Nonconverting Keywords and the Search Continuum, Part 1

What if I were to tell you the majority of online purchases from search advertising happen after a searcher has conducted at least 12 nonconverting searches? Imagine, 12 searches on keywords that don’t produce a single transaction. But these eventually led to a search that does.

What would you do differently? How would you change your pay-per-click (PPC) bidding strategy?

A recent study conducted by comScore and sponsored by Overture suggests exactly that. If you believe the data, it has broad implications for search advertising campaigns.

The study generated only limited coverage, partly due to the timing of its release (during the holidays) and partly because it was so rich with stats the most significant findings may have been overlooked in a forest of data points.

This particular research is simply the most important search engine marketing (SEM) study to date. Why? It’s the first one to connect nonconverting keywords to conversion outcomes. That’s the first step in connecting SEM with all other marketing.

In a recent column, I posited nonconverting keywords would become the key to paid search growth. I asked anyone with information that could validate this theory to contact me. Diane Rinaldo, director of strategic alliances at Yahoo, pointed me to the comScore research.

The study follows searchers from the initial search for three months. That’s the longest time span of any search study to date, according to Rinaldo.

The study follows searchers seeking consumer electronics and computer products on search terms categorized as follows:

• Header term: High frequency, but very few broad product terms (e.g., “camera,” “computer,” “MP3 player”)
• General term: Lower frequency, but still general product terms (e.g., “CD burner,” “computer memory,” “PDA phone”)
• Specific term: A detailed product query often paired with a brand name (e.g., “HP notebook,” “ZX 500,” “Epson printer”)
• Trademark term: Any search with a specific online retailer destination brand name in the query (e.g., “BestBuy.com,” “CircuitCity.com”)

The semantics and nomenclature of the findings can be a bit confusing. The trademark category doesn’t refer to branded search terms per se (brands can be part of the query construction in specific terms, too). Instead, it refers to a search for an online retailer where a purchase could be made, a destination site. This could be partially why the study received only limited coverage and even less chatter from search engine marketers.

The study asks a number of questions. The most interesting include:

• How many searches are conducted during the buying cycle?
• What is the volume of search activity by term type?
• How do users refine search activity over time?

According to Rinaldo, “When we talked to clients, they kept telling us that their brand name converted at the highest rate. But when we looked at our search data, we kept seeing all of these product searches. We just knew that there was more activity against product searches all combined than activity against brand searches all combined.”

There Is No Search Funnel

The study debunks an entrenched SEM myth, namely, that searchers refine their queries and search again on product-specific terms as they near a buying decision.

This is not the case.

The study finds 83 percent of all searchers start with generic search terms. Though 85 percent may conduct additional queries, all but a relatively small percentage will use these same generic search term types on their subsequent queries. Advertisers, however, mistakenly buy branded terms at the expense of unbranded terms and miss the majority of their converting audience.

Nonconverting Searches Lead to Converting Searches

Searchers who ultimately purchased a product online conducted some 13 searches before ever making the purchase. This means that for every single converting search term, there were 12 prior nonconverting searches — searches that today, most PPC search advertisers would never consider bidding on due to poor post-click conversion performance. The implication is nonconverting terms have a greater value than search marketers currently ascribe to them.

Yet how do most search marketers manage their bidding strategy? Their mantra: If it won’t convert, you must avert!

“If you’re not bidding on general or broader keywords consistently, you’re losing some of your opportunity to influence the purchase decision,” Rinaldo points out.

The majority of searchers in the study who ultimately purchased began by searching broader terms. Yet most search advertisers still believe they should primarily buy product-specific terms to reach buyers.

Extend the Time Horizon and Cookie Expiration

The study indicates we may also need to rethink our paid search time horizon. A significant increase in conversion rate was witnessed after a searcher had been investigating a product for over four weeks. According to Rinaldo, “It’s a long buying cycle. If companies abandon a broad keyword prior to four weeks, they lose a quarter of their audience.”

Many online retailers set cookies to expire in 30 days. The data from this research suggest a 30-day cookie doesn’t adequately reflect the buying habits of consumers who search for these types of products.

“We found the first, or ’start session,’ represented 38 percent of all search activity, but just 3 percent of conversions,” Rinaldo noted. “Contrast that with weeks 5 through 12, which experienced just 22 percent of search activity but 38 percent of the conversions.”

ComScore’s data suggest a need for new thinking about paid search. We’ll explore additional findings and implications for a bidding strategy in part two.

Join us for Search Engine Strategies in New York City, February 28-March 3.

Want more search information? ClickZ SEM Archives contain all our search columns, organized by topic.