If you were an google engineer, how would you?

approach the targeting problem..

[loanuniverse]

It just ocurred to me that the best way to gauge targeting is through CTR. If you make a chage to your targeting {recipe/formula/algo} wathever you want to call it, and then find an statistical drop in CTR across the network then the best reason behind this drop has to be a decrease in target quality.

[runboard]

Not to play devil's advocate, but maybe they use/used conversion as targeting indicator? After all, conversion is what matters.

I see why an ad for Playstation games would convert better on a Christmas gifts site, but get a higher CTR on a warez site. (Just an example obviously).

[icedowl]

Conversion may be what matters, but when the targeting is so far off (as I've been seeing) they'll never get a chance. First step to conversion is for the ad to be interesting and targeted so that it gets a click. Add some variety in the mix too, another thing that seems to have been forgotten. I keep seeing the same-o same-o ads and they get old real fast.

[loanuniverse]

You guys make a good point about conversion, but being completely honest while it would be nice to have a high conversion rate from my ads, I would much rather have a higher CTR. :)

To tell you the truth, I think that the Google geeks {err engineers} are in a better position to determine what should work best {I hope they have the input of a couple of business minded people when they brainstorm}. I also hope that they keep the interests of the publisher's as a high priority when deciding such things. {I am not saying the highest, I mean one of the big priorities}

[nyet]

####
I also hope that they keep the interests of the publisher's as a high priority when deciding such things.
####

Not many customers keep their vendor's interests in mind much. It usually works the other way around.

####
if you make a chage to your targeting {recipe/formula/algo} wathever you want to call it, and then find an statistical drop in CTR across the network then the best reason behind this drop has to be a decrease in target quality.
####

This is just a guess on my part, but because of smart pricing, the new algo must have to decide which sites are more likely to convert for the advertister (to lower bids automatically). I'd bet that this is related to the bad targeting of late. My guess is that the new algo has to 'learn' how targeting affects conversions and I suspect over time it will get better at it.

I think patience is in order....

[edited by: nyet at 3:39 pm (utc) on May 7, 2004]

[nyet]

[double post]

[novice]

"If you were an google engineer, how would you?
approach the targeting problem.."

Allow publishers to choose keywords for their pages. Then figure a formula to pay the publishers a percentage based on ctr or conversions.

[john_k]

Allow publishers to choose keywords for their pages. Then figure a formula to pay the publishers a percentage based on ctr or conversions.

Add to that:
1) Google provides a list of the keywords it has used to match ads to your site and then allows you to block any from the list. This process would probably not need any human intervention by Google.

2) Google allows you to suggest additional keywords that they should use for ad targeting. This would probably require some manual review process by Google.