Tal Zarsky, an Information Society Fellow at Yale Law School is doing a dissertation and paper on data mining and data privacy, and part of this project involves virtual worlds/MMOGs. An initial discussion of his work is available here.
He wanted to follow up on the comments and asked the following about datamining in VWs:
"First I want to thank you all for taking interest in my project.
Also, I want you all to perhaps think of the following scenario: The “gods” do not share the information with others – nor do they try and connect the virtual identities to the real ones. Nevertheless, they still can gain from their omnipresence and constant surveillance of the virtual identities. Here are some ideas:
(1) The “gods” provide specially tailored ads to every player (Adaboy strikes back?) – thus taking the practices of DoubleClick to another level. Here we have to think what the possible detriments are: providing advertisers with greater powers of persuasion? Being put in a “virtual Skinner box”? Having others decide for you what you should consume?
(2) The “gods” use the information they collect and data mining analysis (data mining is a personal favorite of mine, and the main theme in my dissertation) to provide different goods to different players at different prices. These goods could be game related (different playing plans) or non-game related.
What do you think? Am I too pessimistic here? I was wondering if you think the commercial entities will have the ability to muscle their way into the virtual worlds - perhaps by providing 2 pricing schemes for every game - one that will be "ad free" and the other with ads. Since it is clear that we all have a myopia problem when it comes to privacy rights - such a strategy should be fruitful for commercial entities."
Data mining is indeed being performed by virtual world operators at the moment, but not yet to the degree envisaged by Tal. In his 2003 paper "Better Game Design through Data Mining" (http://www.gamasutra.com/features/20030815/kennerly_01.shtml), David Kennerly gives 4 reasons why it is done:
1. To balance the economy
2. To catch cheaters
3. To cut production costs
4. To increase customer renewal
Only the final one falls in the area that concerns Tal.
As for the two specific ideas that Tal asks about:
Tal>(1) The “gods” provide specially tailored ads to every player (Adaboy strikes back?)
This would deeply worry me if the ads were not in context. For a virtual world where reality is already embraced (eg. There, Second Life), it wouldn't be any more of a problem than any other RL data-mining process. For ads that concerned in-world products, promoted using systems that made sense in the virtual world, the same applies. However, when ads for RL products appear in a VW where those products are meaningless (fizzy drinks in a medieval setting), they could seriously break immersion. This applies to all advertising, though, whether data-mined or not.
Of course, if the data-mining is exceptionally good it will be able to tell from a player's playing style whether or not they are likely to be annoyed by the ads, and therefore not advertise to people who fit such a profile. As if...
Tal>(2) The “gods” use the information they collect and data mining analysis ... to provide different goods to different players at different prices.
Players would soon find out that other players were getting a better deal, and would want to know why.
It's fair enough to use data-mining to direct advertising only towards people who will respond to it (so you don't spam the rest), but if you do it too narrowly then that can be almost as bad as if you do it too broadly.
Richard
Posted by: Richard Bartle | Jan 06, 2004 at 05:45
Data Mining and the subsequent analysis is also an excellent way to better understand your players/customers and their behaviors in VWs. Moreover, in VWs that are close approximates of RL (There, Second Life, etc.) the “gods” can unobtrusively collect much more data about individual behavior than any current RL methods AFAIK. For example, is blue Nikes more popular than red? With proper notice, regulatory compliance, and sharing of the aggregate data to the community, VWs operators have a new revenue stream.
But the idea of raw commercialism in VW worries me. Someone probably already explained why in-game span or a 3D Virtual eBay is a bad idea, so I won’t rant.
However, as Richard has indicated, Data Mining and understanding of player preferences and behaviors can help any VW operators to improve the VW experience.
Frank
Posted by: magicback | Jan 06, 2004 at 10:16
Just to be clear, There.com has invited Nike and Levis in while Second Life has not. Second Life allows its users to retain rights to their creations and to make both real and virtual money. Users are on a level playing field and are competing with each other NOT with Linden Lab (the creator of Second Life).
On the other hand, in There.com Levis and Nike are in a privileged position relative to the users since Levis and Nike retain their IP rights and presumably have access to demographic data while users are forces to grant licenses to There.com and don't get to look at the underlying data.
Cory
Posted by: Cory Ondrejka | Jan 06, 2004 at 12:54
I'd be interested if Tal could unpack a little bit the notion that "commercial entities will have the ability to muscle their way into the virtual worlds"?
Virtual worlds (with the exception of some MUDs) are commercial entities. I do get the concern over player privacy (which is currently handled by our old friend the EULA), but I don't get the concern over advertisements.
If Sony wanted to, it could put up ads for its content wares in EQ and SWG, just as TSO did with certain entities. Sony apparently doesn't want to do this, because, as Richard says, it would ruin immersion.
Posted by: Greg Lastowka | Jan 06, 2004 at 13:36
I agree with the others that the two suggestions made by Tal are antithetical to the idea of most virtual worlds. To put it in marketing terms one of the USPs of most virtual worlds is the immersive experience created by the world – creating this experience take time, money, expertise and more money. The reward is revenue stream built on people wanting to experience the environment and supported by networking effects of the social group that it is a nexus for. Using data generated in that environment to do things that would disrupt the environment seems commercially stupid.
OK there is a risk reward but the reward from putting ad into most of these worlds given the rate or ad click through revenue or (where ads are for the same companies products) the chance of creating positive brand awareness seem negligible.
Any brand relationship especially one related to a service is a trust relationship. The idea one is a long term relationship with strategically valuable customers. Given what MMOs (in the main are) the last thing they want to do is start to disrupt that trust.
‘There’ is different, so possibly is The Sims Online. These are much more analogues of our commercial world, to some degree it would be odd not to see advertising in these games, just as it would in say a contemporary football sim.
Which brings me to data mining - traditionally data mining is not (or was not) the type of application outlined above. That is much more CRM data analysis. What I understand data mining to be is the technique of taking large amounts of data and running analysis against those data to _discover_ relationships.
The intention is not at all to relate any fact to any individual, that’s quite beside the point. The motivation is to discover patterns that may be useful to you.
The FMGC sector is the classic users of data mining. The famous examples are of shopping patterns where they discover things like x% of people that by product A also buy product B – if I remember correctly nappies and beer was one that actually hit the news in the UK. With this kind of data you can do all kinds of things such as place nappies next to beer in your shop, or, coming to the point at last, sell this information to nappy and beer makers so that they can think of things like cross promotional opportunities.
evil_marketing_ren: so what you do in an MMO is much more subtle and insidious. You sell data mining analysis to companies that sell to your target demographic and you don’t tell your players coz its not about _them_ really is it, its about nodes and clusters and statitical fits and things, not about people.
Ren
www.renreynolds.com
Posted by: Ren | Jan 06, 2004 at 15:46
evil_marketting_ren: "You sell data mining analysis to companies that sell to your target demographic and you don’t tell your players coz its not about _them_ really is it, its about nodes and clusters and statitical fits and things, not about people."
I really don't have much of a problem with this. If anything, it is the opposite. I *want* better directed ads. Sufficiently well directed ads stop being ads, and start being helpful news items. If Blizzard releases a new game, it isn't an Ad when they tell me of this. It is a news item that I wanted to hear about.
My real concern over data mining isn't that it will allow the marketers to better market. It is that it said marketers will entirely fail to understand the results of the data mining, and market in an even more obnoxious and annoying fashion.
- Brask Mumei
Posted by: Brask Mumei | Jan 06, 2004 at 17:40
Brask> I really don't have much of a problem with this. If anything, it is the opposite. I *want* better directed ads.
One of the points I was trying to make in the preceeding section was that this has nothing to do with directed ads.
ren
www.renreynolds.com
Posted by: ren | Jan 06, 2004 at 17:54
This topic is so broad that I could write for days on just two or three of the 20 topics that have been raised here. While I just don’t have the time to give even one of these topics its due attention, I can’t help responding even briefly to several.
There.com may be unique in that we treat much of our data similar to how a Fortune 500 company would treat any data that they have access to. My guess is that going forward almost all product development proposals will be supported by multiple data reports and constant review of goals and expectations for any development project.
Data Mining for Product Development:
Just to give a simple example, it’s easy enough to use data reporting on how members are using vehicles. In one of the rounds of product creation, we found that Buggies were very popular. The problem then becomes finding out why they are popular. Do they like racing? The skills ladder? Is it cheaper than other vehicles? Are there more places to use it? Are there more events with buggies? Etc, etc. While there were probably 10-20 reasons why we thought they might be popular after running several reports one pint that stood out was that we found out that member would often use them with friends as they were the only vehicle that had dual seats. This lead to the creation of the 5-seat Hoverboat, which is one of the top revenue items today despite its $25.75 (US$) price tag. Going forward, I think there are 20 or so key features in a product like a vehicle that can be measured, analyzed and tuned, and clearly Data mining will be the way that this information is collected in the future.
Advertising:
I think at many levels members expect, if not demand, that basic data mining techniques are employed with some forms of advertising in VWs. Members often have very different levels of passion for the many aspects of any virtual world. While almost everyone wants to know everything about something, rarely does everyone want to know everything about everything. So, data mining becomes very important in filtering the information that we bombard members with.
For example; If we know that members might read one email every 3 days, then it’s in both our interest and the interest of the member that we do our homework on what that email might include. Developers want to know about changes in developer tools/pricing, sport racers want to know about changes in racing, club leaders want to know what changes are in store for clubs and club houses. And God forbid you ever send an email written for a Newbie to a veteran. But clearly, these concepts are applicable to not only emails but also almost any type of advertising going forward.
Comments:
Richard: It's fair enough to use data-mining to direct advertising only toward people who will respond to it (so you don't spam the rest), but if you do it too narrowly then that can be almost as bad as if you do it too broadly.
Absolutely!
Frank: (There, Second Life, etc.) the “gods” can unobtrusively collect much more data about individual behavior than any current RL methods AFAIK.
Yes, and its also possible for VW developers to construct an infrastructure where members are able to collect much more data that has ever been imagined.
www.ThereDaq.com is a site that a single member created that gathers information from our in-world auctions. By using this information member developers know what products are hot, which have the best profit margins, how the market compares today to yesterday, etc, etc. Not to mention that this data is all live and current. Without a doubt, members now have access to data that would make any RL marketing manager feel like a kid in data heaven.
Cory: In There.com Levis and Nike are in a privileged position relative to the users since Levis and Nike retain their IP rights and presumably have access to demographic data while users are forces to grant licenses to There.com.
So, I guess this legal mumbo jumbo is a bit confusing. But, just to be clear, Levi’s and Nike actually don’t have any legal advantages over our member developers. Yes, they retain their IP, so do our member developers. Yes, member developers grant us license to use their IP, but so does Nike and Levi’s. It’s been this way since Beta 2 back in April.
The advantage that is available to Nike and others is that they can purchase abilities that we set aside for them. For example, in the case of Nike the shoes actually let you run faster than member developer shoes. In the case of Levi’s the texture on the jeans is of higher quality than we normally allow.
Ren: ‘There’ is different, so possibly is The Sims Online. These are much more analogues of our commercial world, to some degree it would be odd not to see advertising in these games, just as it would in say a contemporary football sim.
Absolutely. Most of the feedback we have from members is that they would like to see more ‘Brands’ in-world, not less. My guess is that most of our target market is already very accustom to expressing their persona in RL through the use of Branded goods.
One of my favorite requests, that we often hear, is that members would like copies of RL goods in There when they make purchases in RL. For example, you go out and buy a Tommy Hilfiger shirt. The shirt comes with a code that you can enter in There to get free virtual copy of the shirt. You now have a new shirt that you can show off to your friends in There, and talk about. This would be very simple to do and imho would add to our unique virtual experience, not detract.
-Bruce
Posted by: Bruce Boston | Jan 06, 2004 at 19:16
Bruce> One of my favorite requests, that we often hear, is that members would like copies of RL goods in There when they make purchases in RL. For example, you go out and buy a Tommy Hilfiger shirt. The shirt comes with a code that you can enter in There to get free virtual copy of the shirt. You now have a new shirt that you can show off to your friends in There, and talk about. This would be very simple to do and imho would add to our unique virtual experience, not detract.
Excellent, further integration between RL and There will be much more closer such that RL behavioral interactions can take place.
And I can show off my new Armani Suit... or can I buy a Sword of Death +5 in VW and get a RL replica of it? What are the price differentials?
Frank
Frank
Posted by: magicback | Jan 06, 2004 at 20:30
Cory Ondrejka>Just to be clear, There.com has invited Nike and Levis in while Second Life has not.
You wouldn't have any objections, however, if the good people of Nike were to play Second Life and create their own versions of their sportswear?
Richard
Posted by: Richard Bartle | Jan 07, 2004 at 03:31
Regarding my fansite ThereDAQ.com where I produce reports on all the There.com member-auctions, I have had numerous requests from players to sell them data on both their own and others' auctions, products and sales, plus analyses of same.
However, I've so far only supplied data to designers for their own products and auctions, because there is no consensus as to who owns the data (which is publically available to There members). There would however be outrage among the members if I were to sell the sales information of the top designers to third parties.
A friend of mine is providing consultancy to There's member designers using a feed from my auction database, but the market opportunities for this data appear limited unless we were to be granted rights by There to use/sell the data as we saw fit, which is extremely unlikely.
Posted by: Tessa Lowe | Jan 07, 2004 at 17:40
I have been following this intriguing exchange and have a few responses/remarks.
I liked the points re: using data mining for product development – but isn’t that a step away from marketing and the use of this personal information (even though it only pertains to the avatar) to put sophisticated schemes of price discrimination in place? For example – if the “gods” see from the patterns of consumption that specific forms of users are easier spenders than others – they will offer them the products (which could relate directly to the game to answer the immersion question) at lower prices.
I saw that many of you are skeptical about this happening. You believe that an arbitrage market will emerge and render these practices impractical. But, how will this information travel? Why is this world any different than other places where personal information about users facilitates such pricing (whether this is problematic or note is another issue to discuss)?
Here I also want to note the earlier comment re: Data Mining – data mining is indeed best used to pick up on hidden patterns and clusters in information. When applied to a world which is already using such various schemes – these applications will be very efficient in picking out the instances in which the operators could make most money. It will also find the places where brands are constructed in the most efficient way and where people are buying into new notions.
I acknowledge the “immersion” problem as a major barrier to any of these schemes. However, this could be overcome – first as many of you mentioned in games that resemble the “physical” setting. In addition, the “gods” have other ways to access users outside the actual game space. Finally, there is the use of the analysis of personal information (which is not personal per se as it only pertains to the avatar) to market products that involve the game.
Finally, immersion sounds nice – however how about games that will provide 2 pricing schemes – one that allows real world advertisers (and a discounted price) and one without? Wouldn’t the discounted game attract many players and thus lead to the problems I address?
Tal
Posted by: Tal Zarsky | Jan 08, 2004 at 12:59
>Finally, there is the use of the analysis of personal information (which is not personal per se as it only pertains to the avatar) to market products that involve the game.
Noooo. In many respects my avatar is me. This is one of those repects.
>Finally, immersion sounds nice – however how about games that will provide 2 pricing schemes – one that allows real world advertisers (and a discounted price) and one without? Wouldn’t the discounted game attract many players and thus lead to the problems I address?
This assumes certain things about price elasticity. That is, it assumes that a significant barrier to sale or more particularly on-going participation is the purchase price. This may be the case. I doubt it. I think that time requirement is a much more significant issue. Hence I don’t think that price variation within the type of limit that you would imagine would be sustainable though add sales would have a significant penetration effect.
Also I think that the impact would be more complex than this. There is a single community with the game, I feel that having one set of people that have ads in the interface and one that does would have a disruptive effect on the community which is one of the hardest things to sustain in the game.
Ren
www.renreynolds.com
Posted by: Ren | Jan 08, 2004 at 13:21
Ren>Also I think that the impact would be more complex than this. There is a single community with the game, I feel that having one set of people that have ads in the interface and one that does would have a disruptive effect on the community which is one of the hardest things to sustain in the game
But isn't that the beauty of the Internet media - that is allows tailoring of messages so that others do not have to be aware what you receive- meaning that others could have this mixed world of ads, manipulation and data mining and you will be unaware of the fact when you interact with them (as they will have a different screen - yet will interact with you at the same time and in same way).
Note re Avatar info and personal info: I agree with Ren's position (that they are same). However, as the discussion re:personal info laws and avatars has shown, that is not the way most people will view this. My point is that the data mining of avatar info alone could lead to interesting and perhaps even disturbing results. The key question of course, is how.
Tal
Posted by: Tal Zarsky | Jan 08, 2004 at 13:47