It is getting more difficult to control ad display in Google AdWords because they’re giving the system more liberties. There are a few variations of this – I’m calling them “Ultra Broad Match” because it’s funny:

Automatic geo-targeting parsing of queries.

If a visitor searches a general term and Google pinpoints him/her to a location that is in the target-area of one of your campaigns, the query is sometimes treated just as if it had geographic modifiers attached to it. Likewise, if you are bidding on a term that has geographic modifiers, Google might decide to broad match it to a more general query that does not have modifiers attached. For example, if someone is searching for “plumbers” from an IP address that is located in Detroit by Google’s system, then it might be treated just as if he/she searched “plumbers in detroit.” Likewise, if a person searches for “plumbers in detroit,” he/she might be treated just as if he/she searched for “plumbers” from an IP address located in Detroit. In your account, you can be explicitly bidding on the term “plumbers in detroit” in a nationally targeted campaign, but Google may decide instead to show an ad for the term “plumbers” that is in a campaign targeted to the Detroit area. For an advertiser, this means it is very difficult (if not practically impossible) to keep terms that have geographic modifiers separate from terms that do not have geographic modifiers when setting bids.   This is a problem because you might expect a broad matched keyword phrase like “plumbers in detroit” to have better conversion rates than a broad matched keyword like “plumbers” – even if the more general term is in a locally targeted campaign. I can think of at least four reasons why the geo-modified term should deserve a higher bid for a local business:

1. Locating by IP address isn’t always accurate and not particularly specific. Google thinks Pure Visibility’s IP address is located in Detroit, not Ann Arbor, for example. But Pure Visibility and its internet users are definitely in Ann Arbor. So general terms might not always be hitting the right location, whereas a query that explicitly specifies a location is going to be accurate all the time.

2. General terms like “plumbers” don’t express as much commercial intent as terms like “plumbers in [insert town].” For a query with a location specified, it’s probably safe to assume that a search engine user is looking for products/services in that area. For a query that does not specify an location, a visitor could be looking for something entirely different: pictures of plumbers, a list of plumbing associations, an official definition, the name of a newly released Super Mario Brothers film, whatever.

3. General terms like “plumbers” might be more apt to broad match less relevant terms. Thus, giving general terms a higher bid to match the implicit geo-targeting that goes along with them will raise account costs because of less relevant matches.

4. The fact that someone is searching from a particular area does not necessarily indicate that they’d be more interested in products/services in that area. As an example, take the term “vacation resorts.” Who’d want to vacation in the same town they live in (besides me)? Most likely a search engine user using this query would be interested in a resort somewhere far away from where they are.

You can try to battle Google’s automatic parsing by adding campaign level negative keywords. You’d add town names to regional campaigns and exact negative matches of general terms in national campaigns, although I haven’t gotten verification from Google that this will work well.

Expanded broad match of related terms when the exact match of the query is being bid on.

This one is particularly surprising. Running search query reports, I found ad groups matching terms that other ad groups in the same campaign were explicitly targeting. For example, a search engine user searching “toddler care” might be seeing an ad for “infant care” from an ad group with only terms that contain “infant,” even though an ad group with ads about toddlers was bidding on the exact term “toddler care.” If you’ve never seen this page from Google, and you use AdWords a lot, you might want to bookmark it:

http://adwords.google.com/support/aw/bin/answer.py?hl=en&answer=66292&type=f

It changes a lot. Basically it describes how Google decides to show ads. Recently Google added this gem:

Session-based ultra broad matching.

According to Google, “Broad (Session-Based)” matching is this:

When determining which ads to show on a Google search result page, the AdWords system evaluates some of the user’s previous queries during their search session as well as the current search query. If the system detects a relationship, it will show ads related to these other queries, too.

I cannot block enough session-based queries with negative keywords. This is where you get the most bizarre and irrelevant matches. And in the rare cases they are useful and converting, it’s actually a match that would exactly or broad match something you are already bidding on. I don’t understand why Google continues to use this one – there should be an option to turn it off. Especially if your bids are very, very high. When you bid $20+ on a term, you expect the queries it is shown for to be matching pretty closely or at least very closely related thematically, and in this respect session-based queries are a disaster.

Google Ultra Broad Match

In the battle against the Ultra Broad Match, search query reports and possibly a web stats program that isn’t Google Analytics (at least Google Analytics with auto-tagging enabled) is your best bet. Even then, there are going to be problems that arise because AdWords’ system is taking a ton of liberty matching queries – noticeably more than it did before. The more the system relies on the quality score algorithm and the less it relies on advertiser input, the crazier and more frustrating the matching gets.

Tags: Google, Google AdWords