Keyword Testing and Expansion Done Right, Part 2

Finally, you’ve completed your keyword discovery. (If you haven’t, part one of this series explains how.) The next critical decision is determining whether to use a new campaign structure to roll out your new keywords or to feed the keywords into an existing campaign structure (using existing Ad Groups and creative in Google, for example).

Many of the new keywords you’ll add will be phrases if it’s a mature campaign being expanded. Thus, there may be a compelling reason to use an Ad Group that contains similar phrases or root words. My general philosophy in Google is if you really want to know if a new phrase has potential, put it in its own Ad Group with custom-written titles and descriptions. To determine if a phrase or word deserves that level of attention, I look at the following factors:

• Search impression volume. The minimum will vary by industry or client and has more to do with the rest of the campaign and the customer’s potential importance than any arbitrary number. I may not recommend a custom Ad Group for a dating site when a phrase gets only 50 searches per month, for example. But for a site that sells music equipment, I may recommend a custom listing for the phrase “marshall ms2 practice amp,” which probably gets about 100 searches per month across the major engines. A searcher seeing a custom-written ad for the item he seeks is very likely to respond. If the merchant/advertiser can fulfill that need, everyone’s happy.
• Customer’s importance. If a business-to-business client sells “double beam spectrophotometers,” I don’t care if that search term gets well under 100 searches per month or not. The searcher needs to understand the advertiser has what she seeks. A custom-written ad accomplishes this objective more effectively than a generic one.
• Resources. If you have internal or SEM (define) agency resources available to implement best practices, take advantage of them. Alternately, if you’re resource-constrained and must concentrate on how to get your campaign’s core keywords as efficiently as possible, temporary shortcuts on keyword expansion make sense.

One factor cited by those advocating the use of existing Ad Groups in Google for keyword expansion is the click-history/Quality Score issue. Though it’s true an existing Ad Group in an existing campaign structure may have a good Quality Score, if your new keywords differ from your existing keywords in regard to relevancy of creative, then a new Ad Group will generally perform better and more efficiently over time.

Another key factor in campaign structure in Google is themed grouping. Making sure an Ad Group is themed improves targeting (and therefore efficiency) in Google’s contextual AdSense network (should you opt in to it).

MSN uses an Ad Group-style method of clustering ad creative with keywords. In MSN, the “Order” is similar to the Ad Group, as many keywords share the same ad. Of course, in Yahoo there’s currently a one-to-one relationship between keywords and their ad creative. The next generation of Yahoo’s system may have a different structure.

Selecting Landing Pages

Regardless of whether you plan a sophisticated landing page test, you clearly have to start somewhere. Often, the winning landing pages for new keywords are the ones that currently win for you. So compare the new keywords to those in your current campaign and imagine the searcher’s intent. If you have already-live keywords that target a similar search intent, a good place to start in respect to landing pages is with the current winners.

Where to Start a Bid

Once your keywords are uploaded into an existing or new campaign structure, you must decide where to start bidding. Clearly, each keyword must have a bid range that meets your overall visibility, ROI (define), and long-term profit objectives. The trick is to learn as efficiently as possible where that bid range is. I mention a bid range because many marketers find themselves altering bids on a frequent basis or pausing campaigns to take advantage of dayparting conversion effects (large changes in conversion rate).

Your objective should be to bid as high as possible (given a specific ROI objective) to answer the question, “Will this keyword work for me?” Bidding too low has several negative side effects, including:

• A low number of clicks per day means you must wait a long time for sufficient data to accrue.
• It’s easier to convince the engines that use hybrid algorithms that your ads are relevant if you get a high CTR (define) early on (this shouldn’t happen, but it seems to).
• Conversion rates may be lower at the lower positions.

Look at your current campaigns and the CPCs (define) you pay for similar keywords. Boost that CPC as much as you think you can based on potential increases in conversion rate. Regardless of your bid strategy, testing new keywords is an investment. It may require significant monetary investment before you find the winners and weed out the losers.

Data Scarcity: Clustering by Keyword Stem and Landing Page

Keyword expansions often occur in the tail of the search distribution curve, meaning the keywords and phrases aren’t often searched. Given a CTR of 2-15 percent and a conversion rate of 3-5 percent (both aggressive estimates), it would take years to know if a keyword resulted in profitable sales. Data modeling can help. Cluster keywords by stem (root word) or landing page. The data you get by looking at the cluster isn’t perfect, but sometimes it’s the best you’ve got.

When Keywords Fail

You tried a keyword. The data look bad and the target ROI isn’t likely to be met, even at lower bids/positions. Do you toss it? Perhaps, but before you discard that keyword, look carefully at the ad and the landing page. A mismatch or any other factor that provides a poor user experience can sabotage a perfectly good keyword.

Go forth and expand your keyword list. But know ahead of time the monetary and HR costs associated with testing new keywords.

Meet Kevin at Search Engine Strategies in Toronto, April 25-26, 2006.

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