Relevant Content Is King
This series explores many opportunities to grow sales by shifting our focus away from our companies and our products to our customer’s priorities and their interests. With customers firmly in control of the buying process, identifying and meeting our customer’s priorities and preferences are paramount.
For those of you making money by presenting content to consumers, it should come as no surprise that your customers will find greater value in your services if you present the content most relevant to them. Anyone who has uses Pandora, Netflix or Amazon instantly understands this concept and the additional customer engagement which results. Relevant Content is King!
Everyone should be very motivated to get the right content to the right user at the right time. Delivering relevant content is a simple concept one old as commerce itself, but making it happen is not simple so we recommend doing so in these three phases:
1. Define the level of content relevance you can or want to deliver.
2. Create or modify your existing site layouts to allow specific users to see specific content which is identified as relevant.
3. Deploy the technology engines which deliver the right content to the right user at the right time.
Phase 1 requires you to define the level of relevancy in the content delivery you plan to deploy, you should scope the level of effort you are willing or able to pursue. Naturally, the more effort you apply the more valuable the outcome as measured by sales growth. Consider these options:
Content centric similarity is the easiest form of content relevance to deploy because it is not based on human behavior at all. Instead, content centric similarity is attributing based and any other content that shares that attribute are deemed “similar.” For example, the Rolling Stones and the Beatles are both British rock bands who performed in the 60’s so their music is “similar.” Naturally, using this content-centric approach requires a very intuitive categorization paradigm allowing users to easily discover content that is meaningful to them.
User designated data captured by explicitly asking the user their content preferences yields very accurate data, but not all users will supply it. This is akin to using search key words to match ads. Directly asking a user what content — in an elegant manner - ensures they user will see exactly what they want to see. This can be accomplished in a number of ways including survey made during a registration process or during usage in real time.
Behavior based similarity is another user-centric approach to determining relevancy which yields more accuracy but is harder to implement than the content-centric approach or explicit data approach. However, it generates the most data on which you can base your content displays, in particular because many users will not explicitly indicate what they like. When a consumer knows exactly what she is looking for, she searches for it. But when she is not looking for anything specific, she browses. It is the browsing that holds the golden opportunity for a recommendation system, because the user is not focused on finding a specific thing - she is open to suggestions. Behavior-based similarity requires tracking how users interact with your content and finding patterns of usage that other users duplicate. The user is not actively involved in setting the attributes for content relevancy, instead their actions (what click on, hover over, or display) determines what else may be relevant to them.
Phase 2 may require changing your content display pages post registration that are able to dynamically display relevant content based on a trigger or signal from a third party system or from a internally operating program. Deploying dynamic web pages can be complex, but the benefits of delivering relevant content to users are numerous and far outweigh the costs. Deploying dynamic display pages based on relevant content not only increases the number and duration of page views while improving interactivity, it also makes large website easier to maintain and less complex to update.
Phase 3 involves the selection and deployment of a rules based decision engine (also called recommendation engines) and it is the most technically challenging part of this process. There are a two different ways to pursue Phase 3 — buy or build — and each has their advantages and disadvantages.
Buying a pre-built recommendation engine such as those made by Strands (http://recommender.strands.com) or Istobe (http://istobe .com) allows you to quickly add content relevance capabilities with limited capabilities. More elaborated solutions for behavioral targeting which offer far more rich and robust solutions are sold by vendors including Art Technology Group (www.atg.com) which is now part of Oracle, Coremetrics (www.coremetrics.com) which is now part of IBM, Omniture (www.omniture.com) which is now part of Adobe, Avail (http://www. avail.net) and Click Truth (www.clicktruth.com).
Prebuilt software for the User designated data approach mentioned above includes Pop-UpBooster (www. popupbooster.com) offers a great and very easy mechanism to ask users what they want in real time. While these solutions offer various capabilities and features, they will cost you to setup and to run because they are robust and well tested.
Alternatively, there are several open source solutions for deploying your own recommendation engine. Here again there are two choices to consider. First, and recommended, are those open source solutions which are relatively simple to deploy, and EasyRec (http://www.easyrec.org/) has a good reputation.
For more sophisticated outcomes or for the most complex solutions sets, consider the Apache Mahout project Taste (http://mahout.apache.org/) which has its roots in the same technology as Amazon, or using an open source rules engine to create your own logical approach. The three most popular rules engines are from Vistology (www.vistology .com), HP Labs (www.jessrules.com) and Sandia National Labs (jena.source forge.net).
The three phase process outlined here is proven and reliable, but the big picture should not be lost in all these details. Expanding customer engagement online directly drives increases in customer spending which directly drives profit. Treating each customer as an individual is the single best way to deliver value to the market and those companies who deliver the most (perceived) value earn the biggest slice of the pie.