Weigh the costs and benefits of over-segmentation from geo-targeting
Published: November 28, 2012
Author: Sean Marshall
I love simple campaign structures. If you’ve read my blog before, you’ll know that I’m a firm believer in streamlined accounts w/ few campaigns (I’m willing to bet I used “firm believer” in this same context in prior blogs. Apologies to you and my editor). The question is, what do you do once you’ve made it so simple? Well, make it complicated again.
The real advantage in having a basic campaign structure is deploying big, gnarly, account-level optimizations that would otherwise create clutter – perhaps none of those optimizations more important than geo-targeting.
The edges you can gain from geo-targeting usually fall in these buckets:
1) More relevant (localized) messaging
2) Bidding efficiency to either:
– Hit different CPA/ROAS targets
– Make up for CPC or CVR deltas by market
– Hit budgets
Maybe splitting #2 into three bullets is cheating, but it usually comes down to messaging and economics. Analysis tools are strong enough that you can see how different segments perform without breaking up campaigns so those two are all that’s left.
The challenge in all this is figuring out what warrants segmentation. I think you could make a case that any ad could benefit from some localization. Even things that have nothing to do your location can be aided by this – it’s about relevance. Problem is, text ads might not be the best medium for this. It’s one thing to show me a commercial of the new Buick Tank5000 driving over the Golden Gate Bridge and down the Embarcadero; it’s quite another to tell me that Dyson vacuum cleaners work even better in your neighborhood than others (they don’t). So what kind of criteria should you use?
Like all things PPC, make it about the numbers. By now, I hope you know if location has an impact on how you pitch your product (no, Dyson vacuums don’t work better in San Francisco), so it should now be a matter of how much it impacts things. Best way to find out – take a single large city (90% of the time it’ll be New York) and break it out in a different campaign. This is your testing ground.
It’s safe to assume that, if your product benefits from a local message, you’ll see gains in the NY market by breaking it out and customizing the messaging. Now it’s time to quantify. While I wish I had a magic formula for this, it comes down to counting time invested and comparing that to the impact your optimization has on performance.
As with all segmentation, there are major trade-offs. Every campaign you add (most importantly – duplicated w/ geo settings altered) results in more work. While it’s not quite double, you need to worry about the same things (budgets, URLs, ads, KWs, negatives, etc.) many times over. If you do it in one place, you’re doing it everywhere. Wait…did I launch that Philly KW in the Austin campaign, er…or was it mobile? Now it’s time for a gap analysis and possibly relaunching terms, using time you could have spent on a new landing page – driving the same incremental improvements you got from your geo work, in a fraction of the time (possibly).
Now, you can create great process to mitigate this and, combined with technology, really limit the extra time needed to manage the details of these duplicated campaigns. There are a couple of other things to consider.
Localization can be a black hole. There really aren’t any limits to what you can do with proper geo-targets – custom offers chief among them. The problem with custom offers and variants isn’t that they don’t drive improvements; it’s usually that those improvements can’t be quantified. While it’s one thing to pull a dimensions report and evaluate the performance of a single ad in 5 states, it’s impossible to compare your 20% off offer in Georgia vs. free shipping in Texas. Again, you can create robust process around A/B testing these ideas in all of these places, grounding your now ultra-varied, customized offers in each market in data, but that’ll lead you to the last problem: statistical significance.
Data is a funny thing. It tells compelling stories about our customers and how to effectively market to them. But when data is scarce, it becomes less compelling, and tests are impossible to judge. Therein lies the ultimate curse of geo-targeting: things can get too segmented.
Imagine you’ve found a scenario where localizing ads has a clear impact on ROI. It’s a data set with absolute certainty, which then leads you to implement the most comprehensive geo-targeting scheme possible and leveraging this finding in as many markets as Google will allow. You’ve probably created something so complex it’ll be impossible to optimize. Is your plan to roll up performance by geo and duplicate bids in all campaigns? Do you plan to turn your KW launches into 10-hour exercises, requiring every single ad to be customized? How will you run definite ad copy tests without rolling data up again?
At the end of the day, there are tradeoffs to anything you do in paid search. A simple two-part question I like to ask myself whenever I plan client roadmaps is this: how much of the account am I impacting by doing this, and how big of an impact can I expect? So long as I can impact 50%+ of the account and expect double-digit deltas, that’s what I’ll do. And you should too.
– Sean Marshall, Director of Client Services