• http://twitter.com/larrykim Larry Kim

    nice write up, brad. thanks. couple of questions here. Say i have a campaign that was previously just targeting USA. Would you recommend splitting it up into 50 different campaigns, say one for every state, to find out differences in performance based on geography, then bid differentially based on ad performance on each location? Earlier when it required campaign segmentation i would have never given it a thought – way too much overhead. now, i’m tempted to try it out.

    Also, do you happen to know what happens with overlapping geographic regions. Like say i have one region as Boston, the other as Massachusetts. Does it resolve overlaps? how? (eg: pick the higher bid or pick the more specific region?)

  • http://www.searchermag.net/ Vance Woodward

    Excellent overview of some of the new tools (and challenges) now at our disposal. While I think it can sometimes be unclear what motivates Google to make these changes, I find they can generally be trusted to always be improving and strengthening their advertising platform. This is excellent to help people (especially new users) understand some of the changes currently being made http://www.searchermag.net/google-adwords-consultant/

  • http://twitter.com/Kevin_Lee_QED Kevin Lee

    Interestingly over the last two years the Didit team and I worked to perfect a system by which we optimize based on conversion rate, order size and LTC all based on geography (auto-cloning the master campaigns into hundreds or thousands of child campaigns all synched to the master). The essential element that predicted variation was audience (demographics a big factor). So, we were able to build a huge data warehouse that uses first and third party data to isolate neighborhoods at a granular level and big higher there. The new Enhanced platform just makes it easier to bid boost within a campaign as opposed to the cloning system. However, one can ten ad creative by geography and get the QS higher in a separate campaign making that superior in many cases.

  • Terry Whalen

    Hi Larry, you should be able to get geo-specific performance data without splitting the US-targeted campaign into 50 smaller state-targeted campaigns. Just go into Dimensions/geo and you can see the state-by-state performance. Then, after you’ve enabled the campaign as an Enhanced Campaign, target the campaign to each state (rather than leaving it set to target ‘U.S.’), and then you can tweak bids at the state level, based on the performance data you saw in the dimensions tab.

  • Terry Whalen

    Brad, this is written clearly – it helps to clarify all this stuff in my own head! Thanks!

  • http://twitter.com/bgtheory Brad Geddes

    Hi Larry,

    Terry mentioned how to get the data; so I’ll skip that one.

    As far as bid modifiers go; Google will try and use the ‘most specific one’. So, if you target Massachusetts, Boston, and Hyde Park (or a location extension radius); and the user in in Hyde Park, that bid should be used even if its lower than the other location’s bids.

    Now, it’s almost impossible to test this; so for now; I think we’ll have to take Google’s word for it.

  • http://twitter.com/bgtheory Brad Geddes

    I think Google is starting to realize that a lot of the crazy segmentation many of us do is for demographic purposes. I think we’ll see demographic bidding for search (its on display now; but its pretty small inventory) sometime this year and more bid modifiers roll out over the next year or two.

  • RenaldosWalkman

    Just to be sure, is this Location bid multiplier applicable to desktop bids in addition to mobile bidding? I had heard it was only for mobile bids.


  • http://twitter.com/bgtheory Brad Geddes

    The campaign bid adjustments are used in every auction regardless of device (if you’e -100% mobile; you won’t be on mobile). So, even on desktops, Google uses the location and ad scheduling bid adjustments.

  • http://www.LeadDiscovery.com/ Jerry Nordstrom

    Kevin – Have you and or your team been able to model campaign structures and potential outcomes in a visual manner with this data?