• by Steve Smith , Staff Writer @popeyesm, October 24, 2008

Clothing sizes for women are notoriously, maddeningly arbitrary. Two years ago, myShape.com launched its personalized shopping service for women that matches clothing to very specific shapes and sizes. A patented matching engine literally shows a member only garments that fit and flatter them. Several months ago, the site added to the personalized shopping service a recommendation engines from Baynote that layers into the member's view items that other viewers of specific products also considered. According to Baynote and myShape, the result was a 16% lift in average order value, a 30% rise in revenue and a 200% lift in conversions.

We spoke with myShape's Mercedes De Luca, customer experience and chief information officer, about layering recommendations with personalization -- and the kinds of novel services that might evolve from this combination.

BI: What is that matching technology?

De Luca: First we match on measurements. Women give us their measurements when they become members and have personal shopping on our site. Those measurements allow us to also assign a shape, one of seven shapes. Then there are "style preferences," which is what women tell us about themselves and their style. Some women only like to wear cotton fabrics and some don't want to wear skirt length below the knee. So we take measurements, the shape, and style preferences and we put those all together into our secret black box and we match them to the garments. What happens as a result is that women will only see the things that will fit and flatter them in the personal shop.

BI: And you know the precise measurements of garments rather than just the label sizes.

De Luca: We have been able to make partnerships or have agreements with the vendors on getting the information we need. We have a very extensive intellectual property and proprietary system that has all of this information in it. Many manufacturers don't operate sites and can't get into retail stores. We offer them a perfect opportunity for them to be viewed -- but also because of the matching to find the right person for their designs.

BI: The personalization features demand a lot of information from users. How do you get people to invest in this at the front end?

De Luca: Part of it is that the women are rooting for us. Women are tired of going into stores and having to pull size 10 and 12 and 8 into the dressing room because you don't know which one will fit. Or just finding the right size is not in the store. We see ourselves bringing personal shopping to people, that promise and that intimacy. That is why we have been able to develop the trust. Women are realizing that the more information they do give us, the richer and the deeper we can be in terms of meeting their needs. Women are really motivated to eliminate the disappointment of getting something that doesn't fit.

BI: Since personalization already is built into the framework, what does a recommendation engine like your new Baynote partnership add?

De Luca: It is interesting. They brought to the table for us taking the community recommendation of the wisdom of the group. You look at this purple blouse, and other women who also looked at this purple blouse are looking at these silver dangling earrings. If we just did that it would be fine. But we only show the recommendations inside the personal shop. It means that other women who looked at this purple top that you are looking [at], which is in your personal shop, also looked at these other items that are also in your personal shop. And we make sure these things are in stock. So if I click on one of those recommendations like a shawl or jacket, then I am brought to the product page for that item and automatically given the size recommendation for that one.

BI: Is that recommendation grounded in people of similar size? Is that pool of recommenders also probably in your shape as well?

De Luca: Shape wise it is probably more likely because similar shapes of different sizes would see the same garment if we had it in their size.

BI: It seems like an interesting way to parse a crowd. Offline, many women will identify with one another's shape or size challenges and then recommend their solutions. It would be an interesting way to parse the online crowd according to groups that are shaped in a similar way or facing the same issues. Then the recommendations would flow out of their solutions.

De Luca: Exactly, and in this case that is correct. In order for the item to be in your personal shop it does need to match on the shape. Women will say that something makes them look short or broader hipped and someone will say you should wear a lighter color on top or a different neck. I may be helping another woman who many not be as savvy.

BI: You have had the recommendations in place for several months. So here is an instance where we really can see what effect a recommendation engine like this has when it is dropped into a personalization model. What results have you seen so far?

De Luca: We have seen good results in terms of a lift on average order. I think we were looking at 16% over average orders. So we were really happy with that. We are still working on our own information in terms of what we observe and how people shop. So I think there is still more to gather to understand more.

BI: Is there ever a problem of scale, where there just aren't enough people who view the same item to make the recommendations very material?

De Luca: It is funny you say that, because it is less about the scale. The only time I don't see a lot of recommendations is in the very instant that we put up a new garment. We are posting continuously throughout the day new garments. Something doesn't have to be up very long before it has six to eight recommendations, and that is surprising to me. I am looking at something that went up , and it already has all eight thumbnail items in the recommendations.

BI: Are you gaining any new insights from these recommendations about your customers?

De Luca: We see some things, but want to do more. What has impressed us about it is that the crowd does a pretty good job. It's not just finding things that are alike; it is finding things that are also complementary. That has been interesting to see how people choose to pair things. And that is a direction we wanted to go in the future and have on our roadmap so that fits in with the other things we are doing for members.