Geo-granularity in Paid Search

We all know by now that search is an effective channel for local marketing. Marketing through search at the local level is best practice for businesses with physical locations (and even for some without), but doing that at scale can be tricky. If you have a business with multiple locations, you’ll want to ensure that people searching in a particular area get served an ad that calls out the city they’re in. Ideally, you would show offers for that city and drive to the landing page specifically for that city. The most straightforward way to accomplish this is to create one campaign for each city and set the geo-target to that city alone.

geo-targeting (2)

So what if you have 1,500 cities? And then what if you want to split your budget by brand vs. non-brand? This means you have to create 3,000 campaigns. If you have 50 ad groups per campaign, you’re now dealing with 150,000 ad groups and potentially millions of keywords. This can start to get pretty messy for day-to-day management.

The level of geo-granularity when it comes to structuring paid search campaigns is an on-going struggle. As the space becomes increasingly competitive, appearing as the most relevant ad on the SERP is critical.  To accomplish this, should you geo-target strictly at the city level and create one campaign per city? Should you create one campaign at the state level and geo-target only the list of cities where you have locations? Or should you target the entire state and separate your locations using ad groups?

I tested two of these scenarios.

Test 1:

For a company with 40 locations in the state of Texas, for a 14 day period, I created one campaign for each city and set the geo-targeting for each campaign to that city only. For the next 14 days, I created one campaign, and geo targeted the same list of 40 cities, but all within one campaign. Keywords, ad copy and max CPC bids stayed the same.

Metric 40 Campaigns – 40 Cities 1 Campaign – 40 Cities % Change
Impressions 37,533 68,232 81.79%
Clicks 711 1,353 90.30%
CTR 1.89% 1.98% 4.68%
Avg. position 2.2 1.7 20.83%
Leads 12 123 925.00%
Conversion Rate 1.69% 9.09% 438.64%
Search Impr. share 62.41% 61.71% -1.12%
Search Lost IS (rank) 37.12% 31.51% -15.11%

The results in this test show that having one campaign with multiple geo-targets yielded better results in this case. Gains were recognized in impressions, clicks, CTR, average position, lead volume, cost per lead, conversion rate and lost impression share due to rank. The down side is a slightly higher CPC and higher impression share loss due to budget. However, the results for the last 14 days were much more desirable overall than those from the first 14 days.

Test 2:

For a company with 64 locations in the state of California, for a seven day period, I had one campaign geo-targeted to all of California. For the next seven day period, within the same campaign, I changed the geo-targeting from all of California to the list of 64 cities. As with the first test, other variables remained the same.

Metric 1 Campaigns – 1 state 1 Campaign – 64 Cities % Change
Impressions 76,827 52,083 -32.21%
Clicks 882 879 -0.34%
CTR 1.15% 1.69% 47.01%
Avg. position 2.1 1.9 -12.53%
Leads 72 99 37.50%
Conversion Rate 8.16% 11.26% 37.97%
Search Impr. share 50.73% 59.94% 18.15%
Search Lost IS (rank) 49.27% 40.06% -18.69%

In this test, the differences in results are not as dramatic, but they’re still there. With far fewer impressions, we were able to maintain the same number of clicks, which means a higher click-through-rate and higher conversion rate. Cost-per-click also came down, average position went up and most notable are the gains in impression share and the drop in lost impression share due to rank.

In summary, from these two tests, we see that creating one campaign for each city may be too much of a granular choice for geo-targeting. At the other extreme, broadening the geo-targeting to cover the entire state may be too broad. The reasons why this is happening are completely open to speculation. The former may be severely restricting the number of auctions you enter, while the latter may be opening you up to too many auctions, reducing your relevance/QS, and increasing costs.

The middle-ground seems to be the sweet spot here. In both tests, one campaign with a list of city level geo-targets had the best results. What kind of results are you seeing from your geo-targeting testing? Please share in the comments below.

Adiela Aviram

Share on FacebookShare on Google+Tweet about this on TwitterPin on PinterestShare on LinkedInEmail this to someone

Related Posts:


  1. Norm
    2013-06-22 16:27:00

    Doing this analysis lays a foundation for significantly enhancing results at scale. It would be interesting to see how the local results differed by device. Think of the messiness now. Nice analysis.

  2. Scott
    2013-05-15 20:44:53

    This is an important topic. I think the degree to which local is important to a given client is also a factor. Another interesting test would be to look at city versus market-level targeting for bricks and mortar-focused retailers.

  3. Leah
    2013-05-14 18:26:02

    Wow, Thanks so much for the great insight Adiela. I guess it is all about finding the right balance between broad relevance and granularity.

  4. Natalie
    2013-05-14 15:42:07

    Very nice! I agree with finding a middle ground with geo-targeting. I've also found the best results with bucketing my geo-targeting at the city level and having ad groups that speak to those. At the end of the day, I think it comes down to relevance. When you target at the city level you have better (if not total) control of how you appear to searchers in those cities.

Leave a Comment