Internet Marketing Blog

I’ve been doing keyword research for a client and found this little gem of a bug (feature?) in the Overture’s Keyword Suggestion Tool. Try searching on “space background” and compare the results to what you find with “space backgrounds”. Note the first result in each case (as of May 31, 2006) shows 70,799 searches. Well, that’s as we’d expect since it stems plurals. But look at the keyword: one includes “my” and the other doesn’t. The popularity of “My Space” is being combined with the results. Not a surprise on the one hand, but it could be very misleading on the other. It is very odd however that it should show “my” in one case and not the other.

If you are only using Overture’s tool, you could be getting a very misleading picture of search engine user behavior. I highly recommend using a service like WordTracker to at least augment this data. An even better technique is to use AdWords with exact and phrase matching (broad matching provides data that can be as misleading as Overture’s).

I took a brief look at Google Trends in a previous post. A recent blog comment over at SEOmoz has me thinking more about its potential utility.

Stemming

The comment points out how Google Trends is a great stemming research tool using this example. While the trends tool doesn’t provide absolute traffic statistics, their is substantial value in knowing the most common variants so that we can incorporate them into SEO efforts. For instance, if you’re selling digital cameras, this graph strongly suggests sticking to the singular form. Here is one more interesting example.

De-obfuscation

Many keywords have more than one meaning or use, and we are only interested in the one that is most pertinent to our business. With Google Trends we can test whether or not a general term correlates stronlgly or weakly with a more specific term. As an example, suppose we want to know the relative popularity of cable versus DSL. Unfortunately “cable” has so many meanings that we don’t really know if search statistics accurately represent demand for cable Internet service or if there is substantial demand for cable tv, cable ties, undersea cable, etc. We can tease out more information with a more specific query. If we compare the first query and this new one, you’ll see many similarities among the peaks and valleys, indicating a pretty good correlation between the word “cable” and our intended meaning, thus we can cast a less critical eye on the cable versus DSL data.

Correlation Testing

Suppose your are considering extending your advertising to reach people who are looking for products and services that might complement your own. You might be selling patio furniture and looking for a way to reach homeowners in the late spring and early summer; perhaps there is an opportunity with families that send kids to summer camp. You can test the timing like this. Since the timing is nearly perfect, it might be worth advertising to parents of camp-age kids.

AdWords is about to roll out click-to-play video ads. Unlike the rich media advertisements and banners you see on Yahoo! and other content sites, the new video ads will not play until a user clicks on them.

The video ads share many characteristics with AdWords’ existing ad formats — text, image, and Flash — including:

• Ads will compete with others through the AdWords content network.
• Advertisers can bid on a CPC or CPM basis.
• Ads can be distributed by both site- and keyword-targeted campaigns.
• Ads can be geo-targeted.

The new video ad format, like images and Flash, provides advertisers with a much stronger branding opportunity than text-ads do as well as an opportunity to present product demonstrations and complex messaging to potential customers.

To get the best performance from your AdWords campaigns, it is important to use the keyword matching options to target the specific phrase variations that are the most appropriate and cost effective for your business. We have created the AdWords Keyword Matching Options Generator to quickly and easily create the matching variations for a list of keywords. Give it a try and let us know what you think.

AdWords has recently added a new method for ad rotation that spreads ad impressions out over the course of a day.

Previously your ads would be shown until your daily budget ran out. If your daily budget wasn’t high enough to cover the entire day, the ads would stop running. Among other problems, this resulted in west coast markets being exposed less to advertisements due to earlier exposure to the east coast and corresponding depletion of the ad budget part way through the west coast day.

AdWords now provides an option to spread the budget over the course of the day, showing your ad intermittently to try and preserve your budget while still keeping some impressions running at all hours. This new method is presumably based upon past campaign performance, which means it may take some time for the system to “figure out” the proper spacing for new campaigns or after significant changes.

We are just starting to implement the new ad placement strategy, so we don’t know yet how effective it is at hitting the daily target by the end of the day, or if ad spacing really is as even as we’d like (it may need to “catch up” on slow days with many impressions late). If we see notable behavior, we’ll write about it here in the blog.

Rumor has it that Microsoft has bought DeepMetrix, a web analytics firm that competes with ClickTracks, WebTrends, etc. No press release at the moment that I’m writing this, but this post in Threadwatch suggests that it will be rolled into MSN adCenter.

Search Engine Watch has been playing with the new Google Trends tool.

I was eager to see what kinds of numbers it provides. Answer: none. Everything is shown relative to itself, no absolute numbers, no keyword versus keyword statistics.  However, if you want to know what cities, regions, or languages dominate a keyword, you can now see this. The tool also clearly illustrates seasonal variations and responses to major events, without actual search volume numbers.

Some things I just learned:

• Vancouver and New York City are the biggest chocoholics
• Shalom is surprisingly popular in Portuguese
• Chrismas Trees are most popular in December

Google has made their traffic estimator accessible without an AdWords login. The tool is a bit misnamed in that it estimates ad clicks and costs; it does not provide an estimate of the number of queries.

It may be possible to elicit some useful keyword statistics from this tool, particularly the relative popularity of different search terms. Using a very high bid will provide a top-position estimate of the number of clicks for any keyword, and this quantity is a small percentage of the total number of searches on that keyword. If — and this is the critical assumption — the percentage of searchers who click on ad that is assumed by the tool is the same for every term, then we have a measure of relative keyword popularity. Without the percentage, we can’t know the absolute number.

Let’s hope they have built in enough capacity for the flood of initial use and inevitable creation of automated screen-scraping tools that will try to take advantage of it.

Other tools:

• WordTracker
• Overture Keyword Selector Tool (free)
• Trellian Keyword Discovery

What should be the proper nomenclature for advertising on search engines? Corollary: what does Search Engine Marketing mean? I am personally in favor of a semi-concerted effort among those of us who practice search engine advertising to call it that (SEA), not search engine marketing (SEM).

Just as marketing is much broader than advertising, search engine marketing (SEM) has historically been and should, in my opinion, be a broad term covering various disciplines like search engine advertising and SEO. Initially this seemed to be the usage promulgated by practitioners, but lately it seems to me that the media and online marketing neophytes have latched onto the idea that SEM refers specifically to advertising programs like AdWords, Yahoo Sponsored Search, and MSN adCenter.

What are your thoughts?

We have found reporting errors in another Yahoo Sponsored Search account that we manage. The report center reports are clearly missing the vast majority of clicks, in one case showing $11 in fees for a month in which more than $750 in charges were accrued. This is the 2nd time in as many months that we have encountered this issue.

There is also a interface bug on the first of every month on the main startup screen. Selecting Month-To-Date refers to the prior month, and Previous Month will summarize the month before that.

Yahoo’s support team is aware of both issues.