How to Choose Tools That Customize Online Shopping

By Michelle Megna

A recent study by Forrester Research analyst Sucharita Mulpuru explores the concept of e-tail "personalization," or the capability to provide customers with customized products and offers. The report, "Which Personalization Tools work for E-Commerce and Why," outlines this growing trend, citing relative newcomers in the field who now offer such services at more affordable prices, even for small businesses. (Some of these players, such as Aggregate Knowledge, CleverSet, Baynote and Mercado, have been covered here at ECommerce-Guide. See our "Related Stories.")


Since shoppers generally embrace personalized buying experiences, Mulpuru explains how Web shop owners can decide which tools will help them customize their transaction process. First, however, she provides a comprehensive overview of the trend, outlining exactly what personalization encompasses, why e-tailers should be aware of it, the factors boosting its popularity and what types of goals it accomplishes.

To start, she defines the term personalization. This can mean one-on-one interactions, such as greeting return customers by name or letting them save a shopping cart, or accomplished on a "one-to-many" basis, for instance, by creating different versions of your site for different demographics. For the purpose of the study, she says personalization is "creating experiences on Web sites or through interactive media that are unique to individuals or segments of customers."

Mulpuru says e-tailers should be interested in providing personalized shopping transactions because they increase customer engagement and loyalty through increased relevance. "Enabled by external tools sometimes called personalization engines, recommendation engines, discovery engines, or behavioral targeting tools, personalization allows retailers to increase relevance through activities like matching cross-sells to customers based on interests or customizing click-stream paths based on previous purchase or visit histories."

Specifically, this type of customer interaction is important because shoppers value recommendations. "Seventy-seven percent of customers say that they find recommendations in general somewhat to extremely useful, and roughly one-third of consumers who notice recommendations on e-commerce sites report purchasing a product based on such recommendations," according to the study.

What Took So Long?
So, the question is why have e-tailers been relatively slow to adopt personalization? Mulpuru says the answer is simple: During the past 10 years it was too complex and expensive to set up. Now, though, she cites a "renaissance" in the industry and says the following factors are making it easier for online sellers to use personalization:
Cheaper deployment costs. There are are now tools based on a revenue share of incremental revenue generated through the recommendation engine, eliminating any upfront costs, which is naturally appealing to smaller e-businesses.

Flexibility within the tools. One of the most common critiques of personalization tools used to be that they weren't flexible or adaptable on-the-fly. Given this well-known shortcoming, developers of the new generation of tools have addressed this issue, either work closely with clients to alter algorithms or to provide user interfaces where clients can affect rules independently. E-business executives report that companies such as Aggregate Knowledge and Certona respond very rapidly to client requests for change. Additionally, the fact that many of these solutions are software-as-a-service models enables them to offer flexibility.

Time to focus on the "nice-to-haves." For years, e-tailers focused on basics such as zoom functionality or on-site search tools or even site analytics. The majority have now mastered these "must-have" tools and are now making forays into the next tier of products that employ more quantitative approaches, and personalization is one of these. In the past, companies matched product cross-sells on their sites manually. In fact, a Shop.org survey of nearly 200 online retailers executed by Forrester found that 77 percent of retailers executed cross-sells by hand. Thirty-seven percent of retailers, however, say that they will focus on automated product recommendations in 2008.

The Current Landscape
Next, the report does a good job of providing an overview of the current e-commerce personalization industry, which is complex due to the many different ways it can be accomplished. For instance, while some companies claim to simply help create cross-sells, others promise to make a homepage more effective. According to the study, despite the nuanced differences in all their approaches, there are "four key buckets" that e-commerce personalization tools fall into as outlined here in the report:

Versioning tools. These tools typically personalize an experience by first defining segments of consumers and then serving up different iterations of key pages of Web sites (e.g., a homepage, checkout page or offer page). An example of such an execution would be showcasing different versions of a homepage to different visitors (e.g., new versus repeat) or different offers to different segments of consumers. In some unique situations, the data that informs the outputs can also be used across channels to create unique e-mail programs or even differentiated print campaigns for individual customers. As a result of their approach, these programs typically require extensive creative resources to support the various "versions" of an optimization campaign. For companies that want to slowly test what works first or want to carefully control their messaging, these tools can be extremely effective.

Simple cross-sells. These tools take implicit and sometimes explicit data and simply place what they believe to be the most relevant "adjacencies" in a predefined box on a Web page. These are often low-complexity, inexpensive, easy-to-integrate and simple solutions that help to automate the tedious processes of Web site merchandising or cross-selling. Small to midsize e-tailers typically are the most active customers of these tools, and companies such as Avail Intelligence, Baynote, CleverSet and Loomia are solid providers of such solutions.

Advanced cross-sells. These tools incorporate all of the features of simple cross-sells but also have the capability to push suggestions to other parts of a site (e.g., a homepage or outgoing e-mail programs). Advanced cross-sell solutions run the gamut from souped-up single-cross-sell solutions that can operate seamlessly in different areas of a Web site to more sophisticated solutions that create completely different navigation experiences for different customers. The key element that distinguishes advanced cross-sells is that they take outputs and feature them dynamically in a manner that is more than just "a box on a page."

Interactive filtering solutions. Given the vast assortment of products available online, consumers are often overwhelmed by the process of finding an appropriate match for their needs. Interactive filtering tools ask consumers for specific inputs, usually by posing a series of questions and then matching responses based on their preferences. The key factor that differentiates these tools from the other e-commerce personalization tools is that consumers essentially "raise their hand" and say what sort of information they want, and companies work to provide specific data or products that meets those needs. Companies such as Zafu.com and Karmaloop.com employ interactive filtering tools particularly well.

How to Find the Right Fit
The report goes on to match which types of Web shops would perform best with what personalization tools. If you have lots of resources and a budget to dedicate to personalization engines, versioning tools and advanced cross-sell tools are recommended; if you have an extremely broad and complex inventory of products, simple cross-sell tools, advanced cross-sell and interactive filtering are recommended; and if you have the need to have close control/input of the content at your site, versioning tools, advanced cross-sell and interactive filtering models are best.

So once you're ready to choose a vendor, given the complexity of the market, how do you go about it? This excerpt from the report provides the critical questions to ask:

How much data is gathered and from where? This is perhaps the most important question to ask a personalization tool company, because one of the biggest stumbling blocks of e-commerce optimization is a concept called the "cold start," which essentially means that there is not enough data to provide meaningful recommendations.

In this case, regardless of the sophistication of an algorithm or the number of Ph.D.s who crafted it, sparse data sets will yield poor recommendations, which not only creates a poor customer experience but also does little to drive sales. Companies such as ChoiceStream are able to address this issue by creating entire taxonomies of associations for their clients. Aggregate Knowledge drops third-party cookies onto its network of sites and gathers vast quantities of data throughout the Web that inform recommendations.

How sophisticated is the reporting? Given that lifts in sales can be very subtle and are frequently associated with certain key pages, it is critical to understand how personalization tools interact with conversion. Site analytics and Web site tagging, while helpful, are generally insufficient to gauge the entire effectiveness of a personalization tool. Tools by vendors such as Certona have the capability to drill down their reports to an item level, which can be critical to providing insight into what specifically is working (or not).

How quickly can it be changed/adjusted? One of the downfalls of the early engines was the "black box" or the relative lack of mutability around the formulas. While the "black box" is virtually nonexistent today, alterations to an algorithm are typically made in one of two ways: either by the client company or through a client services function provided by the vendor.

How many clients does a vendor have? Given the relative youth of so many of the e-commerce personalization tools, any company with experience in a given vertical or industry will likely have an advantage over other competitors. Why? The company will probably have already addressed the complex nuances of a particular industry, which means less algorithm tweaking after deployment. Coremetrics and CleverSet are two examples that are heavily focused on the retail sector, while companies such as TouchClarity actually grew from a background in the financial services vertical. Others, such as Aggregate Knowledge and Baynote, have experience working with media and content providers, in addition to focusing on retail. ChoiceStream has perhaps the industry's deepest experience with media companies but is also gaining traction in other sectors within retail, particularly with heavy-traffic, SKU-intensive sites.

How well-capitalized are these companies? The influx of funding into the personalization tool space has already affected the landscape. Acxiom's purchase of Kefta and Omniture's acquisition of TouchClarity make Kefta and TouchClarity more likely to be upsells to existing Acxiom and TouchClarity clients and may affect their future road maps and commitments to clients. Likewise, the relatively rich capitalization from venture capitalists of firms such as ChoiceStream and Aggregate Knowledge puts pressure on the sales processes of less capitalized competitors.

Is the tool a stand-alone tool? As mentioned above, some of the personalization tools sell only recommendation engines. Others, such as Endeca or Coremetrics, sell on-site search or site analytics first and optimization products as incremental (and sometimes separate) features. While it can often be easier to just manage a single vendor, the risk associated with an all-in-one solution is that it is less likely to be "best-in-class," since development efforts are primarily focused on flagship rather than ancillary products.

Are you comfortable with third-party cookies? Some companies rely on third-party cookies as a means of gathering incremental data. This means that some key elements of information — such as where, when, and how consumers click through a Web site — are shared, albeit anonymously, in a larger pool of data to unearth consumer behavior and reactions to different types of content across the Web. Companies uncomfortable with this approach should carefully qualify such companies to evaluate if a relationship make sense. For companies reluctant to pool their data, partners using third-party cookies may not make sense, but for those businesses bold enough to be part of a data consortium, the results can be extremely rewarding.
Evolution of Personalization

Finally, Mulpuru provides some insight on what to look to in the future of personalization. She predicts there will be a shake-out in the industry, with similar companies merging while others are pushed to the margins, as well as, with any luck, simplified pricing.


"The land grab for clients among the numerous vendors in the space has resulted in a dizzying array of pricing structures ranging from revenue shares to price-per-click models to flat fees to equity investment opportunities," she says. "As in the case of most other e-commerce tools such as rich media tools or on-site search, e-commerce personalization tools vendors will be wise to adopt a multi-year, flat-fee model with some variation based on usage of the tool. This will likely be the natural course of action once the tools are established as 'must-haves.'"

Michelle Megna is managing editor of ECommerce-Guide.com.