Persuasive Web: Where Psychology Meets Conversion

DAY 4: Use Data to Sell: Recommenders in Online Apparel Shopping

Posted in 30 Days of Persuasion by persuasiveweb on June 4, 2009
Recommender Agents Can Get Your Customers to Buy More of the Clothing They Want

Recommender Agents Can Get Your Customers to Buy More of the Clothing They Want

When I buy shoes, I think about the outfits I’ll wear them with. When I buy an outfit, I think about what shoes will go with it. In fact, nearly every time I see an outfit I like on a mannequin, I look to see if they’ve paired it up with shoes… just to get an idea as to what I should buy.

(Bear with me, non-shoppers. I’m going somewhere here.)

Traditional print catalogues have been great for showing shoppers the whole ensemble. Sears comes to mind immediately. And even J. Crew, which sells shoes, accessories, and clothing, has recently started selling Timex watches in their print catalogues — assumedly because people like to buy coordinating pieces and feel confident that they’ll look stylish… even if style doesn’t come naturally.

ANTHROPOLOGIE - PURCHASEMaybe that’s why has a “Buy This Outfit” function on their site. Because there are a lot of moving pieces to creating an outfit, and it can be hard for one person to determine which pieces go best with which others. Stimuli overload!

Welcome to the World of Recommender Agents

What happens to online shoppers when they’re faced with lots of information? Huang, Chung, and Chen (2003), referencing other studies and their own, argued that information overload online creates enormous challenges for customers during decision-making… and that intelligent recommender agents can help customers overcome those challenges.

Recommendation systems assess what a customer is currently viewing online or has purchased in the past and work a little behind-the-scenes magic to recommend additional purchases to them.

Amazon Recommends Products Based on Purchasing Histories

Amazon Recommends Products Based on Purchasing Histories

What exactly is that “magic”? Well, intelligent recommenders use knowledge engineering, collaborative filtering, content, or a hybrid of those approaches to make recommendations (Huang et al., 2003) and depend on “machine learning” as well. Most commonly, intelligent recommenders will use association data mining (Huang et al., 2003), or data from other customers, to recommend products/solutions, like Amazon does here:

AMAZON - Customers Who Bought This Also Bought That

And here:

AMAZON - Customers Ultimately Buy This

Simpler recommenders may pull recommendations in from “neighborhoods” (Huang et al., 2003), which are basically pre-set categories of similar or complementary items to the one a user is viewing or has viewed. For example again, Amazon:

Amazon - Buy This Product and Related Accessories

The goal of intelligent and simple recommenders is the same: To help users make more purchasing decisions more confidently on an e-commerce site.

Persuade them: People just like you buy products just like this

The beauty of recommenders is that they are one more tool to help time-harried shoppers, users overwhelmed by information, and users lacking the ability to judge the quality of a product make the decisions they want to make.

We know that, when abundant information hinders rational choice, the brain relies on social influence choice (Donadebian, 2006) to help us decide. Social influence choice or social proof, as we’ve seen already, is great for online decision-making. Recommenders are one more type of social proof… they’re just based on quantitative purchasing data rather than qualitative user assessments. 🙂

Who’s using recommenders to sell more product… and who’s not?

Well, Amazon’s obviously the king of recommenders, as shown in the screenshots above. But I’m not shopping on Amazon right now. No, right now, I want to buy a new pair of pumps… so I’m heading over to Piperlime.

For those who like to shop for clothes, Piperlime is great because it’s actually 5 sites in one: the Gap, Old Navy, Banana Republic, Athleta and Piperlime. That means just one checkout process (one shipping fee, one account setup, one credit card entry) for a range of clothing shopping needs. We love it!

Given that there are so many outfit + shoe purchasing opportunities on Piperlime, we fully expected the site to have a recommender agent of some kind. Even a simple one. ‘Cos it’s a great cross-selling opportunity… and what marketer doesn’t love those words?

We arrived on the site. We found some great heels and, on the product detail page, all sorts of awesome social proof in the form of user ratings and reviews.

A - PIPERLIME - Reviews and Ratings


A - PIPERLIME - Sorting Reviews

Then we added the shoes to our cart and expected to be met with “customers who bought these shoes also bought this bag” or “these shoes look great with this dress from Banana Republic”. …But we got nuthin’. No, we were sent directly into the cart to create an account… with nary a recommended cross-sell to bump up the average sale price.

This lack of recommender represents a persuasion opportunity for Piperlime, Gap, Old Navy, etc., which are already doing great things with social proof. User testing could help determine if recommenders would feel too salesy to users. And an A/B test could also help determine if recommenders on Piperlime lead to any sort of lift, especially today when some might argue that people lack the disposable income to even consider additional purchases.

Apparel etailers who ARE using recommenders, albeit very simple ones, include Bluefly

A - BLUEFLY - Recommender

and J Crew

A - JCREW - Recommender

Those etailers — and others! — must hypothesize that recommenders help users… and the business. An A/B test could help reveal some insights there.

The truth is that recommenders do help users make more confident purchasing decisions. They also persuade users to buy items the might otherwise not have considered. All with the data you’ve probably already got. So use ’em!


Donabedian, B. (2006, July). Optimization and its alternative in media choice: A model of reliance on social-influence processes. Information Society, 22(3), 121-135. Retrieved May 10, 2009, doi:10.1080/01972240600677771

Huang, Z., Wingyan Chung, B., & Chen, H. (2004, February). A Graph Model for E-Commerce Recommender Systems. Journal of the American Society for Information Science & Technology, 55(3), 259-274. Retrieved June 4, 2009, doi:10.1002/asi.10372


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: