You Say Tomato, I Say T-Shirt

Jack Aaronson

Updated · Jul 31, 2002

The Internet hosts two basic types of users: browsers and searchers. Personalization projects tend to focus on browsers, not searchers. The problem is 80 to 90 percent of people who go online to buy products are searchers.

This week, we will discuss how to make searching more usercentric.

If you sell products or services on your site, make sure your search engine uses more than your internal categorization and taxonomy to index them.

Here’s the quick and easy way to figure out the other keywords you need to use:

  1. Assemble five friends (through email is easiest). Make sure these friends are end users, not coworkers. If you sell in the business-to-consumer (B2C) market, these people are folks such as your mom and your neighbors. If you operate in the business-to-business (B2B) market, they should have industry knowledge but shouldn’t work with your company.
  2. Send each friend pictures of your products. Don’t tell them what the products are called, and don’t send descriptions. Make sure the products are varied (in other words, don’t send the same friend five shirts unless shirts are all you sell).
  3. Ask each friend for six or seven descriptions of the product, including nouns and adjectives. Descriptions should be one or two words long.
  4. Ask each friend for four or five scenarios or actions in which the given product would be needed. Scenarios should not exceed five words.

After you try this exercise with 30 to 40 products, you’ll most likely begin to see a trend in language. That is the taxonomy your users employ to describe your products.

Your test users might use words already in your taxonomy that mean something different to you. As an example, I tried this test with a Banana Republic shirt. The descriptions I got back were: “black,” “cotton,” “t-shirt,” “undershirt,” “no-neck” (as opposed to v-neck), “collarless,” “plain,” “simple,” and “casual.” The scenarios were: “dance club,” “dancing,” “school,” “fall,” “walking around town,” and “dinner with friends.” “T-shirt” and “cotton” are words in the current Banana Republic taxonomy, but they point to different items. Technically, the company’s understanding of these terms is more accurate than a customer’s.

It doesn’t matter. Every company has its own terms and references. Users can’t be expected to know (much less remember) what your company calls things. I attended a conference given by ClickZ’s own Bryan Eisenberg and brother Jeffrey, and they made this point very clearly during their sessions as well. All of these words and scenarios must be input into your site’s search engine so people like my friends can find your products. After all, my friends represent your buyers much more than you and your employees do, never mind the database guy who came up with your taxonomy in the first place.

Knowing how users described the shirt I sent them, the second test was going to Banana Republic (both the stores and Web site) to try to find the shirt. Its site lacks search functionality (which I couldn’t believe). I couldn’t find the shirt on the site by browsing. The site categories include Ts and polos. Nowhere was the shirt to be found.

I called a local store here in New York. I spent at least five minutes trying to explain the shirt to the man on the line. Finally, I realized because I actually had the shirt in front of me, I could give him the product number.

Guess what the shirt’s called? A “Pima Cotton Crew.” Umm, OK. I went back to the Web site. “Crew” isn’t one of the options on the site. So is “crew” more like a “polo” or a “T”? I looked in both sections but couldn’t find my shirt. With no search functionality, I couldn’t enter “Pima Cotton” to see if it was somewhere else. Do I have to know what everything is called to shop at Banana Republic? I shouldn’t have to.

Needs-based design is the second half of the equation. I have talked about this before, but not in the context of searching. After I asked my five friends (in the above example) to act like a consumer-focused thesauri, I asked them for five scenarios or actions. These are the “user needs” surrounding the products. Car manufacturers understand selling a car that’s good for “off-road terrain” to “sportsmen” is a great way to sell a car. What features make up an “off-road terrain” car? I am not sure. I will bet you four-wheel drive, some fancy braking system, and other “rugged” features are part of those cars and trucks.

But the user need (the problem) was “outdoor sports and stuff like that.” The features (the solution) were “four-wheel drive” and “antilock brakes.” By letting users know you understand the problem, you are also telling them you understand the combination of features that go together best to form a solution. If I were buying a car for “off-road mountain climbing tours,” I would trust an SUV geared toward “off-road extreme sports” more than a model that let me pick and choose every option, without helping me understand the circumstances under which that feature would be most useful.

I Say “Tomato,” You Say “Pima Cotton Crew”

All product and services sites would do well to keep user-scenario design in mind. It’s much easier for me to search “dressing for a first date” or “going to a friend’s graduation party” and get outfit recommendations than it is for me to perform a search on “shirts” and “pants” or to browse the site not understanding the difference between the types of designs a particular store has or where it’s socially acceptable to wear the different styles.

You’re too close to your own products and services to do this effectively. Unless your products are so large and varied you need to hire someone, learn from getting five friends to help you with the experiment I above.

Once you do, let me know if you have any startling revelations.

Until next time….


Reprinted from ClickZ.

  • Retail
  • More Posts By Jack Aaronson