Customer Service and Search
Everyone knows that it is always less expensive to keep an existing customer happy than to gain a new one. Today's competitive marketplace, combined with an increase in online shoppers, makes it more critical than ever for retailers to provide a positive and meaningful customer experience at their Web sites.
Consider the following scenario: a potential customer visits a site and looks for a "trigger" word - the word that most naturally describes the product she is seeking - on the home page to direct her to the product she is looking to purchase. If the customer does not find that "trigger" word, she will move on to the search box. It's at that very moment that the search box becomes an online salesperson. Whatever search results the customer receives directly reflects the retailer's brand and can make or break a sale. Imagine if a customer was shopping for a "pocketbook" at a brick and mortar store, but the salesperson wasn't familiar with this term and could not translate "pocketbook" to "purse" or "handbag." The same scenario applies to online shopping. If a customer searches for "pocketbook" and gets an error message - when in fact the online store carries this product - odds are that the customer is going to become frustrated and go to a competitor who "has" the product.
A recent Forrester Research study found that 70 percent of Web site shoppers couldn't find products they sought with search - even when the online retailer stocked the items and they were available. Customers will usually try one more time after a failed search, and then they leave the site if that search also fails. Why does search seem to fail so often? Five problems most commonly emerge:
1. Dependence on the completeness and consistency of the product description
The accuracy of text-searching product descriptions is solely dependent on the accuracy and completeness of the text in the product descriptions. These descriptions are sometimes derived from the manufacturer, whose descriptions will vary dramatically. Other times the descriptions are developed by the retailer's merchandising or marketing people, who often use original or unique-sounding descriptions. Product descriptions therefore naturally tend to lack uniformity, breadth and depth, and this directly impacts the precision of search results. The old phrase "Garbage In/Garbage Out" applies in this case. The harsh reality is that text-searching the product descriptions employs the least accurate information available and ignores the other two sources of product information: product attributes and merchandise categories.
Since conventional word-matching search technology - the legacy search software in many cases - does not make use of information stored in product categories and attributes of the product catalog, it is important to ensure that all terms in these areas are mirrored in the product description. On the surface this sounds like a reasonable solution, but in practice it is extremely onerous to maintain. Each time a new product category or product attribute is introduced, all possible keywords, attributes, or possible searching terms must be entered for each product in the product descriptions. This places undue burdens on description authors and is a significant undertaking, since most retailers add new products regularly.
2. Dependence on exact word matches
Text searching itself is an inherently brittle process. It will work well in one case and extremely poorly in another. The problem is that the customer cannot distinguish between the two cases, and therefore the search results appear to be arbitrary; as a result, a customer using text search is often left feeling very frustrated. For example, if the customer searches for "pants," she may receive satisfactory results, but if she searches for "slacks" or "trousers," the customer will be informed that no products meet that criteria. Clearly, this is confusing to the customer, who knows the products are out there somewhere.
In many cases the customer is required to use a different form of the same word. For example, searching on "hiking boots" may not find products described as "hikers." This, of course, makes no sense from the customer's perspective, but is a frequent occurrence if the retailer's site search technology doesn't handle basic query checking automatically.
3. Inaccurate handling of numeric attributes
Numeric attributes are extremely problematic when performing text searches because the text search process has no understanding of what the numbers mean. For example, if the customer is looking for "a 3 inch screw," the text search will display all screws that have a "3" anywhere in the description.
Products can have a number of numeric attributes (e.g., length, width, or weight) and the inability to distinguish among them is a serious limitation. This phenomenon is guaranteed to yield inaccurate results in virtually all cases.
4. Inability to express more complex selection criteria
There are a number of more complex selection criteria that cannot be expressed through text search. Therefore, these queries always yield inaccurate results. One simple example would be searching for a product "under $50." It is impossible to express the restriction "price < $50" in the traditional word-matching search paradigm. Customers have been conditioned to use simple keyword searches (typically one or two words) by traditional word-matching search tools that originated for Internet use. However, these terse searches tend to produce less focused result sets than customers could achieve were they given the advice - exactly the opposite of most search tools - to try adding MORE details to their query. Of course, that would only work if the search software on the site could handle the longer query, which few technologies on the market today can.
5. Inability to search on intended use of the product
As if the above limitations were not enough, the traditional search process fails to address one of the most common search phenomena. Customers often express their product needs, not by describing the product itself, but by describing their intended use of the product (for example: "hunting boots," "shoes for electrical work" or "a gift for a 13 year old girl"). None of these searches will yield any results because none of these words are likely to match words in the product descriptions.
Once again, the idea of putting additional content into the product descriptions seems like it might solve the problem. However, the idea of putting every intended use of every product into the database is impractical. The result is that this very common way for customers to express what they are looking for is doomed to failure using the traditional search process.
How many salespeople would allow 70% of their customers to leave a brick and mortar store without locating the product they were specifically hoping to find? With the number of online shoppers growing exponentially, e-commerce sites are faced with the task of ensuring this does not happen.
Customers should demand the same level of service from a Web retailer that they expect from its brick-and-mortar counterpart. One way to achieve this optimal level of customer service is to add intuitive and precise search solutions that can respond to a broad spectrum of customer queries as accurately and effectively as an experienced salesperson would. Many e-commerce sites have yet to implement a search technology that even comes close to this capability, but the innovators who have are now reaping the benefits through increased revenues.
As an example, Lands' End was ranked as one of the most successful in online conversion rates by Forrester Research. Lands' End now allows customers to ask questions as if they were speaking directly to a salesperson. A customer can enter specific searches such as "women's hiking boots under $40" and the customer is immediately shown only those items that fulfill those parameters. Executives at Lands' End note that this capability provides the kind of shopping experience that their customers have come to expect, and is an important reinforcement of their brand position. Furthermore, making this change has paid off handsomely with an immediate increase in conversion rates directly attributed to successful searches.
Precise search technology pays for itself, not only by helping customers find what they have in mind, but also by increasing customer loyalty through an enhanced experience with the brand. Is there anything more important to an online retailer?
[Steve Morse is Director of Product Management, EasyAsk Inc.]