How Accurate is Google's Customized Location Targeting?
A few weeks ago I did a little test to compare Google cost and traffic estimates to actual keyword performance using historical data from one of my accounts and comparing it to data provided by Google’s Traffic Estimator for the same list of keywords. For the group of keywords I used, the actual average daily cost over a 30 day period was roughly half of Google’s low daily cost estimate and the high cost estimate was just over three times actual cost.
Of course, I was optimizing the keywords during this 30 day period to meet my ROI goals, which may explain the discrepancy between estimated and actual cost. It is likely that had I just tossed my keywords in an ad group and let them run wild for a month, the estimates provided by the Traffic Estimator would likely have been closer to actual keyword performance during that period.
I was surprised to find out a few days ago when I was pulling some estimates for a potential new client that customized location targeting is even less accurate than estimates pulled at the national level.
To illustrate, below are estimates for 4 different geographic areas for the same group of keywords:
Entire United States
Largest Metro Areas in PA (Erie, Harrisburg, Altoona, Philadelphia, Pittsburgh and Scranton)
15 mile radius around Danville, PA
For a larger list of keywords, search volume estimates are more valuable for getting a general idea of which keywords are likely to generate the highest click volume and may warrant having their own ad group to capitalize on Quality Score, assuming a good CTR can be achieved with highly targeted ad copy. In this case, there isn’t too much variation among keyword search volume estimates, which range from 0-3 across all 4 areas.
CPC estimates are roughly the same for all 4 areas pulled. I’m no expert on tutus, but $0.90 seems pretty reasonable to me.
Estimated Clicks/Day is where the data starts to get a little wacky. Obviously it isn’t possible for more clicks to be generated daily around tiny little Danville, PA than in the entire state of Pennsylvania.
Cost estimates are equally as unreliable.
To be fair, the Traffic Estimator does have a disclaimer on the results page, letting the user know that “All estimates are provided as a guideline, and are based on system-wide averages; your actual costs and ad positions may vary. To view estimates based on your keywords’ performance history, use the Traffic Estimator within the appropriate ad group.”
Does pulling estimates based on performance history generate more accurate data? Nope.
Here are the estimates I got in one of my ad groups when I ran the estimates from within that ad group (using the previous interface):
Just in case you can’t see that image clearly, the estimated search volume is 0 and there isn’t enough data available to estimate CPCs. Actual performance for this same group of keywords over a 30 day period was as follows:
Clicks: 269 (about 8 per day)
Average CPC ($2.10)
So it seems that pulling estimates at the national level (and considering the estimates to be high) is the best way to get useful data from the Traffic Estimator.