Geo-Targeting: An Under-Utilized Lever in Search Marketing

11/11/2010 11:46:00 AM


If you talk to most search engine marketers about how to most effectively control performance, they are likely to talk about 4 main levers of control. These levers most likely are (in no particular order of importance):

  • Keywords
  • Cost per Click (CPC)
  • Ad Text
  • Landing Pages
Marketers can evaluate each of these variables over time (using data) to figure out the most effective combination that will provide the most efficient (and scalable) results. Some sophisticated marketers also tinker with “day-parting” but I wouldn’t put this in the mainstream of variables as this often limits scale in the goal for effiicency.

It’s also true that a number of retailers use geo-targeting in Adwords - but largely for “Online to Store” testing, in-store promotions, new store openings, or something similar. And while I’m probably most passionate about Retailers under-valuing their SEM efforts from the lack of inclusion of brick-and-mortar store benefit -- I also think that traditional retail SEM can be enhanced through the use of geo-targeting.

It’s for these two reasons that I think a key trend in 2011 Retail SEM is going to be geo-targeting. Why not get a head start on this and begin implementing some levers during your Holiday 2010 initiatives?

Recently, we’ve worked with retailers who are looking to grow top-line e-commerce sales but keep marketing efficiency constant. We recommended taking a more local approach (City or State) to SEM rather than their traditional “nation-wide” view. While this definitely takes a bit more elbow-grease to execute, we’ve seen great results.

As an example, when analyzing conversion rate and average order value, we saw that one particular retailer was perfoming 40% better in one city (say, Washington DC) over the national average, yet their MAX CPC’s were set at the same value nationwide. While the actual CPC may vary in each city - the max CPC set by the advertiser, theoretically should be 40% higher than the national average. By bidding more aggressively in areas where we knew conversion rates (or average order values) were higher - we could meet the objective of growing volume but keeping ROI the same.

After seeing this data, we picked a small group of markets (5-10) to test the theory out. We created separate bids and budgets for these high-performing areas and ran campaigns for two weeks to see if the increased bids did in fact grow volume while keeping ROI the same. We were not surprised when our hypothesis came true as we grew revenue a similar ROI. We could simultaneously lower MAX CPCs in poor converting areas (or low AOV) - but decided to wait to do this for the next round of testing.

A few tips on this type of testing based on our experience:
  • Identify good geographic areas of testing: look at Conversion Tracking and/or sales data broken out by state of city and identify top performing areas; cross reference that with Google Insights for Search to see which states or cities index high for specific categories or keywords and look at front-end metrics such as clicks or CTR to see which categories or important products resonate best with consumers.
  • Choose a category / top volume keywords that historically convert well - and be sure to break this out with it’s own budget and bids.
  • Make sure to compare similar products/keywords nationally vs market, rather than looking at the account in aggregate.
  • Be sure to mirror settings from the national campaigns in your newly formed campaigns (e.g. include site-links, product extensions etc. in your geo-targeted campaigns)
  • Be aware of any budget caps that may limit the potential of the test.
  • Start small (5-10 markets) - Iterate, test, and expand.
  • Try testing call-outs of city names in the ad copy to see how CTR is effected.
  • In our experience it’s probably better not to exclude test markets from the national campaign - especially if you are increasing max CPC of the locally targeted ones.
  • Try expanding local reach by geo-targeting with Display to add in another layer/medium for relevant presence.