December 23, 2009
Case Study

Attributing Conversions to Assisting Keywords Lowers CPA: 5 Steps to Optimize Bids

SUMMARY: Although the last click before a conversion is the metric most search marketers use for attribution, it’s often just as important to understand the preliminary keywords that assist that final click.

Read how a marketer uncovered the non-branded and long-tail keywords that assisted their search conversions. By attributing revenue to these preliminary searches, they’ve optimized their keyword bidding strategy and reduced cost-per-acquisition 25.6% while increasing sales.

Roman Godzich, Director, eMarketing, American Management Association, knew that search marketing was a crucial piece of his team’s marketing strategy for training seminars. But the high price points of these sessions -- typically $1,700 or more per seminar -- meant the purchasing process wasn’t an impulse buy.

Customers typically had to get their supervisor's approval on a training purchase, which meant a longer research process often involving multiple searches and comparison of options. Yet Godzich’s team recorded most of its search-generated conversions from branded keywords or direct navigation.

"We started to realize that the conversation we have with customers [before purchase] becomes more and more important," says Godzich.

The team wanted to understand which non-branded keywords played a role in the research and education process, and how those search terms drove the eventual, branded searches that led to sale.


The team undertook a methodical testing and measurement process to determine which generic search terms played the biggest role in driving branded search and ultimately, conversions. Once they could attribute more conversions to non-branded search terms, they adjusted their PPC advertising budget to favor terms that delivered the best return.

Here are the five steps they took to gather data and put it to work:

Step #1. Implement technology to track visitors from entry search to conversion

To understand what role preliminary searches played in the conversion process, the team needed a new tracking system. Most search engine tools, such as Google Analytics, override a searcher’s cookie and attribute a visit to the last search term used -- eliminating previous search history.

They implemented a new tracking tool that tagged each user with a cookie that retained their search history. The system identified which term resulted in a first site visit, and then updated that record with subsequent search activity and website behavior.

Step #2. Establish groups of non-branded keywords to test

The team had roughly 50,000 active keywords in its PPC strategy. However, many generic terms generated a low volume of searches. This meant there was a small sample size for the team to measure when trying to determine a phrase’s impact on conversions.

In response, the team organized their non-branded and long-tail keywords into groupings of similar terms based on searcher behavior, to measure the impact of these keyword clusters. They established 26 keyword clusters in all.

Rather than organizing groupings around product portfolios or industry terms, they looked for relationships that indicated a searcher’s intent, such as:
o Communication skills vs. communication courses

Clusters around terms such as "courses" indicated a searcher interested in purchasing some kind of training, whereas "skills" might be used by a searcher looking for free, educational resources on the Web.

Step #3. Track search activity to determine which keyword groups assisted conversions

The team monitored search activity for roughly six months to accumulate data on which non-branded clusters led to branded searches and eventual conversions.

"We had to take into account the issues of latency," says Godzich. "Our product is not an impulse buy, it’s a considered purchase."

When the team saw a non-branded search term that led to an eventual conversion, they attributed the revenue from that conversion to the initial search term. This calculation allowed them to predict how much revenue they could expect from a non-branded search cluster and adjust bids accordingly.

Step #4. Adjust keyword bids according to assist value

The team began shifting its PPC budget toward the non-branded clusters that were delivering the most conversions.

They added the revenue attributed to those non-branded search terms when calculating their return on ad spend, to determine how much more they could bid on the entire non-branded cluster.

The team also reduced their bids on -- or eliminated -- keyword clusters that didn’t show a tendency to lead to branded sales and conversions.

Step #5. Track impact of bidding strategy changes

Once they had established new bid thresholds for non-branded clusters, the team monitored the ongoing performance of those keywords to determine if the new ad positioning was delivering results.

Key metrics they tracked included:

- Traffic from the ad

The team watched to see if the higher bid amount and positioning generated enough traffic to make the investment worthwhile.

- Sales conversions

The team also monitored conversion rates to see if the cluster continued to perform as they had predicted from previous tracking tests.

Proper conversion attribution and keyword optimization delivered significant improvements:
o Cost-per-acquisition for non-branded search terms decreased 25.6%
o Sales from non-branded search terms increased 15.2%

"We’re happy in two ways: We not only increased sales, we also reduced costs," says Godzich. "During a recessionary period, that’s a gift."

The data helped the team identify and remove some 20,000 non-performing non-branded terms.

Although they did not have to significantly increase bids on the strongest performers -- non-branded budgets increased from 70% to 75% -- the ability to weigh that spending toward the highest-performing clusters delivered the gains they were seeking.

Useful links related to this article

New Chart: Search is Generating the Bulk of B2B Leads...But How Good are They?

How to Use Niche Keywords to Boost Search Traffic: 5 Steps to a 20% Lift

WebMetro: Provided the team’s new analytics platform and manages their PPC strategy

American Management Association

Improve Your Marketing

Join our thousands of weekly case study readers.

Enter your email below to receive MarketingSherpa news, updates, and promotions:

Note: Already a subscriber? Want to add a subscription?
Click Here to Manage Subscriptions