| 5:42 am on Oct 30, 2007 (gmt 0)|
I tried it but after typing in car sign my search for new york did not bring up any ads relating to vehical signs or any combination of vehicals or signs.
[edited by: Kobayashi at 5:43 am (utc) on Oct. 30, 2007]
| 5:46 am on Oct 30, 2007 (gmt 0)|
I tried it as well, I don't seem to get the result you mentinoed. It just give me what I expect fro Car Sign and New york independently.
Could they be testing it on you? :)
| 6:07 am on Oct 30, 2007 (gmt 0)|
I tried your example too and couldn't replicate the results.
| 6:19 am on Oct 30, 2007 (gmt 0)|
There is an article on it that talks about adverts now being based on 3 factors...
1. Viewers IP
2. Page content
3. Your browser history...
On the Adsense help site people are complaining about getting adult adverts, when its actually a factor of their history rather than their page content.
| 8:22 am on Oct 30, 2007 (gmt 0)|
You agreed to that when you installed the Google toolbar. You can also disable it but for best results let it gather data over a few months of searches, it gets better.
If you share the computer disable it, other people searching for other stuff will make it learn a mix of stuff that you don't want. Disable it at work too, especially if you work for a sign company.
| 11:36 am on Oct 30, 2007 (gmt 0)|
Oh well, I guess they're just targeting me! I guess I should be paranoid. I don't have the Google tool bar, by the way.
Here's another one:
query #1 - "california redwoods"
query #2 - "used cars"
I received 3 ads for "Los Angeles Used Cars", along with some decently geo-targeted (East Coast) ads.
All I'm saying is that whoever is bidding on "Los Angeles Used Cars" is going to have a disappointing CTR, because there is no real correlation between my research on california redwoods and my research on used cars.
| 11:43 am on Oct 30, 2007 (gmt 0)|
rehabguy, you are entirely right; and this has been brought up a few times in the past (I recall at least a month ago, but I can't find the thread at present).
It is a predictable effect and has nothing to do with the Google Toolbar.
Another example for the mix: search 'car' then 'light'. Now 'light' again. Notice first 'light' search includes vehicle / car lights; second does not. Tested with google.co.uk. Additional bonus example: 'insurance' then 'car'.
It works particularly well when a given word can have multiple meanings (or advert clusters?); it takes context from your previous search term.
| 12:26 pm on Oct 30, 2007 (gmt 0)|
behavioral marketing = BM.
the human mind just ain't that simple, trying to anticipate what someone wants now based on what they said they wanted before is fruitless... 21 years of marriage speaking here.
| 3:52 pm on Oct 30, 2007 (gmt 0)|
I completely AGREE behavioural concept is NOT good. At least for now.
This is like when you put a banner on your site and less than 1% of visitors actually click on it, even when you believe there is a connection between the content and what that banner offers.
If Google is doing this on regular base, they should exclude those impressions from CTR and quality score, just like they exclude partner search networks.
| 4:03 pm on Oct 30, 2007 (gmt 0)|
Google already takes the ad position into account when determining the quality score, so I would hope that they do something similar for the ads that are driven by their behavioral targeting "feature". Maybe one of the AWA's can clarify that.
| 4:53 pm on Oct 30, 2007 (gmt 0)|
|Check out Google's new "behavioral targeting" system: |
Do a search on "car sign", then do a search for "new york"
Your search for new york will bring up ads for "New York Vehicle Signs" even though you may have been really looking for information about New York in general.
I see it exactly how you described it.
Expanded broad match + behavioral targeting = lower ROI for us and more $$$ for google.
| 2:12 am on Oct 31, 2007 (gmt 0)|
Google will not fully implement something like this without hard evidence it improves the overall relevance of the algorithm. Will it be right every time? of course not.
There are many instances where subsequent queries are related especially if they are within a certain timeframe.
| 2:20 pm on Oct 31, 2007 (gmt 0)|
"There are many instances where subsequent queries are related especially if they are within a certain timeframe."
I agree. But there's also many unrelated queries that occur within that same timeframe. And there's no way for a machine or person to know which is the case without asking the searcher. If you want to allow people to drill down, add a feature that lets them choose to do so.
I think this will have to go way too far before all the people now writing reasearch papers on it and those salivating over demographic data and analysis get a decade or so to completely screw this up. "Precise targeting" is now sold by spyware and adware companies galore, everyone wants to own my desktop - consumers will eventually demand that parties stop listening and observing and recording every move they make. But, for now, it's buzz, good, buzz, social, buzz, target, buzz, demographic cartesian cross-section, buzz...
The superiority of asking people to tell you what they're looking for is plainly evident before us - the search engine wars and their billions in profits are proof. They're all thinking they can do better if they observe us and make guesses about what we want... anyone who has ever been an employee knows the difference between your boss asking what you may want and need to get your job done better -and- one who partially observes some aspects of what you're doing and then makes all kinds of nutty conclusions about what you need and want...
| 3:00 pm on Oct 31, 2007 (gmt 0)|
What do you expect from a bunch of idiot PHDs
| 3:52 pm on Oct 31, 2007 (gmt 0)|
I'd like to know how Google puts this in your stats? If Google uses your ad for a search that is not in your Keywords because of their BM - Where will it show up in your stats?
| 4:15 pm on Oct 31, 2007 (gmt 0)|
|If Google uses your ad for a search that is not in your Keywords because of their BM - Where will it show up in your stats? |
Just check the Average CPC and Quality Score columns.
I sure miss the good old days of linear bidding. That way, the market could just figure it all out.
| 2:53 am on Nov 3, 2007 (gmt 0)|
Broad Match of late has gotten VERY broad match. That is why if you are going to be bidding on broad matched kw's, you need an outrageously huge negative keyword list...AND, you need to continuously add to that negative keyword list on a monthly basis by running "search query" reports...You can make Great money with broad match bidding...but you can make boatloads more by adding those hundreds of negative keywords over time...
| 4:09 am on Nov 3, 2007 (gmt 0)|
I understand the science of behavioral targeting, but I find it a little too clever, a little too sneaky. Once you know why "incredibly relevant" ads suddenly appear on websites you visit, when they're not relevant to its context, usually I am more inclined to think, "nice try" instead of "I need to click on that ad."
| 4:22 am on Nov 3, 2007 (gmt 0)|
I get the ads described. I would think that only broad match would be caught up in this.
Edit - I also tried "penguins" and then a major car brand and make. Got some ads in lower positions for hockey stuff. That is totally crazy. Surely they tell that the two searches have zer relationship and thus should not pull up ads for each other.
| 4:54 pm on Nov 3, 2007 (gmt 0)|
Yes, behavioural targeting is way ahead of it's time. I mean, the algos should (with some level of accuracy) be able to determine if two consecutive searches are related...or not related at all. But it seems they can't even determine that? It's a great way to move forward, but don't shove it down our throats until you've got it right, Google et al ;)