, making the link is easy: It is one of the few sites in which this can be done unilaterally. However, linking is generally done to give someone access to part of one’s journal, and linked members’ posts appear on one’s own space. This makes “friend” a relatively significant signal, as friending someone both reduces one’s privacy and publicly connects one with that person’s writing (Fono & Raynes-Goldie, 2007).

The meaning of these links is also personally subjective. For some people, listing someone as a “friend” on a social network site is an indication of personal and positive acquaintance. Others are far more casual, willing to add friends indiscriminately (boyd, 2006). This has ramifications for the reliability of the profile itself. Viewers may trust the self-created content of a profile if they believe that its links are to people who know that user well, while links that they believe have only minimal connection add little credence.

SNSs are designed for different audiences. LinkedIn is for professionals. It has no photographs, the profiles are resumés of education and work, and the comments are in the form of testimonials from co-workers. Identity is firmly tied to one’s professional self, and there is limited ability to explore other people’s networks. MySpace, popular with young people, has a very different atmosphere. Its profiles feature photographs, music, and embedded programs, and users can explore the network far beyond their own acquaintances (although they can choose to make their profile visible only to direct connections). This open interface makes it a rich environment for the jokes, links, and software that function as information fashions (discussed below).

Identity in MySpace is fluid. Some profiles are real people, presenting themselves much as they would offline. Some are commercial entities, such as bands, charitable organizations, or celebrities; still others are fictional personas, made for creative experimentation or as fronts for spam. No single design is ideal for all sites. What is important is that designers be fluent in not only the fonts and colors that make up the graphical design of the site, but in the social costs and benefits that shape its emerging culture.

Once a culture takes hold on a site, it is very hard to change. People building social media sites should be careful to think through the implications of their UI (including such mechanics as keeping score and exposing popularity) as their choices will likely have long term implications that can’t be easily reversed by a subsequent tweak to UI.

I’d love to hear other examples of sites where the dominant culture is a function of UI.

UPDATE: Bokardo has a good related post on how changes to Digg’s UI changed its culture

Games 2.0: Raph Koster at GDC Prime December 11, 2007

Posted by jeremyliew in asynchronous gaming, game mechanics, games, games 2.0, gaming, web 2.0.
add a comment

Last week Raph Koster was a last minute substitute speaker at GDC Prime. You can get a copy of his (long) presentation here.

Some choice quotes from his presentation – slide 93 (The game has changed):

The hot platform is the net
The hot audience is the non-gamer
The hot feature is other players
The hot technology is connectivity
The hot game is a mini-game

and slide 49 (Successful mass market interactive entertainment):

* Asynchronicity
* Indifference to rendering
* Minimal controls
* Platform agnostic

Model first – The system is the game
Universal inputs – Any button will do
Long phases – Take your time
Short decisions – Be done fast
Massively parallel – Side by side
Extended accumulated state – Save your profile
No roles – Classless
Representation agnostic – Draw it however
Open data – Change it however

For those of you interested in games 2.0, it is well worth reading the whole thing.

Games 2.0: Asynchronous gaming November 29, 2007

Posted by jeremyliew in asynchronous gaming, business models, casual games, facebook, game mechanics, games, games 2.0, gaming, user generated content, viral, viral marketing, web 2.0, widgets.

I am not a hard core gamer by background; more of a casual gamer. But casual gaming is now widespread; we’re all gamers now. My interest in the area has grown out of my interest in social networks and social media. I’ve long noted the increasing application of simple game mechanics to social web sites and how this can meaningfully increase the levels and types of interactions that users have with each other and the site.

As an investor in Flixster and Rockyou, both highly viral Facebook and Myspace “app” and “widget” makers, I’ve been tracking closely the spread of emergent user behaviors in these social networks. One Facebook app that really caught my attention is Scrabulous, an online scrabble game that can be played asynchronously, ie players don’t all have to be online at the same time.

Online multiplayer games have long been popular at all the big casual games portals. Multiplayer gaming can be viewed as user generated content for games, one of the drivers of Games 2.0. These games typically have a “lobby” where players can meet and match up before entering into a game against each other in real time.

Making the gameplay asynchronous fits better with the “continuous partial attention” world that we increasingly live in. The reason I never became a hard core gamer is that the serial monogamy requirements (one game at a time, total dedication, long periods of gameplay coordinated with others) doesn’t mesh well with my lifestyle. Scrabulous is a better match for the “play a little bit when you have some time, at various points throughout the day” life that many of us lead. Single player casual gaming (whether Bejeweled online or Brickbreaker on the Blackberry) has been filling that need for many players. These are fun, and at least have the “high score” dynamic, but they lack the social aspect that turn based asynchronous games offer.

Asynchronous games also make it easier to play against friends. You don’t have to coordinate to be online at the same time. Playing friends makes games more fun, and gives them a social aspect (the games have context if you have an ongoing relationship with an opponent). Playing with friends also offers an opportunity for true viral growth for the game, as players invite their friends to play.

Although these turn based multi-player games (especially those derived from boardgames) have some social dynamic, they lack the breadth of social interaction of synchronous MMOGs (not just the direct social interaction, but also the perfomative aspects of gameplay) that help make them such compelling experiences. Part of the appeal of MMOGs (whether World of Warcraft or Puzzle Pirates) is knowing that you’re “in game” with thousands of other people at the same time, each of them interacting with the same universe that you are.

So what would an asynchronous massively multi-player game look like? It can’t be turn based because most players would spend most of their time waiting for someone else to move. That’s not fun. It would have to be time based instead. Players would need to make their moves against a real world clock. Games like Duels.com (swords and sandals PvP fighting game), Manager Zone (soccer manager game) and Kings of Chaos (real time strategy game) all employ this dynamic. Massively multi-player games offer even more opportunity for viral growth because a players invitation ability is unbounded by the number of seats at a board game.

This led to me think about games using the framework below:

narrow games framework

I think that we’ll see a lot more innovation in the two sections of asynchronous multiplayer and massively multiplayer games over the next few years. I’m actively interested in investing in these areas. What are the most interesting such games that you see?

Web 2.0 has been driven by variablization November 26, 2007

Posted by jeremyliew in advertising, business models, distribution, platforms, viral, viral marketing, web 2.0.

The last couple of years have seen an explosion of innovation on the web that has broadly been labeled Web 2.0. There has been a lot of debate about what exactly constitutes web 2.0 but what hasn’t received as much attention has been what changes have enabled these Web 2.0 companies to arise – what is different now from the mid 90s and Web 1.0. I think the change can be summarized in one (somewhat clumsy) word: Variablization.

Variablized Development Costs

In the 90s we used software development models, primarily waterfall models, where a usable product wasn’t available until close to the end of the development period. Most code was written from scratch, with little reuse or public domain code, and large teams were necessary.

With the popularization of agile programming methodologies, widespread use of open source software, greater ability to use offshore development resources on a consulting basis, and a culture of “open beta”, the costs of developing a website or web service have become both lower and more variable. Ideas that look promising but fail to capture user interest in beta can be identified much earlier and at much lower cost, and resources can be shifted to more promising avenues.

Variablized Content Costs

In the 90s almost all content was created by professional editors and writers, employed by companies. To launch, they had to create a critical mass of content, which cost a certain amount.

Recently, with lower expectations out of beta products, the widespread adoption of user generated content and emerging best practices in how to use user generated content, the costs of content creation have dropped dramatically and become variable.

Variablized Marketing Costs

In the 90s, there were only two ways to get a large number of users. The first was offline marketing – the famous Pets.com superbowl ad approach. Expensive, and with a high minimum level of spend required to break through the clutter. We all know how that worked out.

Overture and Google have changed that landscape. Their CPC model means that you can spend as much as you choose to gain new users, and that your marketing spend can be completely variable.

Variablized Distribution Costs

The second way to get a large number of users in the 90s was to get a distribution deal with one of the big portals – AOL, Yahoo or MSN. In those days, this was the only way to reach a large number of internet users effectively, and you typically had to sign up for a multi year, multi million dollar deal to do it.

As social network platforms open up, and as the basic principles of viral marketing become better known, distribution has become variable, if not free.

Variablized Monetization

The vast majority of Web 2.0 companies rely on advertising as their business model. I think this is because advertising is the one business model that has become variable (relative to the 90s). Back then, to sell online advertising, you both needed to have substantial scale, and you needed to have your own sales force.

Today, thanks to ad networks and CPC contextual targeting (not just Google’s adsense, but also Quigo, Yahoo’s Publisher Network and others), even the smallest of websites can start earning advertising revenue.

There have not been equivalent innovations for subscription and ecommerce business models, and as a result, we’ve seen far fewer web 2.0 companies that use those models.


These changes in cost structure are a useful lens through which to view the current startup environment. It’s been said before that it is cheaper to build a company than ever before. While that is partially true, it is not the whole story. Digg has raised over $10m, Youtube over $12m, Photobucket and Rockyou (a Lightspeed company) over $15m, and Facebook has raised over $275m. (With the exception of Facebook) while these are lower than the amounts raised by companies in the 90s, they are still large numbers. Variablization of costs only makes costs go away when usage is low. In other words, while it still takes money to succeed, it is cheaper to fail than ever before.

Luckily for VCs like me, that means that successful companies will still need to raise money!

Social Design Best Practices November 5, 2007

Posted by jeremyliew in business models, facebook, game mechanics, google, myspace, open social, product management, social media, social networks, viral, viral marketing, web 2.0, web design.
add a comment

Bokardo notes a set of social design best practices as recommended by the Google OpenSocial team:

1. Engage Quickly – (my interpretation: provide value within 30 seconds)
2. Mimic Look and Feel – (make your widget look like the page it is in)
3. Enable Self Expression – (let people personalize their widgets)
4. Make it Dynamic – (keep showing new stuff)
5. Expose Friend Activity – (show what friends are doing)
6. Browse the Graph – (let people explore their friends and friends of friends)
7. Drive Communication – (provide commenting features)
8. Build Communities – (expose different axes of similarity)
9. Solve Real World Tasks – (leverage people’s social connections to solve real problems)

Worth reading the full text from OpenSocial

Three ways that a conference lobby is like Facebook October 21, 2007

Posted by jeremyliew in communication, Consumer internet, performance, social networks, web 2.0.

I spent three days last week at the Web 2.0 Summit, mostly in the lobby of the Palace hotel. The lobby served as the crossroads for the conference; all attendees passed through there and many never seemed to leave it! It was a great venue to catch up with friends and industry contacts among the attendees and lobbyconners.

It struck me that the conference lobby was like a social network in three ways:

Public Communication as Performance

At Web 2.0, if you wanted to have a private conversation, you would leave the lobby and find some place more private. In a social network, if you wanted to have a private conversation you would send a private message. But if you were OK with others seeing your conversation, you would stay in the lobby, or post a public message on the Wall/Comments. The Performance aspect of communication is seen both online and offline.

Serendipitous communication

In ordinary life, you communicate with far fewer people than you’d like to. You forget, you get busy, and you don’t reach out to people that you’d like to talk to more often. But in the lobby of a conference, you’re always accidentally running into people that you’d love to talk to but don’t usually see. This is one of the biggest benefits of conferences.

Similarly, social networks bring up opportunities to communicate with people that you may not have connected with in a while. Perhaps you see one of their comments posted on a friend’s MySpace page, or you get an update on them from the Facebook feed, and are prompted to ping them. I’ve been communicating more regularly with ex colleagues and extended family because of Facebook.

Lightweight Interactions

Over the course of two days at a conference you’ll see the same people a number of times. After you’ve talked, there is only so much you can say the next time, so your interactions tend to get lighter weight. You want to acknowledge each other but not necessarily get involved in a long conversation. So you smile, shake hands, clap shoulders, bump fists, wink, wave, or kiss cheeks (gender specific!) instead. It is the same rationale that leads you to text a friend instead of call.

Social networks provide similar lightweight opportunities for interaction. Facebook’s poke is the simplest example. Although Kara Swisher thinks that many Facebook apps are childish, I think they are providing an avenue for lightweight interactions between friends. Whether you’re buying someone a drink, biting them to turn them into a zombie, hugging, slapping or tickling them, the subtext of “I’m thinking of you” is there.


People building social media companies and other companies that require user interaction should bear these examples in mind. It is hard to create new mental models of behavior for users. As always, if there is an offline parallel for the online behavior you want from your users, you’re more likely to succeed. These three elements of social network behavior have clear offline parallels.

Wisdom of Crowd or Crowdiness of Crowds II October 9, 2007

Posted by jeremyliew in attention, Consumer internet, social media, social networks, user generated content, web 2.0.

In May I posted about a NY Times article that showed that making popularity data public made hits bigger and that talent was only one factor in this equation – the taste of the early adopters was more significant.

A recent Wharton research paper comes to a similar conclusion. Paid Content summarizes the results:

— “One, some common recommenders lead to a net reduction in average sales diversity. Because common recommenders (e.g., collaborative filters) recommend products based on sales and ratings, they cannot recommend products with limited historical data, even if they would be rated favorably. In turn, these recommenders can create a rich-get-richer effect for popular products and vice-versa for unpopular ones. This finding is often surprising to consumers who express that recommendations have helped them discover new products.
— In line with this, result two shows it is possible for individual-level diversity to increase but aggregate diversity to decrease; recommenders can push each person to new products, but they often push us toward the same new products.
— Result three finds that recommenders intensify the effects of chance events on market outcomes. At the product level, recommenders can ‘create hits’ out of products with early, high sales due to chance alone. At the market level, in individual sample paths it is possible to observe more diversity, even though on average diversity often decreases.
— Four, we show how basic design choices affect the outcome. Thus, managers can choose recommender designs that are more consistent with their sales or product assortment strategies.”

These are largely consistent with my conclusions from May for people who run social media sites:

1. If you’re trying to iterate towards a “best answer” then keep feedback loops to a minimum, at least before users “vote” on their own. (e.g. Hotornot, espgame)
2. If you’re trying to create “hits” out of some of your content (and don’t care if it’s the “most worthy” content – you only care that they are hits), then display feedback and popularity constantly, as this will effect user behavior and exacerbate the size of the hits (e.g. Youtube, Digg, American Idol?
3. If you want to “guide” user behavior in a certain direction, provide feedback that validates or shows the popularity of that behavior. This is consistent with my prior post on game mechanics applied to social media: keeping score.