Google recently launched a new report in AdWords: the geographic performance report. This is going to be a fairly brief post, so let me tell you exactly what I think about this report: if you have the Google Conversion Tracker installed, you are an absolute idiot not to check this report on a regular basis.
Could I make that point any more forcefully? Geographic performance will show you – down to the suburb if you want it to – the relative performance of different geographic regions. Let me give you an example of the power of this report. I have a client that has a service business in Arizona. They were buying keywords on a national level and appending the keyword with their location name (for example: “Tucson Chiropractor”). We bid lower nationally because the conversion rate was lower than a local campaign.
I ran the geographic performance report last night and did an analysis of performance by state and then rolled up the data to regions (i.e., West, East, Midwest, South). The south and the west were performing well, the midwest was slightly above average, and the east was absolutely horrendous. In fact, the east was 4oo% over the amount we were willing to pay for a lead. By pausing these eastern states, we’ll be able to raise our bids in the south and west. In other words, we’re no longer paying for bad traffic, but we’ll get more market share of the good traffic.
This geographic performance report is an absolute must. Congrats Google, you are once again light years ahead of your competitors. Now here’s two suggestions for the next iteration of this tool – first, enable geo-parting – allow us to vary our bid by state, city, or metro area. You could even offer a “radial bidding” tool, which would allow me to bid more for clicks coming from a short distance from my location, and less and the radius expands. Second, why am I only able to run a report on a daily basis and not summarize data up to the month or even all time? Someone at Google owes me a t-shirt for this suggestion!
Terry Whalen
September 5, 2008 at 5:55 pm
Word up – good post – there must be a funny story as to how this report can only be run on a ‘daily’ unit of time basis. That makes zero sense. But my Google agency contact assures me it’ll be fixed. For now, folks may want to import data into Excel and do a pivot table in order to look at consolidated data by geography.–Terry
David Rodnitzky
September 5, 2008 at 5:57 pm
Great point Terry. I meant to mention that pivot table idea – and thanks for telling your Google contact to get this fixed (while you are at it, ask them if they can change the time of day/hour of day report to include conversion data!)
nate
September 8, 2008 at 4:37 pm
Hi Terry,Regarding your point about geo-parting your bids…you can do that in AdWords, although it's a bit manual (especially if you intend to get really granular w/ your targeting).Set up your campaign & adgroup structure in a way that supports whatever your strategy is and then adjust your bids as necessary for that segment. For example, you could create a Campaign called Arizona Geotargeted and then a bunch of adGroups inside for specific metros like Tucson, Phoenix, Flagstaff etc. This could certainly be improved, but it works at a basic level.Nate
nate
September 8, 2008 at 4:45 pm
Clarification – I’m pretty sure geotargeting only works at the campaign level, not the adgroup level.
CustardMite
September 9, 2008 at 2:08 pm
My only concern is that if you look at the data, you’re bound to see some ‘patterns’ due to random variation. In your case, one of the regions was always going to be the weakest performer, even if the average performance of all of them were the same. I’m sure you made sure that the pattern you saw was statistically significant before making any changes, but I suspect that many people will start changing their bids on whichever regions are performing best/worst, without taking into account Data Snooping Bias.Here’s Wikipedia’s explanation – not sure that there’s much I can add…http://en.wikipedia.org/wiki/Data-snooping_bias
David Rodnitzky
September 9, 2008 at 5:48 pm
Great point custardmite (and I love your user name . . .). In the case I referenced above, the data was definitely significant – indeed, I looked at a 2.5 month time period and had thousands of clicks and hundreds of conversions with which to play.I do think, however, that the risk of data snooping is an acceptable risk – if the data is wrong 1 out of 20 times but right 19 out of 20, then I am willing to take the chance that the data will sometimes not show the real results.And I also believe that any bid management must be an ongoing approach that looks at your trailing data over a set time period – so if I make bid adjustments that totally mess up my ROI, over time my bids will re-adjust as the truth unfolds.