How Keyword Length Affects Conversion Rates
A study of ROI trax® data confirms Oneupweb's hypothesis that, in general, keywords with longer strings have higher conversion rates. In addition was found that- conversion rates peak at four-word keyword phrases (or strings), noticeably dropping with longer phrases.
- single keywords have very high conversion rates, until you remove corporate names from the data. Once the corporate names are removed, single keywords have lower conversion rates than two-, three-, and four-word keywords.
The Study.
In January 2005, Oneupweb conducted a study to determine if the number of keywords in a search query was related to conversion rates. Hypothesis was that the longer the keyword string, the higher the conversion rate. Oneupweb focused study on data generated by natural or organic search engine results listings, not sponsored listings or pay-per-click advertising.
Using aggregate data collected by ROI trax® Oneupweb gathered traffic and conversion data related to search engine keyword searches for the months of July, October and December, 2004. Oneupweb's database includes data from hundreds of companies, both Business to Consumer (B2C) and Business to Business (B2B). Some are online retailers, others use websites to generate leads or provide information. All the companies in the study track natural search engine traffic and conversions using ROI trax tools.
The data was divided into categories by the number of words in the search query: one-word phrases, two-word phrases, three-word phrases, four-word phrases, and five-word phrases. For graphic clarity and because the data volumes were much lower, keyword phrases six-words and longer were combined and averaged together. Then was calculated the average conversion rate for each category.
"Conversion rate" is defined as the percentage of unique visitors that purchased or converted. Each client in the study defines their specific conversions; for example, some conversions were actual product or service sales, while others valued online actions such as downloading white papers, filling in forms or requesting proposals. The monetary value of the sale or conversion was irrelevant to this study; Oneupweb focused on the number of sales or conversions as a percentage of traffic.
"Traffic" refers to the number of unique visitors coming to the site for the first time. Repeat visitors aren't tallied again as traffic and so don't affect the traffic data of this study.
What did Oneupweb learn?
Oneupweb believed that conversion rates would increase as the number of words in a search query increased. And for the most part, the data confirms this hypothesis. There were two surprises in the data from the Full Database. First, one may noted that single-word keywords had a higher conversion rate than two-word keywords. And was surprised to learn that after four keywords, the conversion rate dropped. The graph (Figure 1) shows just how similar these trends are from month to month. Keep in mind the averages represented here are from hundreds of thousands of keyword searches in a single month.
 |
Figure 1. Full Database (corporate names included).
|
||||||||||||||||||||||||||||||||
|
Figure 2. High-Traffic Keyword Review Defined by Total Unique Visitors (corporate names deleted).
|
||||||||||||||||||||||||||||||||
Data from all three months shows that single-word keywords have two to five times higher conversion rates than two-word keywords. In October and December, single-word keywords had higher conversion rates than any other category.
The drop in conversion rate of query strings longer than four-terms is consistent across all three months. There are several situations that may be supporting this trend:
- Searchers may be looking for very specific products, services or information that they don't find
- Searchers using longer strings may be asking specific questions, indicating that they are still investigating a product or service; they may use a shorter string later when they are ready to buy or take action
- Search engines may not be effectively interpreting five-word searches, returning less relevant results and discouraging conversions
For more insight into this trend, looked at the High-traffic Keywords data (Figure 2). Again, the data from month to month showed similar trends. And, data from all three months exhibits the same surprising drop in conversion rates after an apex is reached at four-word keywords.
The most remarkable difference from the results of the Full Database to the High-traffic Keywords, is the single keyword data. In the High-traffic Keywords, Oneupweb removed corporate names and low traffic-keywords from the analysis. Without these keywords contributing to the results, the trend indicates that single keywords don't have the higher conversion data shown by the previous group. The trend here is more congruous with our hypothesis that longer keyword strings have higher conversion rates, with the exception of the drop after the apex.
Traffic issue
To further understand the business impact of keyword length, Oneupweb also reviewed average traffic per keyword. Again, looked at the Full Database and the High-traffic Keywords. A review of the Full Database revealed that single-keywords have on average the highest average number of unique visitors (Figure 3). Keep in mind that the Full Database includes corporate names.
|
Figure 3. Average Traffic Per Keyword. Full Database (corporate names included)
|
|
Figure 3. Average Traffic Per Keyword. High-Traffic Keywords (corporate names deleted)
|
The average traffic data for High-traffic Keywords showed a different trend. Single and two-word phrases have very nearly the same average with two-word phrases having slightly more traffic. Generally, shorter phrases had more traffic than longer keyword phrases.
Hypothesis was that traffic would drop as the length of keyword increased. The data confirmed this hypothesis and the traffic trend drops steadily as the length of the keyword phrases (or strings) increases.
(Note: As expected the High-traffic Keywords analysis has a higher average number of unique visitors per keyword than in the Full Database. The Full Database is averaging hundreds of thousands of records, whereas the High-traffic Keywords Analysis by definition is limited to records with high traffic.)
Oneupweb also examined the top 100 keywords as determined by traffic. In doing so, we can see if the relationship between keyword length and conversion rate holds for high-traffic keywords. The focused list allows to remove mitigating factors like irrelevant results, either from errors the user made in the query or errors the search engine made in interpreting content. By focusing on the top of the list, we removed the effect of that data.
Also, a company's name is a high-traffic generating search term. People searching for a particular company's name are predisposed to making a purchase (or conversion) specifically from that company. Most of these corporate keywords were 1-word (40 percent) and 2-word (31 percent) phrases (19 percent were 3-word phrases; 7 percent were 4-word phrases and 3 percent 5-word phrases).
How to apply this information
Based on these findings, Oneupweb is recommending that, for optimal conversion rates, it is worthwhile to focus on expanding optimization to three-word, four-word, and as appropriate, five-word strings. Oneupweb is not suggesting companies ignore single keywords in their campaigns - but instead, to add incremental effort to optimizing relevant multiword keyword strings ranging from two to five words in length.
None of this analysis accounts for factors like search engine results position, website usability, seasonality, or the competitive climate. And so Oneupweb caution marketers from thinking that optimizing for longer keywords will magically increase conversion rates, on its own. To be successful, an effective search engine optimization and marketing plan is a complicated analysis of the market and website factors affecting a business in its industry. Being optimized for longer keywords is only one factor in the mix.
Using hypothetical example, "Acme widget" is a relevant term for Acme. Additional, longer strings might include, "what is Acme widget", "definition of Acme widget", "usage tips for Acme widget" and "Acme widget white papers", or "maximizing performance using widgets".
