PPC Geo-Bidding, Simplified

Recently, I came across a social media post that — honest to goodness — went on at length about the price the poster would be willing to pay for a new brand of ketchup.

“I’d need 50 percent off to give it a try,” opined the timid tomato tester. “If I turn out not to like the product, it’s just a whole lot of wasted time, and a waste of food .”

“This calls for snark!” some alien voice urged me.

I’m not proud of it, but there was only one possible reply: “Dude. It’s just ketchup.”

Unlike the man-child clinging to the safety of the almighty Heinz brand, some people actually get paid to overthink stuff. That appears to be the case in the marketing profession.

Despite some underlying complexity in the permutations, probabilities and platforms, there are many topics in digital marketing today that could use a little dumbing down — at the very least to overcome paralysis, but also to avoid doing fake work or engaging in “doctor that actually makes the patient sicker” activity (of the type Nassim Taleb has been so eloquent in calling out).

Bidding accurately on different geographic segments is one of those topics. Let’s go.

Dive In! See, You’re Already Swimming!

In AdWords, assuming you set up your main location (i.e., your nationwide “catch-all” — say, United States) along with a few additional sub-locations of interest (a handful of cities or states, using the +LOCATIONS button), you’re already geo-targeting.

Plus, you’re already geo-bidding as soon as you enter your first bid adjustment (adjusting your core CPC bids by some percentage, for ad viewers associated with that geo segment). That wasn’t hard, was it?

(Note: Bing Ads offers essentially parallel functionality.)

I consider this screen to represent the core of any AdWords geo-bidding strategy today; it’s available on the Locations tab as one of three prominent bid adjustment opportunities (alongside mobile bid adjustments and ad scheduling).

Somehow this powerful, basic functionality gets lost in the shuffle of PowerPoint decks seemingly bent on adding complexity for its own sake.

One conference presentation I recently reviewed finally got to this screen on slide 34 of a 40-slide presentation. By this time, attendees were no doubt visualizing the complimentary samosas, fancy ketchup and drinks available at the cocktail reception, convinced they could never handle all the complexity of data manipulation required to be “good at geo.”

Worse, little to nothing was said about this important geo-bidding screen. What should you set up, and why?

What To Set Up, And Why? No One Is Really Saying.

The above AdWords screen, obviously, is no great secret. The question is, can you be doing more with it? Can you be doing that more effectively? And are people in Phoenix as stingy as I say they are? And if Pittsburgh is such a goldmine for my clients, why the heck can’t Pittsburgh be bigger? Do special features like demographics, places of interest, colleges, central commercial areas and ZIP codes really help? When?

Another question that nags at a lot of marketers is this: What if I do it improperly? Won’t things get worse?

Faced with a lack of resources and no clear methodology to manage from, many avoid the task entirely. Or they’ll throw a few (or all 50) US states into the mix, then abandon the effort.

I’m convinced there is a solid lift to be had from geo-bidding accurately. But so far, few of us in the industry have produced usable case studies to show clearly what kind of lift geo-bidding is capable of.

Given the difficulty of A/B testing campaign-level settings, most case studies would have to be taken with a grain of salt anyway.

Beyond a certain point, there is only wheel-spinning, busywork and regression to the mean, as with so many other marketing boondoggles.

Experiment. There’s Limited Risk.

It’s worth asking — to channel Taleb once more — is this something you’d fuss with if you had real skin in the game? Not as a technocrat, but as a business owner?

In the financial, medical and environmental realms, there are awful consequences to “blowing up,” even if blowups or meltdowns are rare (black swans). Yet fast talkers and advocates of shiny new things pursue slight gains too ardently. Taleb refers to that as “the convexity of risk.”

In AdWords, you don’t get quite the same potential for catastrophe. So if you can tune out the overly eager purveyors of shiny objects, use some of your spare time to tinker. You may find you can create incremental, reliable lift without endless effort.

The Cool Thing: Geo Data Gets Beyond Black-And-White Thinking

Many marketers aren’t aware of the power of the data we have at our fingertips today, and how easy it is to tap that power.

Last year, we worked with a client who told us to exclude a number of states from their financial lead generation effort because 30 years of direct mail had taught them those areas don’t perform for them.

We are generally against putting the cart before the horse in this way. If René Descartes himself sat us down and told us that logic dictates we should shut down potential goldmine states in favor of a highly convoluted and unproven theory about how certain personas might come up with a search query, we’d introduce him to David Hume and a pitcher of ale and return when he’d come to his senses.

Basic Principles: What Are We Trying To Accomplish?

To put the exercise on a solid foundation, consider the following basic principles:

• The purpose of geo bid adjustments is to maximize PPC campaign volume and/or ROI by bidding accurately on some configuration of different geographic areas (a number of states and metros within the US, for example).
• This “bid adjustments” functionality, called Enhanced Campaigns when first rolled out by Google, is a powerful advancement over the old methodology where you had to set up a separate campaign for every geo-specific bid strategy you wanted to deploy. Granted, bid adjustments may not be the only geo-strategy in many accounts. But for many PPC accounts — possibly the majority — it’s a great time-saver to lean more on bid adjustments and less on elaborate account structures, geo-specific ad copy and so forth.
• At a certain point, you must accept that complex stories are irrelevant to this exercise because of what it entails: many of the resulting actions will be small bid adjustments of less than 10 percent. In much rarer cases, those adjustments may be 20 percent, 40 percent, or all the way up to 100+ percent if you are looking at a highly localized type of business. In all cases, what you are doing is fiddling with bids. It’s that simple. Hearing an elaborate story about neighborhoods and personas does zero to alter the course of events. Just normalize each segment to hit your target KPI on all of them.
• Geography is not being used as a bid factor for its inherent characteristics, presumably, but because it is a good enough proxy for propensity to purchase. That propensity doesn’t derive solely from income, but from a mix of demographic and cultural characteristics, including the nature of employment or common pastimes.
• For simplicity’s sake, it’s worth remembering that we are essentially on the lookout for differing conversion rates (though you can opt to manage to ROAS, CPA or whatever you like, of course). A greater search query volume, because “people like salty snacks in this region,” doesn’t necessarily translate into more dollars to the business, since we’re paying for clicks.
• The behavior of the segments has to be significantly patterned in a manner distinct from just random data fluctuations to be worth adjusting your bids to. Put another way, long-term patterns that are distinct enough from the mean to build up a high statistical confidence level warrant attention. Stuff that just bounces around short-term but results in regression to the mean shouldn’t be “chased” — at best, you’re getting no farther ahead; at worst, you do even worse than if you had not managed it at all.
• Following from that, I’ll save you some time: If you’re getting excited about how to best market to a bazillion ZIP codes, keep randomness and statistical confidence in mind. Maybe don’t bother unless you’re very advanced and have a very large account.
• Behavior will vary from industry to industry, from account to account and from campaign to campaign.

Slightly More Advanced Principles

Now consider the following slightly more advanced principles:

• A thematically organized account may help, as long as the resulting campaigns are large enough. Poorly organized accounts — say, accounts that have unnecessarily large campaigns — may wind up “blending” interesting behaviors to a less interesting aggregate. This can mask interesting behaviors that break down by, say, type of product.
• This stuff isn’t all that easy or common to automate, but there’s no doubt you can and should automate it, past a certain point of time being wasted.
• Big-city dwellers may exhibit behavior that meaningfully departs from the “hinterland” (rest of the state); in this case, managing a combination of cities, metros and states may be important.
• For your particular account, combined with your unique insights into how certain parts of the world “tick,” you might be able to hit on some clever approaches to geo-bidding in highly populated areas. Get creative. Pick a suburb you know, and add it as a geo-segment, along with the DMA and the state.

Don’t worry about data, though, if you don’t add a given segment. You can look up the past data at any level using “View Location Reports,” available from the same AdWords screen. That might be an awfully important place to research your strategy!

Sunset Superfluous Stuff

Finally, after a considerable period of time, consider sunsetting the pieces of your geo-bidding edifice that do nothing to further the cause. If a state or metro area is going to regress to the mean (for, say, the whole United States), then maybe you’re better off admitting that it isn’t interesting to enough to manage separately.

If you no longer want to manage a geo segment separately, simply “remove” it (not the same as excluding it). The “unmanaged” segments now simply pool in with the catch-all (e.g., “United States”) and should — if you’re adept at reading the data and making a solid prediction — make management simpler with no loss in performance.

With the time you save, you might just be in a good enough mood to pay full price for condiments. Pass the ketchup.