Do Actions Speak Louder Than Words?

There’s a burning issue among web marketers about the best way to build online profiles.

One camp supports development of user profiles based on “declarative” data, where the user actively provides personal data. Another camp would rather develop user profiles from “behavioral” data, where the profiling system collects the user’s clickstream to build a unique user profile. These people contend that users don’t always provide accurate information about themselves.

Who’s right? Well, many of us believe that using both types of data is actually the way to go. (For a brief overview of declarative and behavioral data, see The Start Of A Beautiful Relationship.)
Here’s what Steve Larsen, vice president of marketing and business development at Net Perceptions, thinks: “If I had to choose only one, I would choose implicit [behavioral] data, because what customers do on your web site is more accurate than what they say about themselves. However, web marketers should consider using both declarative and implicit data when building profiles.”

Larsen suggested that people use different sides of their brain to make hypothetical decisions versus decisions in actual situations. Thus, one might contend that actions are more believable than words. Let’s take a look at the pros and cons of each type of profile data.

I Do Declare!

When users fill out a registration or online profile, they are declaring their interests, preferences and other unique information. And when a user conducts an online transaction, they are providing you with additional declarative profile data. On the plus side, you will get a lot of in-depth personal information than you can’t get from their clicks.

Some users take their online profiles seriously because they want the benefits from the profiling system. If I build an accurate profile, then that web site is going to save me a bunch of time by providing me with relevant information. On the down side, some users provide inaccurate information about themselves. While most users don’t actually lie (although many admit doing so), they may be inclined to tell marketers what they want to hear.

If you can do a good job of educating your online customers about the benefits of profiling and your commitment to protecting their privacy, you’ll probably get more folks building their online profiles, as well as more accurate profile information.

Know Me, Know My Clicks

The overall benefit to profiles based on behavioral or implicit data is that actions don’t lie. As I move through a web site, the profile system will track and collect my click-stream. After the system continues to track me, it will use statistical analysis to serve up banner ads or make recommendations.

In the real world, this would be the equivalent of a shopkeeper trailing you up and down the store aisles, observing you as you pick up products from shelves. In some cases, you’ll put some products back onto the shelves, and in other cases, you’ll put products into your shopping cart. So the profile system is akin to the shopkeeper making a notebook of observations for every visit to the store.

However, building profiles with behavioral data alone can be misleading. I browse and shop for so many people other than myself. I browse web sites and make online purchases for both my work life and my home life. And then, I also surf and buy for friends and family – lots of people with lots of different interests. If the profile system is tracking all of this, it does not have an accurate picture of me.

To further complicate things, what if more than one person uses a single computer in the home? In fact, my computer is used by both an Internet consultant/writer and a wildlife photographer. You could deduce that we have very different interests, preferences and other information that makes each of us unique. (Yes, we do fight over the computer often.)

The Best Of All Worlds

So, you can see that declarative and behavioral profile data each have their strengths and weaknesses. Knowing this, it is hard to argue against using both types of data in building customer profiles. Together, they provide a more accurate picture of the user. And, I’d also like to recommend these two additional capabilities of online profiling:

1) User Managed Profiles – This concept lets the user manage his or her own profile(s). These profiles are built with declarative and behavioral profile data, but the user can also modify them.

For example, if you are an Amazon.com customer, you should visit their Recommendation Center. Not only do they make recommendations based on your previous purchases, you can “refine” those recommendations by rating their recommendations in all of their book categories, such as Computers & Internet, Arts & Music, Literature & Fiction, and Travel. You refine your Amazon.com recommendations by rating each recommendation with two criteria: “I Own It” or “Not For Me.” Amazon.com then instantly makes further recommendations based on your refinements for each category.

My recommendation to Amazon.com would be to let me create more than one “Deborah Kania” profile – one for work life, one for home life, and one for gifts by recipient type (i.e., family members, friends, colleagues, etc.).

2) Collaborative Filtering – Basically, collaborative filtering is a process that builds profiles based on your web site interactions and transactions, plus the profiles of other customers like yourself, who have similar interests or tastes. Think of this as equivalent to word-of-mouth recommendations from people who know you or who are similar to you.

For web marketers, these types of web profiling/personalization systems can predict future buying behavior, and they can increase the likelihood of user inquiries or purchases. For the user, collaborative filtering can add a bit of serendipity to recommendations beyond simple rules-based personalization. For example, I am a woman and I like scuba diving, so give me product recommendations for women scuba divers.

As you can see, rules-based personalization is linear, whereas collaborative filtering is not. Thus, collaborative filtering is especially helpful when used on sites that have “fuzzy” subject matter that is more associated with psychographic data (i.e., attitudes, lifestyle, values, etc.) than demographic data (i.e., age, income, gender, etc.). Web sites associated with movies, books, clothing, art, food, hobbies, and leisure activities are perfect candidates for collaborative filtering.

We web marketers have some powerful one-to-one web marketing tools and services available to us, it is really a matter of which will be the best fit. Remember, don’t just base your profiling implementation on technology, but base it also on what you know about your customers, and how and why they buy or interact with your organization.

Next Week: It feels like just the right time for a case study.

Related reading

nurcin-erdogan-loeffler_wikipedia-definition-the-future_featured-image
pwc_experience-centre_hong-kong_featured-image
12919894_10154847711668475_3893080213398294388_n
kenneth_ning_emarsys_featured-image
<