Research as Platform Specific: Avatar-Based Marketing

Part two of three-part Interview with Market Truths’ Mary Ellen Gordon, managing director of the award-winning VW marketing research group. For summary and part one, click here.

PGL: Is there any clear or significant difference between this kind of research in real life (RL) and Second Life (SL)?

G: Yes. We’ve thought a lot about what works in world and what doesn’t. Clients ask, Why not do the surveys in world? The answer is because surveys are not a good format for live interaction. For examples, I cannot show up at Second Fest [an in-world music festival] to ask them a bunch of questions. From a technical standpoint, computing resources are limited. Also, you don’t want to disrupt the experience. Go out for quantitative and in for qualitative. We had to change the way we approached it from a traditional market research perspective. We had to be more approachable and fun and informal. If you did a lot of the traditional method people would think you are odd. “You sound like a recording.” You need to keep the same standards or rigor and reliability, yet not sound stilted in-world. When we end SL focus groups people often say, “That was fun.”

I should have mentioned that another factor in our decision to go out to the Web for most quantitative research is confidentiality. For in-world qualitative stuff, we close down the area within 100 meters of where we are collecting data so we are out of (voice or text) hearing range of anyone who might be in the area and don’t allow other people to run scripts to prevent spy scripts. Even so the Linden Labs Terms of Service agreement (TOS) mean that technically speaking they can access any information they like, so going out to the Web when we’re collecting quantitative information just makes it that much more secure.

In focus groups, the real difference is that in RL you want to keep one conversation going. What we found about SL focus groups, of which we conduct 90% in text [as opposed to VOIP], is that you have a whole bunch of conversations going on at once. One tool we use a lot to give some order is a scripted object that is essentially a ‘raise hand’ indicator that gives feed back about who said what. Then the moderator might start with the three people out the group who said ‘no’ to the questions rather than the seven who said ‘yes.’ [In this example people are using their whole body to talk.] You need to practice in changing your expectation of being in control of everything going on. Also, you can read the transcript in a case you did not catch everything. Or have them fill in note card so you have several different kinds of communication happening at once.

The SL focus groups are strangely self-policing. In a RL focus group or a class for instance you often have one person dominating conversation and the moderator or professor needs to intervene somehow. What happens in SL is that the other participants will moderate. For example, we had one participant in a focus groups on in-world entertainment for whom the answer to every question was ‘Beyoncé.’ After a while, it was the other participants who said, ‘OK, enough with Beyoncé.’

PGL: The SL population seems more keyed in, more active?

G: Pretty much, yes. The SL population is extremely engaged as participants in the research, which makes their responses so much more useful. As a group they are smart and switched on. Who ever is still there are 30 days has gotten past the clunky initiation process. They are more outspoken. When they are engaged they will give thoughtful answers. I have found the most thoughtful responses to open-ended survey questions.

The other thing about this is that they [participants] are really engaged in SL, so if the topic is at all SL related (or we can at least create some sort of SL link), then they really have a lot of things they want to say.

PGL: For Market Truths as a company, do you see yourselves as survey avatars or their RL users? I ask this question in light of Paul Hemp’s article in Harvard Business Review, “Avatar-Based Marketing.”

G: There is way less separation between the two. [Hemp’s article is] speculation. All the empirical evidence that we have points in the other different direction.

PGL: What happens when critics say, ‘You’re not even talking to real people.’

G: What we’ve done is ask questions about both [avatars and players]. It is important to be really clear about whom we are talking about and even among avatars, since many have alts, which avatar.

One of out first MT projects was on the women’s clothing market in SL. We looked at personal style in RL and personal style in SL. What we found it that if you see an avatar dressing conservatively in SL, then 2:1 that person dresses conservatively in RL. The same for avatars who dress hippy or arty, it’s a 2:1 SL to RL. The third category where there was that strong correspondence was casual clothing– people who dress casually in RL also tend to in SL and vice versa. There is a correspondence with risqué clothing too, but it is not as strong. The aspects of dress and appearance that tend to be skewed are things such as size and ages. In general, dressing in SL tends more toward the risqué side of spectrum in general in comparison to real life. When we first when in (2006), you could only find ball gowns or mini skirts. We did observational research from hundreds of clothing stores with thousands of items. When we did the fashion report, we found that what many people wanted to wear were business suits. While business suits were the most common answer to a question about types of clothing respondents wanted but had been unable to find, there were also lots of requests for other more conservative and professional clothing.

Just to clarify, those ratios are in relation to chance – so you’re twice as likely to be able to correctly predict if someone dresses a particular way in RL by knowing how they dress in SL than you would be if you just guessed based on the overall proportion of the population that dresses that way.

I should have also mentioned that we’ve also found links between RL and SL behavior in entertainment (e.g., people drawn to more social forms of entertainment in RL tend to be in SL too; same for people who are drawn to more solitary and discovery-oriented forms of entertainment), real estate (RL real estate experiences influence in world real estate preferences), and brands (different attitudes toward brand activities in world depending on RL brand attitudes). We find the same sorts of patterns in our proprietary customer research.

PGL: What indicates to MT that virtual worlds (or is it only SL) merit their own market survey?

G: We are interested in virtual worlds [VWs] generally. I see this as the way that media is going. It feels like the Web did in early ‘90s––who is using VWs, the penetration. When I first heard about SL from David Rowe, the other MT director who is a technical guy, I spent the whole weekend there. The first time I saw a Web browser it was Mosaic, same experience. Everything is going to go this way. If virtual worlds are going to be where the Web is now in ten years, then a group like ours can potentially help determine how things will work. Early on, MT did Web surveys and we did not like the way it was done and did not trust results.

The other thing I should mention here is that VWs are also just a really useful venue for collecting information. As described in the discussion of focus groups, people tend to give very thoughtful answers, and VWs are also a very good way of reaching particular types of customers. For example, we have done research for a high tech company that wanted to use SL to collect data because it was a good way to tap into ideas from tech savvy people. We can also have focus groups with people from different countries, different states, etc. and manipulate things in 3D. That’s much better than face-to-face focus groups (which are limited to people in a single location) or online focus groups that don’t offer the same degree of immersion and ability to manipulate objects in real time in 3D.

We still do “regular” Web surveys, and do trust them when we have something like a customer list to work from so we’re surveying a list of people we know care about whatever it is we’re asking about, but it’s those surveys that rely on sampling from online panels that I’m much more cautious about for the reasons described previously.

In terms of the psychographic profiles of users, SL is over represented in creative, marketing, and technical fields. We hope to investigate during 2008 the profiles of users on other platforms, such as There.com and Kaneva.